Harmonia · Croatia first
Bounded, expert-informed support for everyday parenting moments.
€400k provisional validation seed · 18-month target runway
Investor send-ahead·Investors and founders
The decision gap
The problem is not a lack of content. It is choosing an appropriate next step under pressure.
Hard moments happen inside routines: leaving, setting a boundary, ending screen time, eating, and going to sleep. Parents need one usable next move—not another library to interpret while the moment escalates.
Existing alternatives
Content, general AI, and professionals each solve part of the job.
- Libraries offer breadth, but still require interpretation in the moment.
- General AI offers immediacy, but its context, evidence, and safeguarding limits are hard to inspect.
- Professionals offer judgement, but are not an always-on workflow for routine decisions.
Initial customer hypothesis
We earn the right to broaden by starting with one narrow, high-frequency job.
Croatian-speaking caregivers of children aged 3–7, testing one recurring non-emergency transition, boundary, or routine challenge.
One cohort · one use case · one geography · explicit exclusions
V1 workflow
Convert context into one bounded next step, then learn from parent feedback.
- Capture minimum-necessary context and the parent's immediate goal.
- Choose a governed content, protocol, refusal, or escalation path.
- Offer one usable action, record the outcome, and adapt only within scope.
Trust boundary
The product's safety boundary is part of the product—not a footnote.
Harmonia does not diagnose, treat, or replace clinicians, emergency services, safeguarding authorities, or trusted adults. Medical, developmental, abuse, self-harm, violence, and child-endangerment indicators require refusal and the appropriate human route.
Two product horizons
Prove the loop now. Expand the governed family system after validation.
- Now: one profile → reviewed recommendation → 10-day plan → outcome → next step.
- Profiles and living plans · longitudinal insight and professional summary · immediate guidance and escalation.
- Quality time · parent education · social-emotional activities · broader ages and multiple children.
Defensibility
The wedge is local trust. The moat must be earned.
- Governed protocol rights and an expert-labelled evaluation set.
- Consented longitudinal outcomes and a documented safety history.
- Repeatable distribution through trusted local relationships.
Evidence today
Founder access to the problem is real. Product-market proof is not.
Harmonia reports live workshops involving 100+ children and founder experience across psychology, family support, business development, and programme delivery.
This may support problem access. It does not prove paid demand, retention, AI quality, safe escalation, or scalable economics.
100-parent · 90-day evidence programme
The first commercial test is designed to falsify the thesis.
- Four-week active protocol plus eight-week follow-up; ≥60% activation and ≥35% week-four retention.
- Test €8.99 and €11.99 monthly; target ≥15% paid continuation by week six.
- Compare expert quality and harmful-output behaviour; require zero unhandled high-severity safety events.
Beachhead market
Size the opportunity from reachable families and channel capacity—not a global category headline.
Regional and English-language markets are contingent upside after the first loop proves safe paid retention and a repeatable route to families.
Eligible households × reachable share × activation × paid conversion × annual price
Go to market
Distribution starts where trust already exists, then expands only after repeatability is measured.
- B2C subscription first; workshops and trusted parent/professional networks recruit the initial cohort.
- Measure qualified reach → activation → retention → paid continuation → contribution margin.
- Treat B2B2C as a 20-account design-partner test—not committed enterprise revenue.
18-month plan
The seed buys evidence gates—not an eight-year forecast.
- Months 0–3: scope, governance, data map, protocol rights, evaluation set, and concierge prototype.
- Months 4–12: 90-day pilot, closed alpha, comparator tests, paid beta, and incident review.
- Months 13–18: repeat the winning channel and launch regionally—or narrow, rework, or stop.
Driver-based economics
Break-even is an output—not a date chosen for the headline.
Base sensitivity: €11.99 monthly price, €7.24 contribution per paid parent, and €25k mature monthly fixed cost imply approximately 3,452 concurrent paid parents for B2C-only break-even.
The deck promises no EBITDA year until measured retention, CAC, support, safety, inference, and staffing inputs replace assumptions.
Team
The execution claim rests on family access, clinical governance, and product delivery.
Sven Špehar Rajević brings business development, marketing, and programme execution. Tihana Starčević brings psychology and family-support experience. Technical, privacy/security, evaluation, and safeguarding ownership must be explicit—not implied by adviser biographies.
Provisional ask
€400k to answer four questions that can unlock scale.
- Will one narrow cohort activate, retain, and pay?
- Can the governed workflow beat a constrained baseline on usefulness and safety?
- Can one trusted channel acquire families without weakening privacy, safety, or support?
- Can the team reach the 18-month evidence gate within one reconciled cash plan?
Appendix · claim discipline
Every material statement resolves to a source, method, assumption, target, or product intent.
- Verified fact: current, retrievable external evidence.
- Internal evidence: method, sample, dates, owner, permissions, and limitations.
- Assumption/target/product intent: visibly labelled and never presented as achieved.
Appendix · competition
The category validates demand—and removes the right to claim uniqueness.
Good Inside publicly markets personalised, real-time support with plans from $23.25/month. Joy announced a $14m Series A in 2025 for an AI parenting product. Human coaching and general-purpose AI cover adjacent jobs.
Appendix · architecture
Each answer path needs a source, owner, version, evaluation set, incident path, and rollback rule.
- Retrieval: curated, licensed, versioned knowledge with named content governance.
- Generation: bounded low-risk explanation with model/prompt/source logging.
- Immediate guidance: deterministic approved protocol—or refusal and human escalation.
Appendix · data and privacy
Longitudinal context is useful only if families retain meaningful control.
- Purpose limitation and minimum-necessary collection.
- Clear consent/lawful basis, vendor and model-training boundaries, and high-privacy defaults.
- Correction, access, retention, deletion, incident response, and accountable ownership.
Appendix · evaluation design
Quality is measured against experts, a constrained baseline, user outcomes, and safety behaviour.
- Blinded expert agreement and task completion versus a constrained general-AI baseline.
- Critical harmful-output rate plus escalation precision, recall, and incident completion.
- Activation, week-four retention, eight-week follow-up, parent confidence, and paid continuation.
Appendix · pilot scorecard
Denominators, thresholds, and failure actions are set before the data arrive.
- 100 consented, ICP-qualified caregivers; ≥60% activation; ≥35% week-four retention.
- ≥15% paid continuation by week six; measure eight-week retention; report refunds and reasons.
- Zero unhandled high-severity events; blinded comparator; pre-registered escalation and harmful-output measures.
Appendix · economic sensitivity
The initial model is transparent enough to challenge—and too early to call a forecast.
- Price tests: €8.99 and €11.99; the model adds a €14.99 high sensitivity only.
- Illustrative contribution: €4.99 / €7.24 / €9.49 after a 25% reserve and €1.75 variable cost.
- Mature fixed cost: €25k/month; B2C-only break-even range: ~2,634–5,008 paid parents; base ~3,452.
Appendix · provisional use of funds
The €400k hypothesis prioritises product evidence and governance before scale.
- €130k core team/delivery; €85k product/engineering; €45k clinical protocol, safeguarding, and quality.
- €30k pilot; €30k channel tests; €20k privacy/legal/finance; €20k tools/admin.
- €40k contingency; €360k operating-spend ceiling over 18 months; no assumed follow-on round.
Appendix · stop conditions
A credible seed plan includes the reasons to stop.
- Pause if clinical/safety governance cannot define a shippable non-emergency scope.
- Narrow or stop if the workflow cannot beat a constrained baseline or retain paying families.
- Do not scale if channel payback, privacy controls, incident handling, or funding runway cannot survive review.
Appendix · full product map
The complete ambition remains visible without pretending the seed builds every module at once.
- Family system: multiple profiles, older-child participation, periodic review, goals, activities, and preserved history.
- Support system: transparent recommendations, 10-day plans, immediate non-emergency guidance, outcomes, and escalation.
- Growth system: progress patterns, professional summary, quality time, parent education, and social-emotional development.
Appendix · external-circulation gate
The local deck is coherent. The company evidence is not yet complete.
- Verify founder/adviser roles, workshop method/permissions, architecture state, and protocol rights.
- Approve clinical/safeguarding ownership, privacy/data map, legal position, and incident process.
- Reconcile the monthly cash ledger, use of funds, price/economic assumptions, and final ask.
Arrow keys, Page Up/Down, Space, Home, End · swipe on touch screens