Section 1: Executive Summary

1.1 Business Overview

Agentic AI Ltd. is an Israeli technology company building Numi — the world’s first agentic AI baby sleep consultant. Numi transforms Dorit Kreiser’s proven sleep methodology (25+ years, over 3,000 families, published book) into an autonomous AI agent that coaches parents through sleep training 24/7 via WhatsApp.

Unlike existing sleep apps that passively track data or reactively answer questions, Numi proactively manages the entire sleep training process: building personalized plans, running real-time countdown timers at 2AM, detecting mistakes, and coordinating all caregivers in the home.

Detail	Information
Legal Name	Agentic AI Ltd.
Product	Numi — AI baby sleep consultant
Headquarters	Israel
Stage	Active POC, pre-revenue. Company incorporation: March 2026.
Founders	Rotem Levi (CTO) & Dorit Kreiser (Methodology & Marketing Lead)
Ownership	50/50 equity, no external investors
Investment	$50,000 owner’s loan (0% interest, priority repayment)
Revenue model	$399/year per child (Israel). Annual subscription.
Team	2 co-founders + 1 full-time engineer from Day 0
Technology	Google ADK, Gemini Pro + Flash, RAG, WhatsApp Business API

---

1.2 Unique Value Proposition (UVP)

Numi is the only AI in the world that proactively manages baby sleep training — built on a proven clinical methodology, delivered via WhatsApp, available 24/7.

The problem: 37% of children aged 4 months to 5 years don’t get enough sleep (CDC, 2025). Human sleep consultants cost $641-962 per package and aren’t available at 2AM when the crisis peaks. Generic sleep apps answer questions but don’t manage the process.

The solution: Numi is an agentic AI — not a chatbot. It initiates contact, manages real-time check-in timers (“Dorit’s Clock”), tracks daily progress, detects mistakes before they derail the plan, and coordinates all caregivers (mom, dad, grandma, nanny) through the same training program.

Capability	Generic Apps	Human Consultant	Numi
Available at 2AM	❌	❌	✅
Proactive coaching	❌	Limited	✅
Real-time timer management	❌	❌	✅
Multi-caregiver coordination	❌	❌	✅
Proven methodology	❌	Varies	✅ (Dorit Kreiser)
Annual cost	$120	$641-962	$399
Sibling coordination	❌	Limited	Planned
Hebrew + WhatsApp	❌	✅	✅

---

1.3 Target Market

Israel → UK → US expansion strategy.

Market	Annual Births	SAM (method-adjusted)	Year of Entry
Israel	181,609 (CBS Israel via Macrotrends, 2024)	~$6.1M/year	Year 1 (2026)
UK	594,677 (ONS, 2024)	~$25.4M/year	Year 2 (2027)
US	3,628,934 (CDC, 2024)	~$221M/year	Year 3 (2028)

Why Israel first: Highest OECD fertility rate (2.91 children/woman), 99% WhatsApp penetration, zero AI sleep competitors in Hebrew, and Dorit’s built-in audience (14K Instagram followers, book readers, active client base). Small enough to dominate quickly; large enough to prove the model.

The global baby sleep coaching market was valued at $1.14 billion in 2024 and is projected to reach $2.44 billion by 2033 at a CAGR of 8.7% (Growth Market Reports, 2025).

---

1.4 Business Goals & Objectives

Year 1 (April 2026 – March 2027): Product-Market Fit in Israel

Goal	Target	How We Know
MVP built and launched	November 2026	Operational WhatsApp product with paying customers
First paying families	120 families (base case)	Revenue covers monthly operating costs
AI accuracy	≥90% (Dorit-validated)	9 out of 10 responses methodologically correct
Gate 1 passed	30 families, revenue ≥ costs, <15% refund rate	6 months from launch

Year 2 (April 2027 – March 2028): Growth + UK Expansion

Goal	Target	How We Know
Israel + UK families	400 (base case)	Growing across two markets
Annual revenue	$190,000 (base case)	Company is profitable
Gate 2 passed	120 families, $48K ARR, <15% churn, NPS ≥ 50	12 months from launch
$50K loan repayment	Underway (50% of monthly profit)	Monthly reporting to Dorit

Year 3 (April 2028 – March 2029): Scale + US Entry

Goal	Target	How We Know
Total families	1,000 (base case)	Three-market operation
Annual revenue	$550,000 (base case)	Strong unit economics at scale
Loan fully repaid	✅ (base: Month 33)	$0 remaining balance
Profitability	$320,000 net income (base)	Sustainable and growing

---

1.5 Financial Highlights

Metric	Value
Total investment	$50,000 (~156,000 ILS) — owner’s loan from Dorit, 0% interest
Monthly operating cost	$3,000 (one of the lowest in the startup world)
Break-even	~100 paying families (0.05% of annual Israeli births)
Gross margin	90.5% per primary child (even with Gemini Pro for AI quality)
Year 1 revenue (base)	$48,000*
Year 3 revenue (base)	$550,000
Year 3 net income (base)	$320,000
3-year cumulative profit (base)	$398,000
Cash profit ROI on $50K (base, 3 years)	8.0x
Loan repayment	Month 33 (base) / Month 24 (optimistic)
LTV:CAC ratio	9.6-28.4x (target: >3x is healthy)
Zero-revenue runway	12 months operational (after $11K one-time setup from the same $50K)

*Year 1 revenue assumes $399/year standard pricing. First 50 founding members pay $299/year as a launch incentive (see Section 6.5), which would reduce actual Year 1 revenue by ~$5,000 to ~$43,000. This discount is treated as a one-time marketing cost, not reflected in projections.

Three-Year Scenario Summary

Metric	Conservative	Base	Optimistic
Year 3 families	500	1,000	1,500
Year 3 revenue	$270K	$550K	$850K
Year 3 net income	$120K	$320K	$530K
3-year cumulative	$115K	$398K	$758K
Loan repaid?	Not within 3 years	Month 33	Month 24
Cash profit ROI on $50K	2.3x	8.0x	15.2x

Conservative scenario honesty: The conservative figures above reflect the multi-market expansion plan. An Israel-only stress test (Section 11) models a more cautious conservative case: 185 families by Year 3, the $50K loan NOT repaid (~$46K remaining), and cumulative losses of -$50K — meaning Dorit’s full investment is at risk. The gate system (Section 8.3) provides early warning at Month 14 and Month 20, allowing both founders to course-correct or wind down before full capital depletion. See Section 9.10 (“What Dorit Should Expect”) and Section 11 for the detailed analysis.

---

1.6 Call to Action

Agentic AI is seeking $50,000 in founder capital to build and launch Numi — the world’s first agentic AI baby sleep consultant. This capital funds a 6-month pre-revenue MVP build and 7+ months of post-launch growth ($11K one-time setup + $3K/month operations = 12 months of zero-revenue runway).

Why invest now:

1. The market is ready. AI is entering parenting (Huckleberry Berry, Nanit’s $50M raise) — but nobody is building an agentic solution. The window is open.
2. The methodology exists. Dorit’s method is proven and ready to encode (see Section 2.5). The AI doesn’t need to invent — it needs to scale.
3. The technology works. A working POC demonstrates that the multi-agent architecture successfully conducts professional sleep consultations using Dorit’s methodology.
4. The audience is waiting. Dorit’s 14K Instagram followers, book readers, and active client base provide customers from month 1. This is not a cold start.
5. The economics are exceptional. ~90% gross margins, $3,000/month costs, break-even at 100 families, and LTV:CAC ratios of 9.6-28.4x.
6. The risk is contained. Maximum loss: $50K. Even in total failure, company IP retains $55-120K in residual value. Both founders draw $0 salary. Every dollar goes to product and growth.

The bottom line: For $50,000, Dorit gets 50% ownership in a company targeting $550K revenue and $320K profit by Year 3 — a potential 8.0x cash return on her $50K investment (base case, 3 years). If the company reaches scale and attracts acquisition interest, the equity upside could be significantly higher. The loan is repaid first, before any dividends or salaries. Monthly reporting provides full transparency. And the product is built on her life’s work.

---

Section 3: Product & Technology

3.1 Product Description

Numi is an agentic AI baby sleep consultant — one product, built on Dorit Kreiser’s proven methodology, delivered via WhatsApp today and a native application in the future.

Unlike existing sleep apps that provide generic tips or passive tracking, Numi is an autonomous agent — a system that doesn’t just answer questions, but initiates action, manages the sleep training process end-to-end, and accompanies parents through the hardest nights of their lives.

The Fundamental Difference: Agentic vs. Reactive

Capability	Competitors (Reactive)	Numi (Agentic)
When it operates	Only when the parent asks a question	Initiates contact proactively — even when the parent doesn’t reach out
Type of help	Answer to a single question	Continuous accompaniment: reminders, encouragement, alerts
Process management	Parent manages the process themselves	Numi manages the sleep training process end-to-end
Problem detection	Responds after the parent reports a problem	Identifies patterns and alerts before a problem occurs
Real-time timer	None	Manages “Dorit’s Clock” — a fixed-interval check-in timer with strategic withdrawal during training nights
Cross-analysis	Looks at sleep data in isolation	Connects nutrition, sleep, development, and siblings

In simple terms: Competitor apps are like a book you open when you have a question. Numi is like a consultant sitting beside you, watching the baby, and telling you what to do — before you even knew there was a problem.

The Professional Foundation

All of Numi’s knowledge is built on Dorit Kreiser’s methodology from her book “Shhhh… At Night We Sleep” and her 25+ years of clinical experience. The AI doesn’t generate generic advice from the internet — it retrieves answers from Dorit’s specific, proven method using a RAG (Retrieval Augmented Generation) architecture. Like asking Dorit herself, except she’s available 24/7.

Who Is Numi For?

Numi is purpose-built for parents who are ready to make a structural change to their child’s sleep. Dorit’s method uses a structured check-in protocol: parents enter the room at fixed intervals when the baby cries, offer brief reassurance (touch, “Shhh”), and withdraw when the crying tone drops — signaling the baby is learning to self-soothe. This is not “cry it out” (CIO) — parents don’t leave the baby alone indefinitely — but it does involve tolerating some crying during the 2-3 day process.

This means Numi is ideal for: - Parents experiencing a sleep crisis (baby waking 4-8 times per night) - Parents open to a structured, time-limited intervention (typically 2-3 days) - Parents willing to follow a specific protocol consistently - Families where all caregivers (both parents, grandparents, nannies) can commit to the same approach

Numi is not the right fit for: - Parents who are philosophically committed to “no-cry” or attachment-only approaches - Parents seeking co-sleeping support or bed-sharing guidance - Families looking for a general parenting app rather than targeted sleep training

Research indicates that approximately 55-65% of parents experiencing infant sleep problems are open to structured sleep training methods (Mindell et al., 2022; Honaker & Meltzer, 2016). This self-selection is reflected in our SAM/SOM calculations (see Section 4.2) and our intake agent’s readiness assessment — which screens for commitment level before onboarding a family, reducing refund rates and improving outcomes.

Platform Strategy

Numi starts on WhatsApp — where 99% of Israeli parents already communicate (JPost, 2025). No new app to download. Zero friction. As the product matures and the user base grows, Numi will evolve into a native application with richer features (dashboards, charts, historical views), while maintaining WhatsApp as the primary real-time communication channel.

The multi-agent system that powers Numi is a headless architecture — it serves only the Numi application. Any other usage of the agent system (white-labeling, licensing to third parties, integration with other products) requires a separate commercial agreement.

---

3.2 Multi-Agent Architecture

The Client Journey — Not Parallel Agents, But a Flow

Numi doesn’t use four equal agents running in parallel. It uses a journey-based multi-agent system where specialized agents handle different phases of the parent’s experience. The system is built on Google ADK (Agent Development Kit) and uses cost-optimized model selection — each agent runs on the cheapest model that can do its job well.

+------------------------------------------------------------------+
|                     Root Agent (Router)                            |
|              gemini-3-flash (cheapest model)                       |
|         Routes conversations to the right specialist              |
+------------------------------------------------------------------+
         |            |            |           |          |
   +-----+----+ +----+-----+ +----+----+ +----+---+ +----+-----+
   | INTAKE   | |CONSULTANT| |AFTERCARE| | LOGGER | | BOOKING  |
   | Agent    | |  Agent   | |  Agent  | |  Agent | |  Agent   |
   |          | |          | |         | |        | |          |
   | Collects | | THE BRAIN| | Daily   | | Logs   | | Escalate |
   | baby     | | Builds   | | check-  | | eat,   | | to human |
   | profile, | | sleep    | | ins,    | | sleep, | | Dorit    |
   | sleep    | | plan     | | tracks  | | poop,  | |          |
   | context, | | using    | | progress| | pee    | |          |
   | readiness| | Dorit's  | | detects | |        | |          |
   | check    | | RAG      | | mistakes| |        | |          |
   |          | |          | |         | |        | |          |
   | flash    | | PRO      | | flash   | | flash  | | flash    |
   | (cheap)  | | (smart)  | | (cheap) | |(cheap) | | (cheap)  |
   +----------+ +----------+ +---------+ +--------+ +----------+

Why This Architecture Is Smart

Cost optimization through model selection: Only the Consultant agent — which needs deep methodology reasoning, RAG retrieval, and the ability to build personalized sleep plans — runs on Gemini 3.0 Pro (the expensive model). All other agents handle structured tasks (collecting data, logging events, checking in) that work perfectly on Gemini Flash (the cheapest model). This means ~70% of all AI interactions use the cheapest available model.

The journey flow:

Phase	Agent	Model	What Happens
1. First contact	Intake Agent	Flash (cheap)	Collects baby profile: name, age, weight, sleep habits, feeding type. Questions are age-gated using Dorit’s methodology groups (3-5mo, 6-11mo, 12-24mo, 2-5yr). Checks readiness for sleep training.
2. The plan	Consultant Agent	Pro (smart)	The critical moment. Retrieves Dorit’s methodology via RAG. Builds a personalized, multi-week sleep plan. Explains the “why” behind each instruction. This is where trust is won or lost.
3. Daily follow-up	Aftercare Agent	Flash (cheap)	Collects nightly reports: bedtime, crying duration, wake-ups, what the parent did. Evaluates progress against expected milestones. Detects mistakes (picked up baby, stayed in room too long). Reinforces consistency.
4. Event logging	Logger Agent	Flash (cheap)	Quick logging of eat/sleep/poop/pee events. Builds the data layer for nutrition-sleep correlation analysis.
5. Human escalation	Booking Agent	Flash (cheap)	When the AI can’t solve a case (no improvement after 7 days, medical concern, parental distress) — schedules a paid consultation with Dorit herself.

The Consultant Agent — Numi’s Brain

The Consultant is the only agent that uses Gemini Pro and RAG retrieval. This is deliberate — it’s the highest-stakes interaction:

What it does: 1. Retrieves relevant methodology from Dorit’s encoded knowledge base (RAG) 2. Builds a personalized sleep plan based on the baby’s profile, age group, and specific challenges 3. Explains Dorit’s core principles: breaking sleep dependencies, the “Shhhh…” method, bedtime routine, night feeding weaning 4. Sets expectations: “Days 1-3: crying is normal — it’s the baby’s way of expressing frustration, not distress. By day 3 on average, the baby will find their self-soothing movement and sleep through the night.” 5. Hands off to the Aftercare agent for daily tracking

Why it needs the Pro model: Generic models can parrot sleep advice from the internet. Pro + RAG faithfully applies Dorit’s specific method — including her unique fixed-interval check-in protocol, the strategic withdrawal when crying tone drops, her strict rules about pacifiers and night feeding, and her warm-but-uncompromising tone. The plan must feel like it came from Dorit, not from a chatbot.

“Dorit’s Clock” — The Signature Feature

Dorit’s methodology uses a unique protocol — not the graduated increasing delays common in other methods (like Ferber). Instead, it follows a fixed-interval check-in with strategic withdrawal:

1. Allow self-settling (~30 minutes): Baby may mumble, move, make sounds — normal. Don’t enter.
2. If baby cries for a full minute → enter quietly (barefoot), touch hand or back, say “Shhh…”, leave immediately. [Note: The exact interval duration follows Dorit’s clinical protocol — to be confirmed with Dorit during RAG encoding.]
3. Continue entering every minute (fixed interval) while baby cries consistently. No limit on entries.
4. Stop entering when: (a) baby stops crying, OR (b) crying tone drops — this signals the baby is finding their “Transition Movement” (self-soothing motion like head rocking, blanket touching, or rhythmic movement).
5. Once stopped or tone dropped — do NOT re-enter, even if crying resumes. This allows the baby to internalize self-soothing.
6. Results in ~3 days on average.

Most parents struggle to execute this at 2:00 AM when they’re exhausted and the baby is crying — especially the critical judgment of when to stop entering.

Numi manages the entire protocol via WhatsApp: counts the one-minute intervals, sends encouragement messages, listens for the parent’s description of crying changes, and tells them exactly when to go in and — crucially — when to stop going in. It documents every wake-up and tracks progress across training nights.

Real example: 02:14 AM. Baby wakes up crying. Mom messages Numi: “He woke up again.” Numi responds in seconds: “Night 2, second wake-up. Wait one full minute. I’m timing it.” After 60 seconds: “Go in now. Touch his back, say Shhh, leave. Let me know when you’re out.” Mom: “I’m out, he’s still crying.” Numi: “Starting next minute. Stay strong.” After several check-ins: Mom: “He’s quieter now, kind of whimpering.” Numi: “That’s the tone drop — he’s finding his self-soothing movement. Don’t go in again. He’s learning. I’m here if anything changes.”

Future Agent Capabilities

The architecture is designed to grow. These capabilities will be added as the product matures:

Capability	How It Works	When
Nutrition correlation	Logger data (feeding events) is analyzed against sleep patterns to detect correlations. “He wakes at 2AM every night he eats late. Let’s move dinner earlier.”	Q4 2026
Siblings coordination	Multi-child profiles share scheduling data. “Your 3-year-old naps at 2PM. Put the baby down at 1:45 so she’s asleep before he gets home.”	Q1 2027
Long-term memory	Persistent family history across months and children. “Last time this regression happened (4 months ago), method X worked in 3 nights. Let’s try it again.”	Q1 2027
Regression prediction	Age-based pattern detection. “Your baby turns 8 months in 2 weeks. A sleep regression is common at this age. Here’s what to prepare.”	Q1 2027
Cry detection via microphone	AI-powered acoustic analysis of baby crying through the phone’s microphone. Detects crying tone changes (intensity, frequency, pitch drop) to automatically determine when to stop check-ins — removing the subjective judgment from exhausted parents. Technology proven viable: Zoundream (92% accuracy, €4M seed), Ubenwa (cry analysis foundation model), ChatterBaby (UCLA).	2027+

---

3.3 The Seats System — Multi-User, Multi-Child Architecture

Seats: Every Caregiver, Their Own Numi

Every person caring for the child gets their own WhatsApp connection to Numi. The AI adapts its behavior — and its agent selection — based on who it’s talking to:

Seat Role	How Numi Behaves	Primary Agents Used	Model Cost
Primary Parent	Full coaching: plan building, “Dorit’s Clock” at 2AM, emotional support, deep methodology guidance	Intake → Consultant (Pro) → Aftercare	Higher
Co-Parent	Coordinated support: “Mom entered 30 seconds ago. He’s still crying. Wait for the tone to drop before we decide. Don’t go in.”	Aftercare (Flash)	Lower
Grandmother	Respectful but firm guidance: “I know it’s hard to hear him cry. We’re waiting for the tone to change. You’re doing the right thing. Here’s why.”	Aftercare (Flash)	Lower
Nanny / Au Pair	Operational mode: clear instructions, no theory. “Nap at 2:00 PM. Put down awake. If crying for 1 minute, go in, touch, say Shhh, leave. Log when asleep.”	Aftercare + Logger (Flash)	Lowest
Night Nanny	Night protocol: “Tonight’s plan: check every minute while crying. Stop when tone drops. Log every wake-up. Alert parents if crying exceeds 20 minutes.”	Aftercare + Logger (Flash)	Lowest

Cost insight: Additional seats primarily use cheap Flash agents (Aftercare, Logger). They don’t need the Pro Consultant — the sleep plan is already built for the primary parent. This makes seats a very high-margin add-on: ~$0.75/month variable cost vs. $59/year ($4.92/month) revenue = 85% gross margin.

Why this matters: The #1 reason sleep training fails is inconsistency between caregivers. Mom follows the plan, then grandma picks up the baby after 30 seconds of crying — undoing three nights of progress. Numi Seats ensures every caregiver gets the same plan, adapted to their role. No competitor does this. Not even human consultants can simultaneously coach mom, dad, and grandma at 2 AM. Numi can.

Multiple Children: Per-Child Profiles

Each child is a separate subscription with its own: - Baby profile (age-gated by Dorit’s methodology groups) - Sleep plan (built by the Consultant agent with Pro) - Aftercare tracking (progress, milestones, reports) - Event log (eat, sleep, poop, pee)

All children share the same family account. Existing seats carry over — grandma’s WhatsApp works for all children in the family.

Cost per additional child: ~$3.16/month (needs its own Pro Consultant session) vs. $299/year ($24.92/month) revenue = 87% gross margin.

Why this is a competitive moat in Israel: With 2.9 children per family (highest in OECD), per-child pricing creates growing revenue per household. And siblings coordination (planned) will be a feature no competitor offers — managing sleep across multiple children in the same home.

Variable Cost by Product Type

Product	Monthly Variable Cost	Annual Revenue	Gross Margin
Primary child (Pro Consultant + all agents)	$3.16	$399	90.5%
Additional seat (Flash agents only)	$0.75	$59	84.7%
Additional child (own Pro Consultant)	$3.16	$299	87.3%
Family bundle (2 children)	$6.32	$599	87.3%

Each seat is a separate WhatsApp conversation. All data feeds into one unified child profile. All margins exceed 84%.

---

3.4 Technology Stack

Layer	Technology	Why We Chose It	Cost
Agent Orchestration	Google ADK (Agent Development Kit)	Open-source, purpose-built for multi-agent systems. Built-in agent routing, session state, tool invocation. Apache 2.0 license — free.	$0
Consultant AI	Gemini 3.0 Pro	Deep reasoning for sleep plan building, RAG synthesis, and methodology application. The only agent that needs this level. ($2/1M input, $12/1M output).	~$2.70/customer/month
All Other Agents	Gemini 3 Flash	Fast, cheap, handles structured tasks: data collection, logging, check-ins, routing. ($0.50/1M input, $3/1M output).	~$0.33/customer/month
Knowledge Base (RAG)	Vertex AI RAG + ChromaDB	Dorit’s book, methodology, and case files encoded as embeddings for semantic retrieval. Dual backend: Vertex AI for production, ChromaDB for development.	~$0.01/customer/month
Communication	WhatsApp Business API	99% open rates in Israel. Parents already use it. No new app needed. Service messages (customer-initiated) are free. Business-initiated utility messages: $0.01/msg in Israel.	~$0.06/customer/month
Database	Google Cloud Firestore	Serverless NoSQL. Free tier: 50K reads/day. Baby profiles, event logs, session state.	$1-10/month
Hosting	Google Cloud Run	Serverless — pay only for what you use. Free tier: 2M requests/month.	$5-30/month
Native App (future)	React Native or Flutter	Cross-platform. Dashboards, charts, historical views. WhatsApp remains the real-time channel.	Phase 2+
Security	AES-256 encryption, TLS 1.3	Full protection of sensitive baby and family data. GDPR-ready.	Included in GCP

Total variable cost per primary customer: ~$3.16/month ($2.70 Pro + $0.33 Flash + $0.01 RAG + $0.06 WhatsApp + ~$0.06 buffer for token overages). Driven primarily by Gemini Pro for the Consultant agent. See Section 9 for detailed cost modeling.

Cost Advantage: Google for Startups Cloud Program

Agentic AI plans to apply to the Google for Startups Cloud Program, which offers: - Up to $200,000 in Google Cloud credits (standard) - Up to $350,000 for AI-first startups (Numi qualifies — AI is the product, not a bolt-on) - Credits cover: Cloud Run, Firestore, Gemini API, Vertex AI, and all GCP services

If accepted, cloud costs drop to effectively $0 for Year 1-2. This is not factored into our financial projections (worst-case assumes $0 credits), but represents significant upside.

Current POC Status

The proof of concept is operational today on Google ADK and demonstrates:

Capability	Status	Details
Multi-agent flow	✅ Working	Root → Intake → Consultant → Aftercare → Logger → Booking
RAG retrieval	✅ Working	Dual backend (Vertex AI + ChromaDB). Retrieves Dorit’s methodology accurately.
Age-gated baby profiles	✅ Working	Adaptive questioning by age group (3-5mo, 6-11mo, 12-24mo, 2-5yr) per Dorit’s methodology
Multi-language support	✅ Working	Hebrew and English with Jinja2 templates + locale system
Baby event logging	✅ Working	Eat, sleep, poop, pee events with timestamps
Aftercare daily reports	✅ Working	Structured nightly report collection with progress evaluation
Response time	✅ < 3 seconds	Target: < 2 seconds
WhatsApp integration	✅ Working	Conversations via WhatsApp Business API

What the POC proves: The core agentic architecture works — Numi can conduct a professional sleep consultation based on Dorit’s methodology, from intake through plan delivery to daily follow-up. The MVP builds on this proven foundation.

---

3.5 Product Roadmap

Phase	Timeline	Features	Status
POC	2025 Q3-Q4	Multi-agent flow (Intake → Consultant → Aftercare → Logger → Booking), RAG from Dorit’s book, WhatsApp integration, age-gated profiles, multi-language	✅ Complete
MVP	2026 Q2-Q3	“Dorit’s Clock” real-time check-in timer, multi-seat support (role-adapted AI), payment integration, per-child profiles, Gemini Pro for Consultant	🔨 Next
Enhanced	2026 Q4	Nutrition-sleep correlation (from Logger data), sleep logging dashboards, weekly progress reports	Planned
Advanced	2027 Q1	Long-term memory, regression prediction, siblings coordination, cross-child learning	Planned
International	2027 Q2-Q3	English localization (UK), native app (React Native/Flutter), premium analytics, cry detection via microphone (AI acoustic analysis)	Planned (requires Gate 2)
Scale	2027 Q4+	US market, B2B partnerships (HMOs, pediatric clinics), multi-language expansion	Future

---

3.6 Compliance & Security

Requirement	How Numi Addresses It
Israeli Privacy Protection Law	All personal data stored and processed in compliance with Israeli privacy regulations. Privacy policy and terms of service reviewed by attorney.
GDPR (for UK expansion)	Data minimization, right to deletion, explicit consent for data processing. Required before UK launch.
Data encryption	AES-256 at rest, TLS 1.3 in transit. All WhatsApp communication is end-to-end encrypted by default.
Health disclaimer	Numi provides sleep coaching guidance, not medical advice. Clear disclaimers in onboarding and throughout the experience. Parents advised to consult a pediatrician for medical concerns.
Data ownership	All data belongs to the company (Agentic AI Ltd.). Users can request data export or deletion at any time.
Access control	Each seat has role-based permissions. Caregivers can log data and receive instructions but cannot modify the sleep plan or access billing.

---

End of Section 3 — Product & Technology

---

End of Sections 1-3

Section 4: Industry & Market Analysis

4.1 Industry Overview

The baby sleep consulting industry sits at the intersection of three converging trends: the growing parental awareness of infant sleep health, the explosion of AI-powered digital health tools, and the shift toward subscription-based parenting services.

The global baby sleep coaching market was valued at $1.14 billion in 2024 and is projected to reach $2.44 billion by 2033, growing at a CAGR of 8.7% (Growth Market Reports, 2025). The broader parenting apps market is growing even faster — from $1.06 billion in 2025 to $5.50 billion by 2034, at a CAGR of 20.37% (Business Research Insights, 2025).

Why the market is growing:

Driver	Data Point	Source
Sleep problems are widespread	37% of children aged 4 months to 5 years aren’t getting enough sleep	CDC, January 2025
Parents are spending on solutions	55% of active parenting app users prefer tiered subscription models with sleep coaching	Business Research Insights, 2025
AI increases engagement	AI-powered guidance led to 50% increase in user engagement in parenting apps	Business Research Insights, 2025
Digital adoption is high	60% of urban new parents use baby-tracker apps	Business Research Insights, 2025
Investors see the opportunity	Nanit raised $50M in Dec 2025; Huckleberry raised ~$14M+ in VC funding; Owlet ($128M revenue) entering AI coaching	PressReleases, 2025-2026

What’s missing in the market: Despite billions in market value and growing demand, no product on the market — globally — offers an agentic AI that proactively manages a baby’s sleep training process. All existing solutions are either passive tracking (Nanit), reactive chatbots (Huckleberry’s Berry), or static content (books, courses). Numi fills this gap.

---

4.2 Target Market Definition (TAM/SAM/SOM)

TAM — Total Addressable Market

Definition: If every parent globally who experiences baby sleep problems used an AI sleep consulting service, how big would the market be?

Component	Number	Source
Global births per year	~140 million	World Bank, 2024
% experiencing sleep problems	~30-37%	CDC 2025, PMC studies
Parents with sleep problems	~42-52 million/year	Calculated
Average willingness to pay	$200-600/year	Based on current market pricing (Huckleberry $120/yr to Nanit $300-600/yr)
TAM	$8.4 - $31.2 billion/year	Calculated

More conservatively, using only the baby sleep coaching guides market: $1.14 billion (2024) growing to $2.44 billion (2033) (Growth Market Reports).

We use the conservative figure throughout this plan.

---

SAM — Serviceable Available Market

Definition: The portion of the TAM that Numi can realistically reach given our geographic focus, cultural adaptation, and distribution strategy.

Important note on expansion barriers: The real barrier to international expansion is cultural adaptation, not language. Sleep training philosophies vary across cultures — Israeli parents are receptive to structured methods, while UK parents lean toward gentler approaches. Each market requires adapting methodology framing, tone, and marketing. This is why expansion is phased.

Numi’s expansion plan is Israel (Year 1) → UK (Year 2) → US (Year 3). Here’s the SAM calculation for each market:

Israel SAM:

Component	Number	Source
Annual births	181,609 (2024)	CBS Israel via Macrotrends
% experiencing sleep problems	37%	CDC, based on NSCH 2020-2021 data — US figure applied as proxy (no Israeli-specific study available; Dorit’s clinical experience confirms comparable rates in Israel)
Parents seeking help (subset of those with problems)	~50% of those with problems	Conservative estimate
WhatsApp penetration	99%	JPost, 2025
Hebrew-speaking (target)	~76% (Jewish births)	CBS Israel
Addressable parents (before method filter)	~25,500	181,609 x 37% x 50% x 76%
Method-willingness filter	~60%	~55-65% of parents with sleep problems are open to structured training methods (Mindell et al., 2022; Honaker & Meltzer, 2016). We use 60% — the midpoint. Israeli culture is generally receptive to structured approaches, but no Israel-specific survey exists.
Israel SAM (parents/year)	~15,300	25,500 x 60%
Israel SAM (revenue at $399/yr)	~6.1 million/year	Calculated

UK SAM:

Component	Number	Source
Annual births	594,677 (2024)	ONS
% experiencing sleep problems	30%	Conservative estimate for UK
Parents seeking help	~50%	Conservative
Addressable parents (before method filter)	~89,200	Calculated
Method-willingness filter	~50%	UK leans toward gentler approaches; the NHS does not recommend leaving babies to cry. Lower willingness than Israel.
UK SAM (parents/year)	~44,600	89,200 x 50%
UK SAM (revenue at £449/yr ≈ $569)	~$25.4 million/year	Calculated

US SAM:

Component	Number	Source
Annual births	3,628,934 (2024)	CDC
% experiencing sleep problems	37%	CDC 2025
Parents seeking help	~50%	Conservative
Addressable parents (before method filter)	~671,350	Calculated
Method-willingness filter	~55%	US is polarized: growing co-sleeping movement (46% of parents, AASM 2024), but also strong sleep training culture. 55% is conservative midpoint.
US SAM (parents/year)	~369,200	671,350 x 55%
US SAM (revenue at $599/yr)	~$221 million/year	Calculated

Combined 3-market SAM: ~$253 million/year

Why we apply a method-willingness filter: Dorit’s methodology is a structured check-in protocol that involves some crying (see Section 3.1, “Who Is Numi For?”). Research consistently shows that ~35-45% of parents are philosophically opposed to any form of structured sleep training or will only accept “no-cry” approaches (Mindell et al., 2022; Blunden & Baills, 2013). Excluding these families gives a more honest and defensible SAM. Future product evolution (gentler protocol options, partnerships with other methodologists) could expand the addressable market back toward the unfiltered figures.

---

SOM — Serviceable Obtainable Market

Definition: What Numi can realistically capture in Years 1-3, given our team size, marketing budget, and growth rate.

Market	Year	Target Families	Revenue	% of SAM	How We Get There
Israel	Year 1	50-120	$15K-$48K	0.2-0.5%	Dorit’s audience, organic social, word of mouth
Israel	Year 2	200-350	$80K-$140K	0.8-1.4%	Continued organic + small paid marketing budget
Israel + UK	Year 2	300-500	$130K-$250K	—	UK launch adds 100-150 families
Israel + UK + US	Year 3	600-1,200	$300K-$650K	—	US entry, growing all markets

Note: All SOM revenue figures are gross revenue (total subscription income), not profit. At 90% gross margin, Year 3 gross profit on $300K-$650K revenue would be approximately $273K-$592K. After fixed costs ($36K/year) and marketing, net profit would be approximately $200K-$500K.

Why these numbers are conservative: Year 1 target of 120 families is just 0.8% of the Israel SAM. Dorit’s 14K Instagram followers alone — converting 1% = 140 families, exceeding the target. SOM assumes minimal marketing spend and organic growth only.

---

4.3 Market Trends & Demand

Trend 1: Parents Increasingly Turn to Digital Solutions

The shift from human consultants to digital tools is accelerating. 60% of urban new parents already use baby-tracking apps (Business Research Insights, 2025). The COVID-19 pandemic permanently changed parenting behavior — telehealth and digital coaching became the norm, and parents never went back.

Trend 2: AI Is Entering Parenting — But Nobody Is Agentic Yet

Huckleberry launched Berry (an AI chatbot) in February 2026. Nanit is building a “Parenting Intelligence System.” Multiple new entrants (Luna Sleep, ARIA, Bambii) are emerging. But all current solutions are reactive — they wait for the parent to ask a question. None proactively manages a sleep training process. This is Numi’s window of opportunity.

Trend 3: Israel Is the Perfect Launch Market

Factor	Why It Matters for Numi
Highest OECD fertility rate (2.91)	More babies = more customers. Average 2.9 children per family makes the siblings feature and per-child pricing especially valuable.
99% WhatsApp penetration	No need to convince parents to download a new app. They already live on WhatsApp.
Zero AI sleep competitors in Hebrew	No direct competition. The market is completely unserved.
High willingness to pay for baby services	Israeli parents routinely pay 2,000-3,000 NIS ($641-962) for human sleep consulting.
Cultural norm of involved extended family	Grandmothers, nannies = natural seat upsell. Israeli families are close-knit and hands-on.
Small market = fast validation	182K births/year is small enough to dominate quickly, large enough to prove the model.

Trend 4: The UK Is the Ideal Second Market

Factor	Data
Large English-speaking market	594,677 births/year (ONS, 2024)
Lower marketing costs	Digital CPM ~47% lower than the US
NHS gaps in sleep support	Parents frequently report inadequate sleep advice from the NHS, creating demand for private solutions
Strong WhatsApp adoption	Growing rapidly in the UK for business communications
Stepping stone to US	Proves the English-language product before entering the massive US market

---

4.4 Market Forecasts

Year-by-Year Revenue Projections (Three Scenarios)

Metric	Year 1 (Israel)	Year 2 (Israel + UK)	Year 3 (IL + UK + US)
Conservative
Families	50	200	500
Revenue	$20,000	$90,000	$270,000
Base
Families	120	400	1,000
Revenue	$48,000	$190,000	$550,000
Optimistic
Families	200	700	1,500
Revenue	$85,000	$350,000	$850,000

Assumptions behind these numbers: - Conservative: Only Dorit’s direct referrals, no paid marketing, slow word-of-mouth - Base: Dorit’s audience + organic social + small paid marketing + word-of-mouth - Optimistic: Active marketing budget + partnerships + PR + strong word-of-mouth - Average revenue per family: $399-475/year (base + some seat add-ons) in Israel; £449+ in UK; $599+ in US - Annual churn: 15% (conservative — annual payment model reduces churn significantly)

Key growth drivers by year: - Year 1: Dorit’s existing audience and referrals. Her Instagram (14K followers), book readers, and active client base provide a warm market that requires minimal marketing spend. - Year 2: UK expansion (requires passing Gate 2). English-language product opens a 3x larger market. Israeli growth continues through word-of-mouth and repeat families (second children). - Year 3: US entry. Proven product in two markets. Potential VC fundraise to accelerate growth. Premium US pricing ($599/year) significantly improves revenue per family.

---

Section 6: Sales & Marketing Strategy

6.1 Go-to-Market Strategy

Phase 1: Israel Launch — “Dorit’s Network” (Months 1-6)

Strategy: Leverage Dorit’s existing audience and brand to acquire the first 30-120 families with zero or minimal paid marketing.

Channel	Action	Expected Result
Dorit’s Instagram (14K followers)	Series of posts: “I put my method into AI. Meet Numi.” + stories showing real-time demo	50-100 website visits/post; 1-3% conversion to sign-up
Dorit’s existing clients	Direct WhatsApp outreach to past clients: “Try my AI — founding member price $299”	10-30 founding members in first month
Dorit’s book readers	Insert/sticker in new book copies; QR code to Numi	Ongoing low-volume lead generation
Dorit’s consultant team	Her 4-5 trained consultants recommend Numi to clients who can’t afford human consulting	Steady referral pipeline
Word of mouth	Happy parents tell other parents. Baby sleep is a topic every parent discusses.	Organic growth multiplier
Super Numi (website agent)	AI on Dorit’s website converts 2AM visitors into leads or paying customers	24/7 lead conversion

Marketing spend in Phase 1: ~0-500/month (organic-first strategy)

Phase 2: Israel Growth — “Expand the Circle” (Months 6-12)

Strategy: After Gate 1 (30 families, revenue covers costs), invest in small-scale paid marketing to grow beyond Dorit’s direct network.

Channel	Budget	Expected CAC
Facebook/Instagram ads (Israeli parents)	500-1000/month	30-50/family
Google Ads (Hebrew: “baby doesn’t sleep”, “sleep consultant”)	$300-500/month	40-60/family
Content marketing (blog posts, sleep tips, Dorit’s expertise)	$0 (Dorit creates content)	$0 (organic)
Parenting forums and communities	$0 (organic posts)	$0
Partnerships with baby stores (Shilav, etc.)	Co-marketing	Low

Total Phase 2 marketing: 800-1500/month

Phase 3: UK Launch (After Gate 2, ~Month 12-15)

Strategy: Adapt the proven Israeli playbook for the UK market.

Component	Budget	Notes
Content adaptation (English, UK cultural norms)	3000-5000 (one-time)	NHS references, British English, UK sleep norms
UK sleep expert/advisor consultation	2000-3000 (one-time)	Local credibility and methodology validation
Instagram/Facebook ads (UK parents)	2000-3000/month	Target: new parents, 25-40, UK
Google Ads (UK)	$1,000-2,000/month	“baby sleep help UK”, “sleep consultant”
UK parenting communities (Mumsnet, etc.)	$0 (organic)	Mumsnet is a major parenting forum in the UK
Total UK launch budget: 12500-20000 (from Israeli profits)

---

6.2 Marketing Channels & Budget Allocation

Year 1 (Israel Only)

Channel	Monthly Budget	% of Marketing Spend	Expected CAC	Expected Families/Month
Organic (Dorit’s network)	$0	0%	$0	5-15
Social media ads	$500-1,000	50-60%	$30-50	10-20
Google Ads	$300-500	25-30%	$40-60	5-10
Content/SEO	$0	0%	$0	2-5 (long-term)
Partnerships	$200-500	10-20%	$20-30	3-5
Total	$1,000-2,000	100%	$25-40 blended	15-35

Why marketing spend is so low: Dorit’s existing audience is the primary acquisition channel. Paid marketing is a supplement, not the driver. This is one of the key advantages of building on an existing practice — CAC is dramatically lower than a cold-start.

---

6.3 CAC & LTV Analysis

Metric	Value	How It’s Calculated
Blended CAC	$25-40	Weighted average: $0 for organic (Dorit’s network) + 30-60 for paid channels. ~50% of families come through Dorit organically.
Average revenue per family (Year 1)	$425	$399 base + ~$26 average from 0.44 extra seats per family (see note below)
Gross margin per family	~90%	Variable cost per family: ~$41.88/year ($3.16/month primary + $0.33/month avg seat). Gemini Pro drives the Consultant agent — worth paying for quality.
LTV (1 year, net of variable costs)	$383	$425 - $41.88 variable costs
LTV (2 years, 85% renewal)	$709	$383 + ($383 × 0.85)
LTV (with second child, 30% probability)	$788	$709 + ($299 - $37.92) × 0.30
LTV:CAC ratio	9.6 - 28.4x	$383 / $40 = 9.6x (worst). $709 / $25 = 28.4x (best). Target: >3x is healthy.
Payback period	< 1 month	Annual payment upfront. CAC is recovered immediately at sign-up.

Where does “0.44 extra seats per family” come from? We estimate ~44% of Israeli families will purchase at least one additional seat (e.g., grandmother, nanny). This is a conservative assumption for the Israeli market, where extended family — especially grandmothers — are deeply involved in childcare (Israel has the highest OECD fertility rate at 2.91 children/family, and multi-generational households are common). At $59/seat, 0.44 extra seats × $59 = ~$26 average add-on revenue per family. If the actual attach rate is lower (e.g., 30%), average revenue drops to ~$417/family — still well above break-even.

Supporting comparable data for the 44% assumption: A 44% attach rate sits between Spotify Family plan conversion (~30-35%, Spotify Q4 2024) and Netflix multi-profile engagement (~50%+) — reasonable given Israeli cultural norms of extended-family involvement in daily childcare. The assumption will be validated during beta (Month 7).

---

6.4 Retention Strategy

Why Retention Is Critical

The annual subscription model means retention = renewal. A customer who renews for Year 2 is pure profit (no re-acquisition cost, near-zero variable cost). The key challenge: the usage is bursty. Parents use Numi intensively for 2 weeks (crisis), then barely at all for months. If they don’t see value during the quiet months, they won’t renew.

Retention Mechanisms

Mechanism	How It Works	Impact
Proactive regression alerts	Numi messages: “Your baby turns 8 months in 2 weeks. A sleep regression is common at this age. Here’s what to prepare.”	Parent thinks: “Good thing I have Numi” during quiet months
Weekly sleep reports	Summary: “This week: 6 continuous sleep nights, 1 wake-up. Trend: improving.”	Ongoing engagement even when things are going well
Nap transition guidance	Numi detects: “Based on patterns, your baby is ready to drop from 2 naps to 1. Here’s a transition plan.”	Tangible value during a developmental shift
Seat engagement	Grandmother and nanny interact with Numi regularly for daily instructions	More touchpoints = more perceived value
Long-term memory	“6 months ago, method X worked in 3 nights. Let’s try it again for this regression.”	Demonstrates irreplaceable accumulated knowledge
Second child incentive	“Expecting again? Add your new baby for $299 — we’ll use everything we learned from your first child.”	Upgrades an at-risk churning customer into a higher-paying one

Target annual renewal rate: 85% (15% churn). This is conservative for a product with bursty usage but strong proactive engagement.

---

6.5 Pricing Strategy

Pricing Architecture (Detailed)

Product	Israel Price	UK Price	US Price	Notes
Numi Sleep Program	$399/year per child (1,245 NIS)	£449/year	$599/year	Core product. 2 seats (parents) included.
Founding Members	$299/year per child (933 NIS)	—	—	First 50 families only. Israel launch.
Additional Seat	$59/year (184 NIS)	£69/year	$79/year	Grandmother, nanny, au pair, night nanny. Own WhatsApp, role-adapted AI.
Additional Child	$299/year per child (933 NIS)	£349/year	$399/year	Sibling coordination included. Existing seats carry over.
Family Bundle (2 children)	$599/year (1,869 NIS)	£699/year	$899/year	Save vs. buying separately.
Money-back guarantee	14 days, full refund	Same	Same	Risk-free trial. If Dorit’s method works in 3-5 nights, almost no refunds expected.

Why $399/Year in Israel

Comparison Point	Price	Numi’s Position
Dorit’s human consulting package	2,000-3,000 NIS ($641-$962)	38-59% cheaper — and available 24/7 for 12 months, not just 4-8 weeks
Huckleberry Premium (annual)	$120/year	3.3x more expensive — but Numi is agentic (proactive), methodology-backed, Hebrew, WhatsApp-native. Entirely different product category.
Nanit (camera + annual subscription)	$420-$600+	Similar price range — but Numi is pure software (no hardware), coaches (doesn’t just monitor), and includes methodology.

Pricing Flexibility for Negotiation

These prices can be adjusted based on market response. The key constraint is that the unit economics must remain viable:

Minimum viable price (Israel)	$249/year	Still profitable at ~100 families. Below this, break-even becomes too hard.
Target price (Israel)	$399/year	Optimal balance of value perception and revenue.
Maximum price (Israel)	$499/year	Risk: perceived as too expensive vs. generic apps. Test needed.

---

End of Sections 4-6

Section 7: Management Team

7.1 Team Philosophy

Numi is built by two co-founders and one full-time engineer — a deliberately lean structure for the pre-revenue stage.

Advantage	Why It Matters
Founder salary cost: $0	Both founders draw zero salary. 100% of capital goes to product and operations.
Zero bureaucracy	Two decision-makers = immediate execution. No committees, no “alignment meetings.”
Complementary skills	Rotem builds the product. Dorit provides the methodology, brand, and customers. No overlap.
Skin in the game	Dorit invests $50K of family capital. Rotem invests 15-20 hours/week of personal time at $0 pay. Both hold 50% equity.
Full-time engineer from day 0	Eliminates single-point-of-failure risk. The second person who knows the entire system.

Core Principle: Numi requires exactly two things — advanced AI technology and a proven sleep methodology. Rotem provides the first. Dorit provides the second. Everything else can be hired — but these two cannot be faked.

---

7.2 Founders

Detailed bios, qualifications, and founders agreement terms are in Section 2.5-2.7. This section focuses on the team’s operational structure.

Rotem Levi — Co-Founder & CTO

Detail	Information
Role	CTO. Builds the product, manages technology and operations.
Current employer	Team Lead, Data Engineering at Bigabid (AdTech)
Time on Numi	Evenings, weekends, holidays (~15-20 hours/week)
Salary from Numi	$0 (Year 1). Funded by Bigabid salary (52,323 ILS/month gross).
Full-time trigger	Net profit ≥ 85,000 ILS/month for 3 consecutive months (~880-920 families)

Why not full-time from day 1: Saves the company $120,000-180,000/year at CTO market rate. The Day-0 engineer handles day-to-day development; Rotem transitions when the numbers prove it (Section 12).

Dorit Kreiser — Co-Founder, Methodology & Marketing Lead

Detail	Information
Role	Methodology Lead + Marketing Lead. The “brain” behind every AI response.
Background	Israel’s leading baby sleep consultant. Nurse-midwife since 1994. Psychotherapist. Doctoral student in infant sleep at Tel Aviv University.
Published work	“Shhhh… At Night We Sleep” — published by Kinneret Zamora, sold at Steimatzky, Book Junction, Shilav. English edition on Amazon.
Audience	~14K Instagram followers. Team of 4-5 trained consultants. Active client base.
Investment	$50,000 owner’s loan to the company
Salary from Numi	$0 (earns through 50% equity)

The Kreiser Family Advantage: Dorit’s husband, Dr. Doron Kreiser, is a senior OB/GYN (Hadassah graduate, Stanford-trained, former Sheba delivery room director). Together they run businesses in maternity and parenting — bringing medical credibility, business experience, and a professional network in Israeli maternal care.

---

7.3 Engineering Team

Day-0 Engineer — $1,500/month

Detail	Information
Role	Full-stack engineer. Learns the entire codebase and business from day 0.
Start date	Day 1 of operations (April 2026)
Cost	$1,500/month (~5,550 ILS). Largest single line item in the budget.
Hiring profile	Junior-to-mid engineer. Python, cloud, willing to learn AI/ADK.
Employment type	Contractor or freelancer (flexibility, no employer overhead costs)

Why hire from day 0 (not later):

Reason	Explanation
Bus factor	If only Rotem knows the system and he’s unavailable, the company stops. Unacceptable risk.
Knowledge transfer	It takes months to deeply understand an AI agent system. Starting early means the engineer is productive by launch.
Velocity	Two developers build faster than one. Rotem architects; the engineer builds features.
Transition preparation	When Rotem goes full-time, he’ll need to manage the engineer — better to have that relationship established.

---

7.4 Responsibility Matrix

Area	Rotem	Dorit	Engineer
AI architecture & design	Leads	—	Supports
Backend development	Leads (critical systems)	—	Leads (features)
RAG & knowledge base	Implements	Provides content	Supports
AI response validation	Reviews	Validates	—
WhatsApp integration	Designs	—	Implements
Infrastructure & DevOps	Leads	—	Supports
Methodology & content	—	Leads	—
Marketing & social media	—	Leads	—
Customer referrals	—	Leads	—
Financial management	Shared	Shared	—
Strategic decisions	Shared	Shared	—
Company operations	Shared	Shared	—

The AI Training Loop

Dorit                              Rotem + Engineer
  |                                      |
  | Writes sleep scenarios              |
  | Provides expert responses           |
  | Defines edge cases                  |
  |                                      |
  +----------> Training Data ---------->+
                                         |
                                         | Encodes into RAG pipeline
                                         | Builds agent logic
                                         | Tests and deploys
                                         |
  +<--------- AI Draft Responses <------+
  |                                      |
  | Reviews professional accuracy       |
  | Approves or corrects                |
  | Identifies new edge cases           |
  |                                      |
  +----------> Feedback Loop ---------->+
                                         |
              Numi improves
              with every iteration

This is an ongoing process — not a one-time effort. As more parents use Numi, Dorit identifies new patterns and the engineering team implements them. The product improves continuously.

---

7.5 Advisory Board

See Section 2.6 for full advisory board strategy.

Agentic AI is building an advisory board in three domains:

Domain	Why	Target Profile	Compensation
Pediatric sleep medicine	Clinical credibility, methodology validation	Pediatrician or neonatal nurse specializing in sleep	0.25-0.5% equity
AI/ML engineering	Architecture review, model safety, scaling	Senior AI engineer from a leading AI company	0.25-0.5% equity
SaaS growth	GTM strategy, unit economics, B2B sales	Former VP Growth at a B2C SaaS company ($10M+ ARR)	0.25-0.5% equity

Dr. Doron Kreiser may serve in an informal advisory capacity on medical and clinical matters.

---

7.6 Hiring Plan

No speculative hires. Every position is triggered by a specific revenue or operational milestone.

Trigger	Role	Type	Cost	Purpose
Day 0 (April 2026)	Full-stack engineer	Contractor	$1,500/month	Core development, bus factor
Gate 2 passed (Month 12 from launch)	UK content/marketing contractor	Contractor	2000-3000/month	English localization, UK market adaptation
~300 families	Part-time customer success	Contractor	500-1000/month	Handle edge cases, onboarding support
~880-920 families (Rotem’s transition)	Rotem full-time as CTO	Employee	60,000-65,000 ILS/month	Full-time leadership and development
Year 3 / US entry	US market lead	Contractor/Employee	4000-6000/month	US GTM, partnerships, cultural adaptation
Year 3+	Second engineer	Employee	3000-5000/month	Scale engineering team

Hiring principles: 1. Revenue first, then hire — every hire is justified by existing revenue, not projections 2. Contractors before employees — flexibility, no overhead, easy to scale down 3. Protect the runway — no hire that threatens the company’s survival 4. Each hire must pay for itself — within 3 months, the hire should generate more value than it costs

---

7.7 Why This Team Works at This Stage

Investor Question	Answer
“Why only two founders + one engineer?”	Because that’s all that’s needed. Technology + methodology = product. The engineer ensures continuity.
“Where’s the marketing team?”	Dorit — with 14K Instagram followers, a published book, an active client base, and a team of 4-5 consultants who refer clients.
“Why is the CTO part-time?”	He dedicates 15-20 hours/week unpaid. A full-time engineer handles day-to-day. Rotem transitions full-time when the numbers prove it.
“Why no salaries?”	Because both founders can afford it — and $0 founder salary means 100% of the $50K goes to product and operations, not paychecks.
“What if someone leaves?”	Founders agreement with 4-year vesting, 1.0-year cliff, IP stays with company, exit clauses defined. See Section 2.7.
“Why 50/50 equity?”	Without technology there’s no product. Without methodology there’s no value. Both contributions are equally irreplaceable.

---

End of Section 7 — Management Team

---

Section 9: Financial Plan & Projections

Financial Executive Summary

Numi is built on a lean financial model designed for capital efficiency and rapid path to profitability. The company is funded by a $50,000 owner’s loan from Dorit, with $0 founder salaries, monthly operating costs of $3,000, and break-even at approximately 100 paying families.

Five key facts:

Fact	Number	Why It Matters
Total investment	$50,000 (~156,000 ILS)	Covers $11K one-time setup + 12 months of zero-revenue operations
Monthly operating cost	$3,000 (~9,360 ILS)	Among the lowest in the industry — almost impossible to kill
Break-even	~100 paying families	Just 0.05% of annual Israeli births (181,609)
Variable cost per customer	~$3.16/month (primary child)	90.5% gross margin — driven by Gemini Pro for the Consultant agent
Revenue from month 7-8	Yes — Dorit’s existing audience	This is not a cold start

---

9.1 Investment Structure

Owner’s Loan — $50,000 (~156,000 ILS)

Detail	Terms
Type	Owner’s loan from Dorit to the company
Amount	$50,000 (~156,000 ILS at 3.12 exchange rate)
Interest	0%
Repayment priority	First — before any dividends, founder salaries, or profit distribution
Repayment timeline	Target: 12-18 months from first revenue
Security	Lien on all company assets and IP
Monthly reporting	Full financial transparency: revenue, expenses, customers, loan balance

Why $50,000 — Not More, Not Less

The $50K covers the absolute worst-case scenario (6-month build with zero revenue, 6 months of slow ramp, and unexpected costs) with a comfortable buffer:

Component	Amount	Calculation
One-time setup (legal, incorporation, compliance)	$11,000	Itemized below
6 months MVP development at $3,000/month (zero revenue)	$18,000	Worst case: full 6-month build
6 months early operations at $3,000/month (slow revenue)	$18,000	Conservative: revenue doesn’t fully cover costs yet
Worst-case buffer	$3,000	Unexpected costs, delays, market surprises
Total	$50,000

Why not more: $50K provides 12 months of zero-revenue runway after $11K setup costs. Revenue starts by month 7-8, extending effective runway well beyond year 1. More capital would be idle and unnecessary. Marketing is funded from revenue once it starts.

Why not less: Cutting to $30K would leave only 3-4 months of buffer after MVP build. One delay or surprise could force an emergency. The extra $20K buys peace of mind and negotiating power.

Why an Owner’s Loan — Not Venture Capital

Criterion	Owner’s Loan ($50K)	Venture Capital ($500K+)
Ownership dilution	0% — Rotem and Dorit hold 100%	15-25% to external investors
Decision-making control	Full — both founders only	Board seats, veto rights, reporting obligations
Growth pressure	None — grow at a healthy pace	Aggressive — investors demand 10x in 3-5 years
Personal risk	Limited — $50K max	High — expectations for $5M+ ARR
Revenue timing	Month 7-8 (existing audience)	Often month 12+ (building audience from scratch)
Return priority	First — loan repaid before dividends	Only at exit (sale/IPO)
Fundraise later?	Yes — if both founders agree, likely Year 3+	Already committed to investor timeline

---

9.2 Startup Costs (One-Time)

Item	Cost	Notes
Legal — founders agreement (attorney)	$3,000	Binding agreement covering equity, IP, exit, governance
Company incorporation	$1,000	Agentic AI Ltd. registration in Israel
Accountant — setup, tax registration	$1,500	VAT registration, bookkeeping setup, tax structure
WhatsApp BSP — initial setup	$500	Business Solution Provider onboarding
Domain, branding, landing page	$1,000	sleep-allnight.com or numi-sleep.com, basic website
Business insurance	$1,000	Professional liability, cyber insurance
Compliance — privacy policy, terms of service	$2,000	Attorney-drafted, Israeli Privacy Protection Law
Setup contingency	$1,000	Unexpected registration or legal requirements
Total one-time	$11,000

---

9.3 Monthly Operating Costs

Fixed Monthly Costs — $3,000/month

Item	Monthly Cost	Annual Cost	Source / Notes
Engineer (contractor)	$1,500	$18,000	Full-time contractor. Largest line item.
Google Cloud (Cloud Run + Firestore)	$150	$1,800	Free tier covers early usage. Scales with users. GCP Pricing
WhatsApp BSP fee	$50	$600	Business Solution Provider monthly fee
AI & Development tools	$500	$6,000	Claude Code (2 developers), AI-powered video/creative editors, GitHub Pro, monitoring tools
Accounting	$300	$3,600	Monthly bookkeeping, VAT filings
Legal retainer	$200	$2,400	Ongoing legal questions, contract review
Contingency	$300	$3,600	Buffer for unexpected costs
Total fixed	$3,000	$36,000

What’s NOT included (and why):

Item	Why It’s Not Here
Founder salaries	$0 — both founders have external income
Marketing	Funded from revenue. See Section 9.4.1 below.
Office space	Remote-first. No office needed at this stage.
Gemini API costs	Variable cost — see below
WhatsApp message costs	Variable cost — see below

Variable Costs — Multi-Model Architecture

Numi uses cost-optimized model selection: each agent runs on the cheapest model that can do its job. Only the Consultant agent (which builds sleep plans using RAG + methodology) uses Gemini Pro. All other agents (Intake, Aftercare, Logger, Booking, Root) use Gemini Flash.

Primary Child — ~$3.16/month variable cost:

Component	Cost/Month	Calculation	Source
Gemini Pro (Consultant agent)	$2.70	~135 consultant calls/month. Avg 2,500 input + 1,250 output tokens/call. $2/1M input, $12/1M output.	Google AI Pricing
Gemini Flash (all other agents)	$0.33	~70% of interactions. Intake, Aftercare, Logger, Root. $0.50/1M input, $3/1M output (Gemini 3 Flash).	Google AI Pricing
RAG retrieval (embeddings)	$0.01	Vertex AI RAG. $0.15/1M tokens. Used by Consultant agent.	Google AI Pricing
WhatsApp messages	$0.06	Service messages (customer-initiated): FREE. Business-initiated (utility): $0.01/msg. ~2-3 proactive/day.	WhatsApp Pricing
Total (primary child)	~$3.16/month

Additional Seat (grandma, nanny) — ~$0.75/month variable cost:

Component	Cost/Month	Why Cheaper
Gemini Flash only	$0.60	Seats use Aftercare + Logger agents (Flash). No Consultant (Pro) — the sleep plan is already built for the primary parent.
WhatsApp	$0.06	Same per-message pricing
RAG	$0.01	Minimal retrieval
Buffer	$0.08	Usage variance margin
Total (per seat)	~$0.75/month	85% cheaper than primary child

Additional Child — ~$3.16/month variable cost:

Each child needs its own Consultant session (Pro) for personalized sleep plan building, so the cost is the same as the primary child.

Variable Cost Summary by Product:

Product	Monthly Variable Cost	Annual Revenue	Annual Gross Profit	Gross Margin
Primary child	$3.16	$399	$361.08	90.5%
Additional seat	$0.75	$59	$50.00	84.7%
Additional child	$3.16	$299	$261.08	87.3%
Family bundle (2 children)	$6.32	$599	$523.16	87.3%

Why margins are excellent despite Pro costs: Gemini Pro is used for only ~30% of interactions (the Consultant). All other agents run on Flash. Even with Pro, annual variable cost per primary child ($37.92) is just 9.5% of revenue.

Gemini pricing note: The variable costs above use current Gemini 3 Flash pricing ($0.50/$3.00 per 1M tokens). The original cost basis (Gemini 2.0 Flash-Lite at $0.075/$0.30) was ~$0.04/month — current pricing at $0.33/month is ~8x higher but still negligible relative to Pro costs ($2.70/month). Google’s model lifecycle is ~6-12 months; pricing changes with each generation. See sensitivity analysis in Section 9.8 for impact of 2x pricing scenarios.

Marketing Costs (Revenue-Funded)

Marketing is NOT part of the $3,000 base operating costs. It’s funded from revenue once the company generates income:

Phase	Period	Monthly Marketing	Funded By
Pre-launch	Months 1-6	$0-200 (organic only)	Minimal — Dorit creates content herself
Launch (Phase 1)	Months 7-10	$0-500	Founding member revenue
Growth (Phase 2)	Months 11-14	$1,000-1,500	Growing subscription revenue
Scale	Months 15+	$1,500-2,000	Established revenue base

Year 1 total marketing: ~5000-8000 (funded entirely from revenue)

---

9.4 Revenue Model

Full pricing architecture and unit economics are in Section 6.5 and Section 6.3, respectively. Key figures for the financial model:

• Average revenue per family (Year 1): ~$425/year ($399 base + ~$26 seat add-on)
• Variable cost per family: ~$41.88/year | Gross margin: ~90%

Revenue Streams

Stream	Year 1	Year 2	Year 3	Notes
Core subscriptions	~90%	~80%	~70%	$399/year (Israel), £449/UK, $599/US
Seat add-ons	~6%	~8%	~10%	$59-79/seat. Grows as families add grandma, nanny
Additional children	~4%	~10%	~15%	$299/child. High in Israel (2.9 children/family)
Family bundles	—	~2%	~5%	2-child discount package

---

9.5 Break-Even Analysis

Monthly Break-Even

Fixed monthly costs:                     $3,000
Revenue per family per month:            $33.25 ($399/year ÷ 12)
Variable cost per family per month:      $3.16

Contribution margin per family/month:    $33.25 - $3.16 = $30.09

Break-even:  $3,000 / $30.09 = 99.7 ≈ 100 families

At 100 paying families, monthly revenue covers all fixed + variable costs. Note: this break-even calculation covers operational costs only (engineer, cloud, accounting, legal, contingency) and does not include marketing spend. With marketing costs of $500-1,500/month (funded from revenue), the effective break-even shifts to approximately 115-140 families depending on marketing intensity. See marketing cost phasing in Section 9.3.

With average seat add-on revenue ($425/year per family, including ~0.44 extra seats — see Section 6.3 for derivation: ~44% of Israeli families add 1 extra seat [ASSUMPTION — based on cultural norms of extended-family involvement in Israel; sensitivity: at 30% attach rate, break-even shifts to ~97 families; at 0% attach rate, break-even = 100 families]):

Revenue per family:   ($425/12) = $35.42/month
Blended variable:     $3.16 + (0.44 × $0.75) = $3.49/month
Contribution margin:  $35.42 - $3.49 = $31.93/month
Break-even:           $3,000 / $31.93 = 94 families

Break-Even in Context

Comparison	Number
Break-even families needed	100 (base subscription only) or 94 (with seat revenue)
Annual births in Israel	181,609 (CBS Israel via Macrotrends, 2024)
Break-even as % of births	0.05%
Parents with sleep problems (~37%)	~67,195
Break-even as % of parents with problems	0.15%
Dorit’s Instagram followers	~14,000
Break-even as % of Dorit’s followers	0.71% (100/14,000)

Capturing less than 1% of Dorit’s own followers covers all costs — including Gemini Pro for the Consultant. This is an extremely low bar.

When Does Break-Even Occur?

Scenario	When	Cash Burned Before Break-Even	Cash Remaining
Conservative	Month 16 (8 months after launch)	~$39,000	~$11,000
Base	Month 13 (5 months after launch)	~$32,000	~$18,000
Optimistic	Month 10 (2 months after launch)	~$27,000	~$23,000

Note: “Break-even” here means monthly revenue ≥ monthly costs (fixed + variable, not including one-time startup costs or marketing). Full capital recovery happens later — see loan repayment analysis in Section 11.

---

9.6 Three-Year Financial Projections

⚠️ MODEL SCOPE: Multi-Market Growth Plan (Israel → UK → US) The projections in Section 9 reflect the full three-market expansion plan: Israel (Year 1), UK (Year 2), and US entry (Year 3). These are growth-plan projections that assume successful gate passage and international expansion. For a more conservative, Israel-centric stress test of loan repayment viability, see Section 11 — which uses the Python financial model with single-market-weighted growth assumptions. The two models produce different Year 3 totals by design: Section 9’s base case reaches 1,000 families/$398K cumulative profit (multi-market), while Section 11’s base case reaches 580 families/$69K cumulative profit (Israel-weighted). See Section 11.5 for a detailed reconciliation.

Revenue recognition note: The projections below use annual upfront cash accounting — each family pays the full annual subscription at sign-up, and the revenue is recognized at the time of payment. This reflects actual cash flow timing and is appropriate for early-stage financial planning. The Python financial model (Appendix A) uses monthly accrual for loan repayment calculations, which produces slightly different monthly revenue figures. Both approaches yield the same annual totals.

Year 1: Israel Only (April 2026 – March 2027)

Months 1-6: Pre-revenue. MVP build. Year 1 includes a 6-month pre-revenue build period followed by 4-5 months of customer acquisition. Months 7-12: Launch → growth. First paying families.

Metric	Conservative	Base	Optimistic
Families (end of year)	50	120	200
Revenue	$20,000	$48,000	$85,000
Total costs	$45,000	$55,000	$57,000
Net income	-$25,000	-$7,000	+$28,000
End cash balance	$25,000	$43,000	$78,000

Revenue calculation (base case): - 50 founding members × $299 = $14,950 - 70 regular families × $399 = $27,930 - Seat add-ons (~44% of 120 families × $59) = $3,120 - Additional children (~5% × $299) = $1,794 - Total: ~$47,794 ≈ $48,000

Cost breakdown (base case): - One-time setup: $11,000 - Operations (fixed): $3,000 × 12 = $36,000 - Marketing (revenue-funded): ~$6,000 (growing from $200/month pre-launch to $1,500/month by M12) - Variable costs (AI/API): ~$1,000 (growing with family count at $3.16/family/month) - Total: ~$54,000 ($47K from seed capital + ~$7K marketing and variable costs funded from revenue; see Section 9.7 month-by-month for detail)

Year 2: Israel + UK (April 2027 – March 2028)

Metric	Conservative	Base	Optimistic
Israel families	150	280	450
UK families	50	120	250
Total families	200	400	700
Revenue	$90,000	$190,000	$350,000
Total costs	$70,000	$105,000	$150,000
Net income	+$20,000	+$85,000	+$200,000
Cumulative profit	-$5,000	+$78,000	+$228,000

Revenue calculation (base case): - Israel renewals: 102 families (85% of 120) × $399 = $40,698 - Israel new: ~178 families × $399 = $71,022 - UK new: ~120 families × £449 ($569) = $68,280 - Seat + child add-ons: ~$10,000 - Total: ~$190,000

Cost breakdown (base case): - Operations base: $3,000 × 12 = $36,000 - Engineer (full year): included in base - UK launch costs: $15,000 (one-time) + $3,000/month × 6 months = $33,000 - Israel marketing: $1,500 × 12 = $18,000 - UK marketing: $3,000 × 6 = $18,000 - Accounting/legal (scaled): $5,000 - Total: ~$105,200 ≈ $105,000

Year 3: Israel + UK + US Entry (April 2028 – March 2029)

Metric	Conservative	Base	Optimistic
Total families	500	1,000	1,500
Revenue	$270,000	$550,000	$850,000
Total costs	$150,000	$230,000	$320,000
Net income	+$120,000	+$320,000	+$530,000
Cumulative profit	+$115,000	+$398,000	+$758,000

Per-Market Family Breakdown (Base Case)

Market	Year 1	Year 2	Year 3
Israel	120	280	400
UK	—	120	350
US	—	—	250
Total	120	400	1,000

Note — Two models, by design: The loan repayment analysis (Section 11) uses an Israel-centric projection (580 families by Year 3) to stress-test loan repayment under conservative, single-market-weighted assumptions. The multi-market totals above reflect the full expansion plan — Israel → UK (Year 2) → US (Year 3). This produces a significant gap in Year 3 numbers (1,000 vs. 580 families; $398K vs. $69K cumulative profit). Both models agree on the most critical metric: loan repayment at Month 33. See Section 11.5 for a detailed side-by-side reconciliation.

Per-market breakdown (all scenarios):

Market	Conservative	Base	Optimistic
Israel Y3	250	400	550
UK Y3	150	350	500
US Y3	100	250	450
Total Y3	500	1,000	1,500

Three-Year Comparison

Metric	Conservative	Base	Optimistic
Year 1 families	50	120	200
Year 3 families	500	1,000	1,500
Year 1 revenue	$20K	$48K	$85K
Year 3 revenue	$270K	$550K	$850K
Year 3 net income	+$120K	+$320K	+$530K
Loan repayment	Not within 3 years	Year 3 (Month 33)	Year 2 (Month 24)
Profitability	Year 2	Year 1 (end)	Year 1
Cumulative profit (3 years)	+$115K	+$398K	+$758K
Cash profit ROI on $50K	2.3x	8.0x	15.2x

---

9.7 Month-by-Month Cash Flow — Year 1 (Base Case)

This table shows actual cash position, accounting for annual upfront payments and timing of expenses.

Month	Phase	New Families	Cumulative	Cash In	Cash Out	Cash Balance
1 (Apr 2026)	Setup + build	—	0	$0	$14,000*	$36,000
2 (May)	Build	—	0	$0	$3,000	$33,000
3 (Jun)	Build	—	0	$0	$3,000	$30,000
4 (Jul)	Build	—	0	$0	$3,000	$27,000
5 (Aug)	Build	—	0	$0	$3,200†	$23,800
6 (Sep)	Build + test	—	0	$0	$3,200	$20,600
7 (Oct)	Beta	5 (free)	5	$0	$3,200	$17,400
8 (Nov)	Launch	25	30	$7,475¹	$4,000	$20,875
9 (Dec)	Growth	15	45	$4,485²	$4,000	$21,360
10 (Jan 2027)	Growth	20	65	$6,980¶	$4,100	$24,240
11 (Feb)	Growth	25	90	$10,475#	$4,500	$30,215
12 (Mar)	Growth	30	120	$17,765**	$4,800	$43,180

* Month 1: $11,000 one-time + $3,000 operating = $14,000 † Months 5-7: $200/month pre-launch marketing begins ¹ 5 beta families convert to founding rate + 20 new founding members × $299 = $7,475 ² 15 founding members × $299 = $4,485 (total founding: 40) ¶ 10 founding × $299 + 10 regular × $399 = $6,980 (50 founding member slots filled) # 25 regular × $399 = $9,975 + ~$500 seat add-ons = $10,475 ** 35 regular × $399 = $13,965 + ~$3,800 accumulated add-ons (seats + additional children) = $17,765

Monthly cash out includes: fixed operating costs ($3,000), marketing (growing from $500 to $1,500/month post-launch), and variable AI/API costs (growing with family count at ~$3.16/family/month).

The key insight: Cash balance never drops below $17,400 (month 7). The $50K provides ample runway even with a full 6-month pre-revenue build. By month 8 (launch), cash flow turns positive on a monthly basis. Total Year 1 revenue reaches ~$48,000 from 50 founding members ($14,950) + 70 regular families ($27,930) + seat/child add-ons (~$4,900).

---

9.8 Sensitivity Analysis

What If Key Assumptions Are Wrong?

Variable	Change	Impact on Break-Even	Impact on Year 1 Cash
Price drops to $299 (permanent, not just founding)	-25% revenue/family	121 families (vs. 99)	-$12K more burn
Growth is 50% slower	60 families instead of 120	Break-even at month 18	End cash ~$28K (safe)
Engineer costs double ($3,000/month)	+$18K annual	103 families	End cash ~$28K (safe)
Gemini Pro costs 2x higher	+$40/family/year	92 families	-$5K additional burn
Marketing spend 2x planned	+$8K annual	122 families	End cash ~$38K (more families too)
Disaster: all of the above	Everything goes wrong	~150 families	End cash ~$12K (still surviving)

Even in the disaster scenario, the company retains ~$12,000 in cash and 4+ months of runway to correct course or wind down orderly.

Conservative scenario note: In the conservative scenario (50% slower growth), cumulative cash may reach a low point around Month 22 before recovering. The gate system (Gate 1 at Month 14, Gate 2 at Month 20) provides early warning well before cash depletion becomes a risk. See Section 10.3, Contingency 2 for the decision framework at each gate.

Price Sensitivity

Annual Price	Revenue/Family/Month	Contribution/Month	Break-Even Families	Viability
$249/year	$20.75	$17.95	167 families	Tight — minimum viable price.
$299/year (founding)	$24.92	$22.12	136 families	Viable for founding cohort.
$399/year	$33.25	$30.09	100 families	Target price. Optimal balance.
$449/year	$37.42	$34.62	87 families	Possible with proven results.
$499/year	$41.58	$38.78	77 families	Premium positioning. Test needed.

Churn Sensitivity

Annual Churn	Year 2 Renewals (of 120)	Revenue Impact	Notes
10% (excellent)	108 renewals	+$5,400 vs. base	Strong product-market fit
15% (base)	102 renewals	Base case	Conservative for annual subscription
20% (concerning)	96 renewals	-$2,400 vs. base	Investigate retention mechanisms
30% (crisis)	84 renewals	-$7,200 vs. base	Product needs major improvement

---

9.9 Burn Rate Summary

Pre-Revenue (Months 1-7)

Metric	Value	Calculation
Monthly burn	3000-3200	$3K base + $200 pre-launch marketing from M5
Total pre-revenue burn (Months 1-7)	~$32,600	$14K (M1: $11K setup + $3K ops) + $9K (M2-4) + $9.6K (M5-7 incl. marketing)
Cash remaining after 7 months	~$17,400	$50K - $32,600 = $17,400 (matches Section 9.7 month-by-month)
Months of additional runway (at zero revenue)	~5 months	$17,400 / $3,000 per month = 5.8 months
Total zero-revenue runway	~12 months	7 months pre-revenue + ~5 months additional at $3K/month

Post-Revenue (Months 8+)

Metric	Value
Monthly costs (ops + marketing)	$3,100-3,600
Monthly revenue (at 50 families)	~$1,660 ($33.25 × 50)
Monthly revenue (at 100 families)	~$3,325 (break-even)
Monthly revenue (at 120 families)	~$3,990 (positive cash flow)
Monthly revenue (at 200 families)	~$6,650 (strongly profitable)

Cost Scaling

As the company grows, costs increase — but much slower than revenue:

Families	Monthly Revenue	Monthly Fixed	Monthly Variable	Total Costs	Monthly Profit	Margin
50	$1,660	$3,100	$158	$3,258	-$1,598	-96%
86	$2,860	$3,100	$272	$3,372	-$512	-18%
100	$3,325	$3,200	$316	$3,516	-$191	-6%
120	$3,990	$3,300	$379	$3,679	+$311	8%
200	$6,650	$3,600	$632	$4,232	+$2,418	36%
300	$9,975	$4,000	$948	$4,948	+$5,027	50%
500	$16,625	$5,000	$1,580	$6,580	+$10,045	60%
1,000	$33,250	$8,000	$3,160	$11,160	+$22,090	66%

Why margins improve with scale: Fixed costs ($3,000/month base) are spread across more families. Variable costs ($3.16/month per family) scale linearly but remain a small fraction of revenue ($33.25/month per family). The main cost increases come from: marketing budget (grows with revenue), cloud infrastructure (scales gently), and eventually team hires (triggered by revenue milestones).

---

9.10 Investment Return Analysis (ROI for Dorit)

From the Investor’s Perspective

Metric	Value
Investment	$50,000 (~156,000 ILS)
Investment type	Owner’s loan (priority repayment) + 50% equity
Maximum risk	$50,000 (in case of complete failure)
Loan repayment (base case)	Full — during Year 3 (Month 33)
Equity value (Year 3, base)	~$1.6M-$2.8M (at 3-5x revenue multiple on $550K ARR)

ROI by Scenario — Cash Profit on $50K Investment

Scenario	Loan Repaid?	When?	Cumulative Cash Profit (Year 3)	Cash ROI on $50K
Conservative	❌ No	N/A (~$46K remaining)	-$50,000 (loss)	-1.0x (loss)
Base	✅ Yes	Month 33 (~Year 3)	+$398,000	8.0x
Optimistic	✅ Yes	Month 24 (~Year 2)	+$758,000	15.2x

Equity upside (separate from cash ROI): If the company reaches scale and attracts acquisition interest, equity value at 3-5x revenue multiple could reach $1.6M-$2.8M (base) or $2.5M-$4.2M (optimistic) at Year 3. However, this is speculative and depends on market conditions. The cash ROI above represents actual profits.

Conservative scenario honesty: In the conservative scenario (50 families Year 1, 185 Year 3), the $50K loan is NOT repaid within 3 years and cumulative losses reach ~$50K. This represents the downside case. The gate system (Gate 1 at Month 14, Gate 2 at Month 20) provides early warning signals to either correct course or wind down before full capital depletion. See Section 10.3, Contingency 2 for the decision framework.

The Asymmetric Bet

Downside:     Maximum loss of $50K (~185K ILS)
              Conservative scenario: loan not repaid in 3 years.
              Even in total failure, company IP has residual value ($55-120K).

Upside:       50% ownership in a company generating
              $550K+ revenue in Year 3 (base case).
              Cash profit: $398K (base) to $758K (optimistic).

Risk/reward:  $50K investment → 8.0x cash return (base case)

What Dorit Should Expect

Scenario	Loan Status	Cash Outcome (3 Years)	What Happens
Conservative (~20% likelihood)	Not repaid (~$46K remaining)	-$50,000 (full investment at risk)	Growth stalls at ~185 families (Israel only). Gate system catches this early — Gate 1 (Month 14) and Gate 2 (Month 20) trigger reassessment or wind-down before full capital depletion.
Base (~50% likelihood)	Fully repaid at Month 33	+$398,000 cash profit (8.0x return)	Steady growth to 1,000 families across 3 markets. Loan repaid in ~2 years 9 months. Company is profitable and growing.
Optimistic (~30% likelihood)	Fully repaid at Month 24	+$758,000 cash profit (15.2x return)	Rapid growth to 1,500 families. Loan repaid in under 2 years. Rotem transitions full-time. Significant equity value.

Likelihood estimates are subjective assessments, not statistical predictions.

Why This Is Low Risk

Protection	How It Works
Tiny break-even	Only 100 families — 0.71% of Dorit’s own Instagram followers
Revenue starts early	Dorit’s audience provides customers from month 8, not month 18
$0 founder salaries	No salary drain. Every dollar goes to product.
Loan repaid first	Before any dividends or profit distribution
Monthly transparency	Dorit sees every expense, every customer, every dollar
Spending controls	Nothing above $3,000 without Dorit’s approval
Residual IP value	Even in total failure, codified methodology + AI system worth $55-120K
12 months runway	At zero revenue after $11K setup ($39K / $3K/month). In practice, revenue starts by month 7-8.

---

9.11 Key Assumptions

Assumption	Value	Source / Basis
Exchange rate (USD/ILS)	3.12	Average 2025-2026
Annual price (Israel)	$399/year	Positioned between generic apps ($120) and human consulting ($641-962)
Average revenue per family	$425/year	$399 base + $26 average seat add-on
Founding member price	$299/year	First 50 families only
Annual churn	15%	Conservative for annual subscription with proactive engagement
Gemini Flash cost (Intake, Aftercare, Logger, Booking)	$0.50/1M input, $3/1M output (Gemini 3 Flash)	Google AI Pricing, Feb 2026
Gemini Pro cost (Consultant agent only)	$2/1M input, $12/1M output	Google AI Pricing, Feb 2026
WhatsApp service messages	Free (customer-initiated)	WhatsApp Pricing
WhatsApp utility messages (Israel)	$0.01/message	WhatsApp Rate Card
Israel annual births	181,609 (2024)	CBS Israel via Macrotrends
UK annual births	594,677 (2024)	ONS
Sleep problems prevalence	37% of children 4mo-5yr	CDC, January 2025
WhatsApp penetration (Israel)	99%	JPost, 2025
Israel fertility rate	2.91 children/woman	Macrotrends, 2024
Blended CAC	$25-40	~50% organic (Dorit’s network) + paid channels ($30-60 CAC)
LTV (1 year, gross)	$425	Single year, no renewal
LTV (1 year, net of variable costs)	$383	$425 - $41.88 variable costs
LTV (2 years, 85% renewal, net)	$709	$383 + ($383 × 0.85)
LTV:CAC ratio	9.6-28.4x	$383/$40 = 9.6x (worst) to $709/$25 = 28.4x (best)

---

9.12 Financial Summary — For Decision-Makers

Why this investment makes sense — in five numbers:

#	Number	What It Means
1	$3,000/month	Fixed operating cost. One of the lowest in the startup world.
2	100 families	Operational break-even (~115-140 with marketing). Less than 1% of Dorit’s Instagram followers.
3	~90% gross margin	Variable cost per family: $3.16/month (Gemini Pro for quality). Still exceptional.
4	12 months	Zero-revenue runway after setup costs. Revenue starts at month 8.
5	8.0x	Base-case cash ROI on $50K investment over 3 years.

The Bottom Line

Question	Answer
How much money is needed?	$50,000 (~156,000 ILS)
What happens in the worst case?	Conservative: loan not repaid, up to $50K at risk. Even in total failure, IP retains value ($55-120K).
What happens in the best case?	50% of a company worth $2.5M-$4.2M+
When is the loan repaid?	Month 33 / Year 3 (base). Conservative: not within 3 years — see Section 11.
When does the company profit?	End of Year 1 (base). Year 2 (conservative).
Who controls the company?	Rotem and Dorit — 50/50, no external investors.
How soon is revenue?	Month 7-8 from formation. Dorit’s audience is the launch pad.
What’s the exit?	Not planned — build a profitable business. VC option in Year 3+ if both agree.

---

End of Section 9 — Financial Plan & Projections

---

End of Sections 7-9

Section 10: Risk Analysis

10.1 Risk Matrix

Every risk below is specific, sourced, and paired with a concrete mitigation. We categorize risks by domain — platform, regulatory, technology, market, financial, and operational.

Platform Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
P1	WhatsApp bans Numi under new chatbot policy — Meta banned general-purpose AI chatbots from WhatsApp Business API effective Jan 15, 2026. If Meta classifies Numi as “general-purpose AI” rather than a structured coaching service, our primary channel is gone.	CRITICAL	Low-Medium	Loss of primary channel. Product unusable until native app built.	Numi is a structured sleep coaching service, not a general-purpose assistant. It supports a specific business function (personalized sleep consulting) with defined workflows (Intake → Consultant → Aftercare), human escalation paths, and template-based messaging. This matches Meta’s definition of an allowed bot. Legal review of WhatsApp Business Solution Terms before launch. Fallback: Accelerate native app development.	TechCrunch, Oct 2025; Respond.io, 2026
P2	WhatsApp API pricing increases — Meta changes message pricing or introduces new fees for AI-enhanced conversations.	Medium	Medium	Variable cost increase. At current volumes, manageable — WhatsApp costs are ~$0.06/customer/month (2% of total variable cost).	Monitor WhatsApp pricing updates quarterly. Service-initiated messages (customer-initiated) remain free. Even a 5x price increase would add only $0.24/customer/month. Build native app capability to reduce WhatsApp dependency over time.	WhatsApp Pricing
P3	WhatsApp BSP (Business Solution Provider) instability — The third-party BSP connecting us to WhatsApp’s API becomes unreliable or shuts down.	Medium	Low	Temporary service disruption.	Use a major BSP with redundancy (e.g., Twilio, 360dialog, Vonage). Maintain ability to switch BSPs within 2-4 weeks. BSP connection is a thin layer — the AI system is platform-agnostic.	—

Regulatory & Legal Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
R1	Israel: Privacy Protection Law Amendment 13 non-compliance — Amendment 13 (effective Aug 14, 2025) explicitly covers AI systems processing personal data. Numi processes sensitive baby health and sleep data. Non-compliance risks fines and enforcement.	High	Medium	Fines up to 3.2M ILS (~$865,000). Reputational damage. Potential forced shutdown.	Conduct a Data Protection Impact Assessment (DPIA) before launch. Appoint a Data Protection Officer (can be external/part-time for early stage). Implement informed consent flows in onboarding. Draft attorney-reviewed privacy policy and terms of service (budgeted at $2,000 in startup costs).	IAPP, 2025; Safetica
R2	Medical device misclassification — If Israeli or UK regulators classify Numi as a medical device rather than a wellness/coaching tool, registration requirements could delay or block launch.	High	Low	Months of delay + regulatory costs. In Israel, AMAR (Ministry of Health) registration required. In UK, MHRA registration required.	Position Numi as coaching guidance, not medical advice. Clear disclaimers: “Numi is not a medical device and does not diagnose, treat, or prevent any medical condition.” No health claims. Engage regulatory counsel before launch to confirm classification. Dorit’s credentials add legitimacy but the product does not perform clinical functions.	ICLG Digital Health, Israel
R3	EU AI Act: high-risk classification — The EU AI Act (enforcing from Aug 2, 2026) classifies health-related AI as potentially high-risk, requiring conformity assessments and CE marking. Applies when entering UK/EU markets.	Medium	Medium	Conformity assessment costs ($10K-25K). CE marking process. Delayed UK entry.	Not applicable for Israel launch (Year 1). For UK (Year 2): conduct classification assessment before expansion. Sleep coaching is wellness, not clinical AI — likely not high-risk. If classified as high-risk: budget for conformity assessment in UK launch costs. EU Digital Omnibus package may postpone deadline to Dec 2027.	Trilateral Research, 2025; SecurePrivacy, 2026
R4	US state AI laws — California (AB 489, SB 243), Texas (TRAIGA), Colorado AI Act, and Illinois mental health AI laws create a patchwork of compliance requirements for US entry (Year 3).	Medium	Medium	Multi-state compliance costs. Disclosure requirements. Potential restrictions on AI-provided coaching.	US entry is Year 3 — 2+ years to prepare. California SB 243 requires chatbot disclosure and minor protections; Numi already discloses AI use and targets adult parents. Engage US regulatory counsel 6-12 months before US entry. Budget $40K-60K for multi-state compliance review.	Wiley, 2026; Cooley, 2025
R5	AI liability for harmful advice — If Numi’s AI provides incorrect sleep guidance that results in harm (e.g., unsafe sleep positioning, failure to escalate a medical concern), the company faces legal liability. EU Product Liability Directive (Dec 9, 2026) explicitly covers AI-caused psychological harm.	High	Low	Lawsuit, reputational damage, regulatory action. EU expanded liability includes psychological harm.	Dorit validates AI accuracy through continuous sampling (the “Dorit Loop”). RAG architecture grounds responses in verified methodology, not generic internet content. Hard-coded safety escalation rules: medical concerns → pediatrician referral; parental distress → crisis hotline. Human escalation path (Booking Agent → Dorit). Professional liability insurance. Health disclaimer in every conversation.	Goodwin Law, 2025

Technology Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
T1	Gemini model deprecation — Google deprecates models Numi depends on. Gemini 2.0 Flash-Lite already deprecated (shutdown Mar 31, 2026). Gemini 2.5 Flash scheduled for deprecation June 2026. Model lifecycle is ~6-12 months.	High	High	Must migrate to new model version. Testing and validation required. Potential temporary quality changes.	ADK framework abstracts model selection — switching models requires changing one configuration parameter, not rewriting code. Run migration testing during Google’s deprecation notice period (typically 3-6 months). Maintain latest-generation model compatibility. Current POC already uses gemini-3-flash-preview — latest generation.	Google AI Dev Forum; Google AI Changelog
T2	Gemini pricing increases — Google increases API prices. Current Gemini 3 Flash is $0.50/$3.00 per 1M tokens (already more expensive than Flash-Lite at $0.075/$0.30). Pro is $2.00/$12.00. Price increases directly impact variable costs.	Medium	Medium-High	Variable cost per customer increases. At current usage, doubling Pro prices would increase variable cost from $3.16 to ~$5.86/month per primary child — still under 18% of annual revenue.	Multi-model architecture provides pricing flexibility — can shift workloads between model tiers. Monitor cheaper alternatives (e.g., Gemini 2.5 Flash-Lite at $0.10/$0.40 per 1M tokens — 5-7x cheaper than Gemini 3 Flash for non-critical agents). Negotiate Google for Startups credits ($200K-350K). At worst: price increase absorbed by margin (90.5% → ~82% at 2x Pro pricing — still excellent). Consider self-hosted models for Flash-tier tasks if pricing becomes prohibitive.	Google AI Pricing
T3	Google Cloud platform dependency — Single cloud provider. If Google Cloud has extended outage or changes terms unfavorably, Numi is affected.	Medium	Low	Service disruption during outage. Vendor lock-in.	GCP has 99.95% uptime SLA for Cloud Run. ADK is open-source (Apache 2.0) — can be self-hosted. Firestore data can be exported. WhatsApp integration is BSP-agnostic. True lock-in is minimal — the custom code (agents, RAG, prompts) is ours. Migration to Azure/AWS would take 2-4 weeks of engineering.	GCP SLA
T4	AI accuracy failures — Numi gives incorrect or dangerous sleep advice. Edge cases not covered by RAG. AI “hallucination” overrides methodology.	High	Low-Medium	Harm to child, liability, reputation damage, churn.	RAG grounds responses in Dorit’s verified methodology — reduces hallucination risk vs. generic LLMs. “Dorit Loop” quality process: daily conversation sampling (first 3 months), weekly thereafter. Hard safety rules for medical concerns. Human escalation path. Target: ≥90% accuracy validated by Dorit. Continuous methodology updates based on edge cases.	See Section 8.4
T5	Data breach — Unauthorized access to sensitive baby and family data.	High	Low	Legal liability, regulatory fines, reputation destruction.	AES-256 encryption at rest (GCP default). TLS 1.3 in transit. Role-based access control. Minimal data collection. No third-party data sharing. Compliance with Amendment 13 and GDPR requirements. Cyber liability insurance (budgeted).	See Section 8.5

Market & Competitive Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
M1	Huckleberry invests heavily in Berry AI / Owlet enters AI coaching — Berry launched Feb 2026. Huckleberry ($15.9M VC-funded, 5000000+ families) could evolve Berry from reactive to proactive. Owlet ($128M revenue, 1200000+ babies) announced AI coaching partnership with webAI (Feb 2026).	Medium	Medium	Direct competition from well-funded competitors.	Berry is reactive (answers questions), not agentic (manages the process). Different product category. Owlet’s AI coaching is early stage. Neither supports Hebrew. Hebrew localization would require methodology partnership — a significant barrier. Numi’s moats: Dorit’s methodology, WhatsApp-native, multi-seat, sibling coordination.	Berry Launch PR, Owlet-webAI
M2	Nanit expands into AI coaching — Nanit raised $50M in Dec 2025 for “Parenting Intelligence System.” Could add coaching to their camera-based monitoring.	Medium	Low-Medium	High-funded competitor with camera data advantage.	Nanit’s DNA is hardware + computer vision. Adding methodology-driven coaching requires a fundamental pivot. They have no methodology partner. Camera data ≠ coaching ability. Different price tier ($420-600+ vs. $399). And they don’t support WhatsApp.	Nanit $50M Raise
M3	New Israeli AI competitor with Hebrew support — A well-funded Israeli startup builds a competing AI sleep consultant with Hebrew support and WhatsApp delivery.	Medium	Low	Direct head-to-head competition in our primary market.	First-mover advantage — Numi will be operational before any competitor can build a comparable product. Dorit’s methodology is an irreplaceable moat (published book, clinical credibility). Building a multi-agent sleep system from scratch takes 12-18 months. Data moat grows with every family.	—
M4	Economic downturn reduces discretionary spending — Israeli families cut “non-essential” spending, including sleep coaching.	Low	Low-Medium	Slower customer acquisition. Higher churn.	Baby sleep is not discretionary — it’s a crisis. Parents in distress at 2AM will pay for help. $399/year is already 38-59% cheaper than human consulting ($641-962). Low break-even (100 families) means the business survives even with slow growth.	—
M5	Dorit’s reputation risk — Controversy about the “cry-it-out” methodology. Public criticism or negative media attention.	Medium	Low-Medium	Brand damage. Customer hesitation.	Dorit’s method is evidence-based and published by a major publisher (Kinneret Zamora). She’s a nurse-midwife and doctoral student at TAU — credentials are solid. Controversy actually drives awareness. Numi’s safety protocols (fixed-interval check-ins with monitoring, not pure cry-it-out) provide defense. Clear disclaimer: “This is Dorit’s methodology. Consult your pediatrician for medical concerns.”	—

Financial Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
F1	Conservative scenario: business not viable — If growth follows the conservative path (50 families Y1, 185 Y3), the $50K loan is not repaid within 36 months and cumulative losses reach ~$50K.	High	Low-Medium	Loan default. Total loss of investment.	Gate system provides early warning: Gate 1 (Month 6 from launch) requires 30 families. If not met → reassess. Gate 2 (Month 12) requires 120 families. If not met → do not expand. M12 go/no-go: <80 families → consider shutdown. The $50K covers 12 months of operating costs (after $11K setup) — ample time to identify if the business works.	Financial Modeler analysis, Feb 2026
F2	Loan repayment slower than projected — Even in the base case, the $50K loan is repaid at Month 33 (December 2028). Dorit may expect faster repayment.	Medium	Medium	Tension between co-founders.	Clear expectations in founders agreement: target 12-18 months from first revenue (Month 8). 50% of net profit allocated to loan repayment. Monthly financial reporting gives Dorit full visibility. If base case timeline is unacceptable, discuss before signing.	Financial Modeler analysis, Feb 2026
F3	Exchange rate fluctuation — Plan assumes 3.12 ILS/USD. Material changes affect all calculations.	Low	Medium	Revenue/cost calculations shift. A 10% ILS strengthening (3.33 ILS/USD) increases effective NIS costs for USD-priced services.	Most costs are in ILS (engineer, accounting, legal). Revenue is collected in local currency per market. Natural hedge: Israeli revenue in ILS, UK in GBP, US in USD. Exchange rate sensitivity: ±10% shifts break-even by ~8 families.	—

Operational Risk

#	Risk	Severity	Likelihood	Impact	Mitigation	Source
O1	Rotem unavailable (illness, burnout, personal emergency) — Rotem is part-time CTO. If he’s unavailable for an extended period, development stops.	High	Low	Development halted. Company loses its technical leader.	Full-time engineer from Day 0 knows the entire codebase. ADK framework is well-documented. Codebase is modular (5 independent agents). Dorit can manage the engineer for maintenance. If long-term: hire a replacement engineer/CTO (funded by existing revenue or remaining capital).	—
O2	Dorit unavailable — Dorit is the methodology expert. If she can’t validate AI responses or provide content, quality degrades.	High	Low	AI quality drops. No methodology updates. Marketing effectiveness decreases.	RAG knowledge base is already encoded — the AI operates from existing methodology. Dorit’s team of 4-5 trained consultants can provide temporary methodology support. Published book serves as methodology reference. Most ongoing work is validation, not creation.	—
O3	Engineer quits or underperforms — The Day-0 engineer is a single point of failure for day-to-day development.	Medium	Medium	Development slows. Recruitment takes 2-4 weeks. Knowledge transfer required.	Rotem maintains architecture knowledge. Comprehensive code documentation. Modular codebase. Contractor model (no employment obligations). Replacement engineer can be hired within 2-4 weeks at $1,500/month.	—
O4	Slow customer onboarding — Parents find the WhatsApp intake process confusing or too long. Drop-off before completing baby profile.	Medium	Medium	Lower conversion from sign-up to active customer. Wasted marketing spend.	5-stage intake designed for natural conversation (not forms). Test with beta families (5 free families in Month 7). Optimize based on drop-off data. Dorit reviews onboarding flow for warmth and clarity.	—

---

10.2 Risk Priority Map

Critical Risks (Require Immediate Action Before Launch)

Risk	Action	Owner	Deadline
P1: WhatsApp policy	Legal review of WhatsApp Business Solution Terms. Confirm Numi’s classification as structured coaching service.	Rotem + Attorney	Before WhatsApp BSP connection (Month 1)
R1: Amendment 13 compliance	Conduct DPIA. Draft privacy policy. Implement consent mechanisms.	Rotem + Attorney	Before launch (Month 7)
R5: AI liability	Draft comprehensive health disclaimers. Implement safety escalation rules. Obtain professional liability insurance.	Both founders + Attorney	Before launch (Month 7)
T4: AI accuracy	Dorit validates 100+ conversation samples before beta. Define accuracy benchmarks (≥85% pre-beta, ≥90% at launch).	Dorit + Rotem	Before beta (Month 7)

High Risks (Monitor Actively, Act if Triggered)

Risk	Trigger	Action
T1: Model deprecation	Google announces deprecation of current model	Begin migration testing within 2 weeks of announcement
F1: Conservative scenario	<80 families at Month 12	Founders review: pivot, extend timeline, or wind down
R2: Medical device classification	Regulatory inquiry or pre-UK legal review	Engage regulatory counsel immediately
O1/O2: Founder unavailability	Extended absence (>2 weeks)	Activate contingency plan (engineer/consultants)

---

10.3 Contingency Plans

Contingency 1: WhatsApp Channel Lost

Trigger: Meta revokes Numi’s WhatsApp Business API access or explicitly bans structured AI coaching bots.

Immediate response (Week 1-2): - Notify all active families via email - Activate SMS-based interim communication - Begin emergency native app development sprint

Medium-term (Month 1-3): - Launch minimal web-based chat interface (ADK supports web deployment) - Accelerate React Native/Flutter app timeline to emergency priority - Maintain all backend AI systems (they’re platform-agnostic)

Impact: 1-3 months of disruption. Estimated cost: $10K-20K for accelerated app development. Revenue may drop 30-50% during transition (some families churn due to inconvenience). The multi-agent AI system, RAG, and all business logic remain intact — only the communication layer changes.

Why this is survivable: The AI system is headless. WhatsApp is a delivery channel, not the product. The product is the multi-agent sleep consulting system. Channels can be replaced. The intelligence cannot.

Contingency 2: Business Fails to Reach Viability

Trigger: <80 families at Month 12 (4 months after launch), OR <200 families at Month 24.

Decision framework at Month 12:

Situation	Action
<50 families	Wind down. Repay remaining capital to Dorit. Dissolve company. IP retained for potential future use.
50-80 families	Extend. 3-month extension with zero marketing spend. Focus on product improvement and retention. If no improvement by Month 15 → wind down.
80-120 families	Continue. On base-case track. Maintain current strategy.
>120 families	Accelerate. Invest in growth. Consider Gate 2 preparation.

Wind-down process: 1. Notify all active families (30-day notice) 2. Offer pro-rated refund for unused months 3. Export family data per privacy policy 4. Shut down cloud infrastructure 5. Repay remaining capital to Dorit (priority) 6. Distribute remaining assets 50/50

Estimated wind-down cost: $5K-10K (legal, refunds, infrastructure shutdown). Remaining capital returned to Dorit.

Contingency 3: AI Safety Incident

Trigger: Numi provides advice that results in harm, or a parent reports a safety-related complaint.

Immediate response (Hours): 1. Both founders notified immediately 2. Pause the specific AI workflow that caused the incident 3. Manual review of the conversation log 4. Contact the affected family directly (Dorit leads) 5. If medical concern: refer to pediatrician immediately

Investigation (Days): 6. Dorit reviews the AI response vs. methodology 7. Identify root cause (RAG retrieval error? Hallucination? Edge case?) 8. Implement fix (RAG update, guardrail addition, prompt correction) 9. Re-validate with 50+ test cases

Communication (Week): 10. If public: transparent statement about the incident and fix 11. Update all families if the issue could affect them 12. Report to insurance provider 13. Legal review of liability exposure

Prevention: - Hard-coded safety rules that cannot be overridden by AI: medical symptoms → pediatrician referral; distress signals → crisis hotline - Never advise on: medication, medical conditions, feeding allergies, physical health - Every AI response includes implicit context: “Based on Dorit Kreiser’s sleep coaching methodology”

Contingency 4: Co-Founder Dispute

Trigger: Disagreement that cannot be resolved through normal discussion within 14 days.

Resolution process (defined in Founders Agreement): 1. Discussion (14 days): Good-faith direct discussion 2. Mediation (30 days): External professional mediator, agreed upon by both parties 3. Binding arbitration: If mediation fails, binding arbitration by agreed arbitrator

Financial protections during dispute: - No spending above $3,000 without joint approval (already a standing rule) - Operations continue under existing budget - Engineer continues work under current direction - No unilateral changes to pricing, strategy, or IP

---

End of Section 10 — Risk Analysis

---