AI Venture Creation: CareQuarter

Executive context

Most startups fail not because founders cannot build, but because they build the wrong thing. The research phase is where ideas get validated or killed: what customers want, how they describe their pain, what they will pay, and what will break trust.

This case study documents an experiment in which Ditto synthetic research compressed the entire customer discovery phase of a new startup into a single afternoon. Claude Code and Ditto were paired and tasked with founding a startup from scratch: identify a market, research customer needs, validate a concept, test positioning and pricing, and produce investor-ready deliverables.

The result was CareQuarter, a care coordination service for adult children managing aging parents. The concept emerged from three iterative rounds of synthetic research conducted in roughly four hours. Every major product decision was derived directly from persona responses.

The experiment was not designed to prove that AI can replace human founders. It was designed to demonstrate what happens when the research bottleneck disappears.

Background: why venture creation is a research problem

The conventional timeline

1. Problem hypothesis (Week 1 to 2): Founders identify a problem based on personal experience or pattern recognition.
2. Customer discovery (Week 3 to 8): Interviews with 20 to 40 potential customers to validate the problem and understand workarounds.
3. Solution design (Week 6 to 10): Concepts shaped by interview findings, competitive analysis, and founder intuition.
4. Positioning and pricing (Week 8 to 12): Messaging tested informally and pricing set by benchmarking.
5. MVP and launch (Week 12 to 24): Build, ship, and discover what was wrong.

The research phases can consume 8 to 12 weeks and produce qualitative data from a small, biased sample. Founders routinely skip or compress these steps because they feel slow relative to the urgency of building.

What Ditto changes

Ditto collapses customer discovery, solution design, and positioning into hours rather than months. It generates demographically grounded synthetic personas who respond to open-ended questions with specificity and emotional texture, making early research feasible before any code is written.

The experiment design

Constraints

• No prior domain expertise: The market was selected during the experiment itself.
• No human intervention: Beyond the initial instruction to "found a startup," no guidance was provided during research.
• Iterative methodology required: Each study had to inform the next.
• Deliverables required: The experiment ended only when investor-grade outputs existed.

The three-phase methodology

Phase 1: pain discovery

Study design

The first study recruited 12 synthetic personas: U.S. adults aged 45 to 65 who were managing healthcare for at least one aging parent. Seven open-ended questions explored where time and stress accumulate, what workarounds exist, and what they would never outsource.

• Which care admin tasks consume the most time.
• Moments of highest stress and frustration.
• How they currently cope or who helps.
• What they wish existed but cannot find.
• What they would never trust a third party to do.

Key findings

• Portal fragmentation: every provider, pharmacy, lab, and insurer uses a different system.
• Prior authorization ping-pong: insurers and providers deflect responsibility and the caregiver must chase.
• HIPAA purgatory: authorizations are signed but not visible in provider systems.
• Friday 4pm fires: discharge calls arrive with nothing arranged and no context.
• 2am worry spiral: persistent anxiety about what is being missed.

What this phase determined

The opportunity was not a care app or a portal. It was care coordination with real authority: a named human who could act on behalf of the family. That insight shaped every subsequent decision.

Phase 2: deep dive on trust and authority

Why a second study was necessary

Phase 1 established that customers wanted someone with authority to act. Phase 2 clarified what that authority meant in practice and what guardrails would make trust possible.

Study design

• 10 additional personas with the same demographic filters.
• 7 new questions on authority, trust prerequisites, deal breakers, and trigger moments.
• Focus on communication preferences and escalation boundaries.

Key findings

Trust architecture is the critical design constraint:

• Start with HIPAA only: power of attorney is too much trust too fast.
• Named person, not a team: rotating staff was rejected across the board.
• Phone and paper first: personas preferred calls and mailed summaries over apps.
• Guardrails are non-negotiable: spending caps, defined scope, clear exit.

Trigger moments were consistent:

• Hospital discharge, especially Friday afternoon.
• New diagnosis requiring specialist coordination.
• Medication change requiring pharmacy, insurer, and provider alignment.
• Parent moving to assisted living.
• Caregiver burnout or health crisis.

Deal breakers were equally clear:

• Rotating or anonymous staff.
• No spending caps or unclear billing.
• Data resale or marketing reuse.
• Complicated cancellation processes.
• Coordinators who recommend but do not execute.

What this phase determined

The service model became a named coordinator, HIPAA-authorized, operating by phone, with tiered authority and customer-controlled guardrails. The product was a person with a job description, not a software platform.

Phase 3: concept test

Study design

The final phase tested whether the concept would survive pricing, positioning, and purchase intent. Ten additional personas evaluated four positioning options, three pricing tiers, and trial preferences.

Key results

Positioning: clarity beats emotion

The winning positioning statement was "Stop being the unpaid case manager." It validated the customer experience without patronizing them. Emotional options tested poorly.

Pricing: validated across all tiers

Purchase triggers: crisis converts

The strongest conversion trigger was hospital discharge, especially last-minute Friday calls with no plan. This suggests crisis moments should anchor acquisition strategy.

Objections and deal breakers confirmed

• No named coordinator.
• No spending caps.
• Data resale or unclear privacy.
• Opaque cancellation.

From research to deliverables

Using the findings from all three phases, the AI systems produced a complete set of startup deliverables, each tied to specific research signals.

Landing page

• Conversion-optimized copy based on verbatim pain points.
• Trust architecture sections addressing every deal breaker.
• Pricing and FAQ mapped directly to persona objections.

Pitch deck

• Problem framing with customer voice quotes.
• Solution slide grounded in the named-coordinator model.
• Market sizing and business model with validated pricing.

Messaging and positioning guide

• Words to use and avoid based on persona reactions.
• Primary and secondary CTAs tied to conversion triggers.
• Retargeting copy options for crisis moments.

Design specification

• Production-ready design system for the landing page.
• Typography, color system, and responsive breakpoints.
• Section-by-section layout instructions.

See the output: https://app.carequarter.pro

What makes this methodology different

Iterative, not one-shot

Each study informed the next. Phase 1 identified the pain, Phase 2 defined the trust architecture, and Phase 3 validated the commercial model. A single 21-question study would not have produced the same depth.

Specificity over generalization

Personas described exact scenarios: Friday discharge calls, prior authorization ping-pong, and HIPAA forms that never surface. That specificity made the findings actionable for product design and pricing.

Negative signal is as valuable as positive

Rejections were clear: no rotating care teams, no app-first experiences, no emotional positioning, no power of attorney on day one. Each rejection narrowed the solution space.

Speed enables rigor

Because each study took minutes, the team could afford to run three distinct studies instead of compressing everything into one unfocused session.

Implications for stakeholders

Founders and product teams

1. Run a pain discovery study before writing code.
2. Run a trust and design deep dive before committing to a model.
3. Run a concept test before launch to validate positioning and pricing.

Venture capital and accelerators

• Require synthetic research before pitch meetings to improve signal quality.
• Use Ditto studies to pressure-test portfolio assumptions at each milestone.
• Validate thesis areas before committing capital.

Corporate innovation teams

A synthetic research sprint can produce the evidence needed to justify or kill an initiative before it consumes headcount and budget.

What this case study proves

• Synthetic research can surface non-obvious insights. "Responsible without authority" redirected the concept toward a human coordinator model.
• Iterative synthetic research produces deeper insight. Trust architecture and pricing clarity required separate studies.
• Research-derived products are more defensible. Every element of CareQuarter traces to specific persona responses.
• The research bottleneck is now optional. The 8 to 12 week discovery phase can be compressed to hours.

Follow-up questions for further research

• How do pain points and pricing sensitivity vary by region and urban versus rural context?
• Would employers subsidize care coordination as an employee benefit, and what framing would drive adoption?
• What would it take to convert customers away from existing geriatric care managers?
• What keeps customers subscribed after the initial crisis passes?
• How quickly do customers escalate from HIPAA authorization to limited power of attorney?

Closing

The headline is not AI founding a startup. The story is that when the research phase is no longer gated by time or access, founders can move from idea to validated venture in a single afternoon.

About the author

Phillip Gales

Phillip is a serial tech entrepreneur that specializes in applying AI and machine learning solutions to antiquated and heavy industries. He has been a senior leader or founder at a number of succesful startups.

Phillip holds an MBA from Harvard Business School, an MEng from the University of Cambridge, and is a Y-Combinator alum

LinkedIn