opportunity-treedraft

Opportunity Solution Tree: Make trustworthy post-session follow-up effortless

Desired Outcome

Outcome Statement: Coaches reliably produce client-ready, trustworthy follow-up in minutes. Current State: baseline unknown (assume 30–60 min manual analysis + 20–40 min drafting). [ASSUMPTION] Target State: < 10 min coach time per session, output sent with light edits. Timeframe: by end of pilot. Owner: Founder/PM.

Why This Outcome Matters

It is the wedge: it attacks the highest-severity pain (after-session work) and exercises the moat (reproducible analysis). If coaches won't trust/adopt this loop, nothing downstream matters.

Visual Tree

                         [Trustworthy follow-up in minutes]

                                       |
        ┌──────────────────────┬───────────────────────┬─────────────────────┐
   [O1 Trust the analysis] [O2 Cut manual effort]  [O3 Keep continuity]  [O4 Stay safe/in-scope]

        |                      |                       |                     |
   Sol 1A/1B              Sol 2A/2B               Sol 3A/3B            Sol 4A/4B

Opportunity Branches

Opportunity 1: "I can't trust AI output enough to send it"

Impact: High · Confidence: Medium (hypothesis) Evidence: page principle "value is in reproducible methodology"; founder emphasis on client-safe output.

Solutions

1A: Reproducible analysis pipeline — multi-step, calibrated, evidence-quoted breakdown. Effort: XL. Riskiest assumption: LLM analysis can be stabilized to "no worse than human." Test: run same session 3×, measure agreement (see metric 10). 1B: Evidence-linked conclusions — every conclusion cites a transcript quote. Effort: M. Riskiest assumption: quotes raise trust enough to send. Test: artifact reaction interviews.

Opportunity 2: "Doing the analysis + write-up by hand eats my week"

Impact: High · Confidence: Medium Evidence: founder; journey map after-session severity 5.

Solutions

2A: One-click editable draft summary — Effort: M. Assumption: editing beats writing from scratch. Test: time "draft+edit" vs "scratch." 2B: Auto transcript + speaker separation — Effort: M (integrate). Assumption: quality high enough for RU/EN. Test: WER + diarization check.

Opportunity 3: "Things slip between sessions"

Impact: High · Confidence: Medium Evidence: page "operational routine" + dossier.

Solutions

3A: Cumulative client dossier — Effort: L. Assumption: coaches open it before sessions. Test: pilot usage. 3B: Commitment tracking + micro-tracking — Effort: M. Assumption: clients respond to check-ins. Test: response rate.

Opportunity 4: "I must not breach trust or step outside coaching"

Impact: High (gating) · Confidence: Medium-High (design principle) Evidence: methodology page (out-of-scope protocol, consent, depersonalisation).

Solutions

4A: Consent + depersonalisation pipeline — Effort: M. Assumption: feasible before external LLM. Test: PII-leak audit. 4B: Out-of-scope alerting (human decides) — Effort: M. Assumption: alerts are appropriate not noisy. Test: review on pilot sessions.

Prioritization

Priority Opportunity: O1 (trust) — it is the moat and the adoption gate. Priority Solution: 1A reproducible pipeline + 1B evidence quotes. Rationale: without trust, the time-saving (O2) is irrelevant because coaches won't send the output.

Opportunity Ranking

Rank	Opportunity	Impact	Confidence	Effort	Score
1	O1 Trust	H	M	H	High
2	O2 Cut effort	H	M	M	High
3	O4 Safety (gating)	H	M-H	M	High (prereq)
4	O3 Continuity	H	M	H	Med

Parking Lot

Live in-session suggestions — sensitive, defer (page warns against interrupting sensitive moments).
Coach marketplace / public showcase — later (studio scope-out).

Experiments Backlog

Solution	Assumption	Test Method	Success Criteria	Status
1A	Analysis reproducible	3× re-run agreement	within noise band, F1 ≥ 0.70 target	Planned
2A	Edit beats scratch	timed comparison	meaningful time drop	Planned
4A	Depersonalisation works	PII audit	100% before external LLM	Planned

Next Steps

[ ] Design reproducibility experiment (metric 10).
[ ] Run artifact-reaction interviews (02).
[ ] Define MVP scope around O1+O2+O4 (see spec/PRD).