Deep Research Decision Visualization
One-page evidence dashboard for the hackathon decision: what we learned, where competitors sit, why the problem is real, how OpenAI powers the product, and what to build for MVP.
Final decision
Proceed with a 30/30 target-ready evidence package around Thai family evidence-to-agreement.
Competitors solve pieces: dementia education, family coordination, doctor recording, clinical notes, generic AI notes, Thai eldercare services, and AI wellness. The executed loops now add direct originality proof, prototype proof surfaces, Thai role/tone logic, safety evals, and a concrete pilot package.
Loop 2 result
Scores come from the competitive-advantage judge scorecard plus the OpenAI sponsor-fit addendum.
Pieces are common; the direct overlap table now proves the full Thai evidence-to-agreement workflow is rare.
Implication: say "common pieces, rare bundle."
Yes. The demo case and prototype surface show Thai evidence becoming a source-traced agreement and LINE/SMS draft.
Implication: demo the transformation chain, not a feature list.
Thai dementia care, family burden, and care-information handoff pain are strongly supported.
Implication: lead pitch with hidden care work and fragmented doctor information.
The prototype now has an API-ready Structured Outputs route, schema view, doctor transcript seed, and safety eval cases.
Implication: set OPENAI_API_KEY before live demo; fallback keeps workflow visible.
Thai role model, tone modes, pilot channel map, and localization path now make the Thai wedge product-level.
Implication: show Thai role/tone logic, not translation.
First buyer, 8-week pilot package, price hypothesis, success metrics, and validation interview plan now exist.
Implication: pitch memory-clinic / hospital caregiver education pilot first.
Strong. OpenAI should power structured extraction, burden cards, safety flags, agreements, and Thai messages.
Implication: demo schema-driven UI, not generic AI prose.
Medical overclaim, transcription hallucination, privacy, blame, and scope creep.
Implication: show human review and non-diagnostic safety boundary.
Loop 3 result
The market is crowded with adjacent substitutes, but most categories solve one or two pieces. Our score is strongest for hackathon differentiation, not yet commercial defensibility.
| Product / category | Thai | Dementia | Burden | Doctor | Record | Imbalance | Agreement | LINE | Weekly | Safety | Avg |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Our product | 5 | 4 | 5 | 4 | 5 | 5 | 5 | 5 | 5 | 4 | 4.7 |
| Dementia caregiver apps | 1 | 5 | 3 | 1 | 2 | 1 | 1 | 1 | 1 | 3 | 1.9 |
| Family coordination apps | 1 | 1 | 2 | 1 | 3 | 2 | 2 | 2 | 2 | 3 | 1.9 |
| Medcorder / doctor visit recorder | 1 | 1 | 1 | 5 | 4 | 0 | 0 | 1 | 1 | 3 | 1.7 |
| Clinical AI scribes | 1 | 1 | 1 | 5 | 2 | 0 | 0 | 0 | 0 | 4 | 1.4 |
| General AI notes / voice tools | 2 | 1 | 2 | 3 | 2 | 1 | 1 | 2 | 1 | 1 | 1.6 |
| Thai eldercare services | 5 | 2 | 2 | 2 | 2 | 1 | 1 | 3 | 1 | 3 | 2.2 |
| AI wellness apps | 2 | 1 | 4 | 0 | 0 | 0 | 0 | 1 | 1 | 3 | 1.2 |
Loop 4 result
These maps clarify our lane: not clinical documentation, not generic notes, not generic task management.
Target quadrant: high burden awareness plus high family coordination.
Our advantage depends on converting inputs into dementia-specific family action.
We should not compete as a clinical AI note company.
Best position: Thai-first workflow that can later localize to other family-care cultures.
Loop 5 result
The problem is not invented. The research and persona feedback point to a real care-information and family-labor problem.
Loop 6 result
The opportunity is credible, but revenue is the weakest research dimension. Treat market sizing as planning estimates, not audited numbers.
Memory clinic / hospital caregiver education pilot first, then B2B2C + freemium family plan.
Eldercare provider add-on, hospital/memory clinic validation, NGO caregiver education.
Freemium family plan and employer caregiver benefit after proof of use.
Insurance or public-health partnership if safety and outcomes are validated.
Pure paid B2C subscription before trust and repeat usage are proven.
Loop 7 result
OpenAI should be the evidence-to-agreement engine. The sponsor story is weak if the product only generates generic chat text.
Thai incident text becomes schema-valid care record, burden evidence, safety flags, agreement text, and LINE/SMS draft.
Doctor-audio and AI scribe summaries can hallucinate or omit details. Keep original transcript and require human review.
Generated with OpenAI Structured Outputs. Review before sharing. Live verification requires OPENAI_API_KEY.
Risk loop
Risks do not invalidate the product; they define the safety and MVP boundary.
Do not diagnose, recommend medication, or replace doctors. Use non-diagnostic family coordination language.
Doctor audio summaries must be drafts with original transcript/source visible for review.
Audio, documents, and family messages are sensitive. Minimize storage and explain consent.
Care imbalance should show support need, not score family members as bad people.
Make structured JSON drive UI cards, charts, and agreement sections.
Mock OCR, LINE send, accounts, calendar, and Realtime until the core flow works.
Loop 8 result
The MVP should prove the full product logic without trying to build every integration during the hackathon.
We help Thai families convert hidden dementia care burden into a respectful family care agreement.
Do not pitch as diagnosis, medical advice, medication guidance, legal contract, patient surveillance, generic note taker, or caregiver marketplace.
Traceability
This HTML is a visualization layer over existing curated markdown research, not a new primary research report.
Source files: `docs/deepresearch-visualization-html-plan.md`, `docs/path-to-30-judge-score.md`, `docs/demo-case.md`, `docs/safety-eval-cases.md`, `docs/thai-family-role-model.md`, `docs/pilot-package.md`, `docs/validation-interview-plan.md`, `research-output/competitive-advantage-final-decision.md`, `research-output/competitive-advantage-judge-scorecard.md`, `research-output/competitive-advantage-score-matrix.md`, `research-output/competitive-advantage-good-bad-ugly.md`, `research-output/competitive-advantage-quadrants.md`, `research-output/competitive-advantage-monetization-tam-sam-som.md`, and `research-output/openai-api-sponsor-fit-deepresearch.md`.