Japan launch runbook Draft

90-day Japan launch

Translation is the cheap problem. This runbook is the expensive one: the legal gates, the cultural pitfalls, the infrastructure retrofit, the buyer choreography, and the reference-customer math that turn a technically-excellent product into a Japanese enterprise deployment without destroying the brand on impact.

Drafted 2026-04-21 · Companion to /business.html + /technology.html + labor-consultation template

90 days
to first signed pilot, deployed + measuring
¥8-14M
estimated Phase-1 cash outlay (advisors + infra + certifications)
6-12 weeks
legal-gate time cost that adds to every enterprise deal
3 gates
consensus points per deal: sponsor + gatekeeper + 代表者

00What "a successful launch" actually means

90 days does not mean "Kashi is available to all Japanese enterprises." It means the scaffolding under the product is credible enough to survive a real deployment. Specifically:

Honest acceptance criterion. Anything that sounds like "10 pilots in 90 days" is marketing, not reality. Japanese enterprise sales cycles are 6-12 months. Success at Day 90 is one deployment that proves the model, plus the infrastructure to run the next 10.

01Phase 0 — Week -1 to Day 0 (foundations)

The clock does not start at Day 1. A week of pre-work determines whether the next 90 days compound or stall.

Week −1 (before Day 1) Foundations that must exist before the clock starts
Retain a 社労士 (labor-law advisor)
Specialization in 労働施策総合推進法 + パワハラ prevention. ¥50-100k/month retainer. Scope: advisory, not representation. Purpose: review every pilot's 就業規則 amendment, brief 代表者 consultation meetings, co-sign legal artifacts. Without this seat, every deal is a legal coin-flip.
Justine
First-pilot target archetype locked
100-300-person tech-adjacent firm. Foreign-owned or foreign-investor-backed (lower 代表者-consultation friction). CEO-led. Willing to be a public reference. 3-5 named candidates on a shortlist by Day 0.
Both
JP data residency scoped
AWS Tokyo (ap-northeast-1) vs GCP Tokyo vs Vercel Tokyo region comparison. Supabase currently in Tokyo region already; confirm Vercel function execution region + edge config. Migration plan written, not yet executed. See /technology.html §7.5.
Co-founder
SOC 2 + ISO 27001 auditor conversations
Get 3 quotes. Vanta / Drata / Secureframe for evidence collection. BDO / Prescient / Sompo for audit firm (JP-familiar). SOC 2 Type II takes ~6 months; ISO 27001 takes ~4-6 months. Don't sign yet; schedule the kickoff for Day 1.
Justine
Reference-model decisions
Decide the pilot pricing concession (free / 50%-off / full-price). Decide reference-customer rights (logo-only / named quote / case study / on-panel speaking). Decide data-rights for future aggregate benchmarks. Write this down as a 1-page term sheet before any pilot conversation.
Both

02Phase 1 — Days 1-30 (legal + trust + pipeline)

Days 1-30 Legal foundation + 5 pilot conversations started
The goal of this phase is not to sign a pilot. It is to make signing one possible without burning a candidate.

Legal & compliance

社労士 retainer contract signed
Pre-arranged monthly hours, SLA on 就業規則 review turnaround (target: 5 business days), scope includes pilot-consultation meeting attendance.
Justine
Labor-consultation packet reviewed by 社労士
Our existing labor-consultation template gets a full pass. 社労士 flags every claim that needs amendment for 労働施策総合推進法 §30-2 compliance. Update the template. Regenerate its PDF.
Co-founder + 社労士
就業規則 amendment template drafted
A model 就業規則 change that a pilot customer can adopt almost as-is. Covers: purpose specification, data-types processed, retention window, access roles, opt-out path, 労使協議 minute template. Reviewed and signed off by 社労士.
Co-founder + 社労士
APPI compliance artifact written
Public notice text (), purpose specification (), third-party transfer policy ( — "none" is the answer, but it must be written down), retention-period justification, opt-out path. Hosted at /ja/appi.
Justine + 社労士

Infrastructure

Supabase Tokyo region confirmed + stress-tested
Verify ap-northeast-1 primary, no unintended replication to US. Measure cold-start latency from Tokyo against Vercel function region. If Vercel US adds >200ms to Tokyo-to-Tokyo round-trip, migrate to Vercel Tokyo region or evaluate Cloudflare Workers.
Co-founder
SOC 2 Type II audit kicked off
Evidence-collection tool deployed (Vanta / Drata). Policy library written. Day 1 sets T0 for the 6-month observation window. Target: Type II report by Day 180.
Justine
ISO 27001 scoping + ISMS policy write
Parallel-track with SOC 2. ISMS scope statement, asset inventory, risk register (extending our existing business.html §9 11-risk register), access-control matrix.
Justine

Pipeline

5 warm intros requested
Using the 30 outbound templates at /packets/outbound-templates.html. Lead with the walkthrough link + hidden-loss portfolio framing, not "¥7.9M hero." Target mix: 2 CEO direct, 2 CHRO via investor referral, 1 academic partnership with a Japanese workplace-harassment research lab.
Both
2-3 first-meeting conversations completed
Live tour: /tour.html + manager mirror + governance page. Closing ask: 30-minute follow-up with their legal + labor-relations counterparties.
Both
Day-30 exit gate — continue? 社労士 engaged + labor-consultation packet co-signed + at least 3 qualified pilot conversations in progress + SOC 2 / ISO 27001 kicked off + JP data residency validated. If any two of these are missing at Day 30, pause and re-plan. Do not proceed to Phase 2 on a weak foundation.

03Phase 2 — Days 31-60 (pilot signed + infra hardened)

Days 31-60 First pilot LOI signed + Zoom/Teams integration live for that tenant
This phase converts one of the Phase-1 conversations into a signed pilot AND prepares the technical substrate that pilot will run on.

Commercial & legal

Pilot LOI signed
1-page LOI (not a contract). 90-day paid pilot, pricing concession locked, reference-customer rights locked, data ownership clear. Signed by the CEO (economic sponsor), not delegated.
Justine
Their 就業規則 amendment drafted jointly
Our template + their 社労士 (or ours) + their legal team jointly produce the amendment. Budget 2-3 weeks of back-and-forth.
Co-founder + 社労士
代表者 selection + consultation meeting held
If no union, pilot company must elect a 従業員代表者 (employee representative) per 労働基準法 §90. Consultation meeting scheduled, minuted, and the resulting 意見書 filed with the amendment. This is where most non-JP-native products fail — skip this step and the deployment is legally void.
Pilot CHRO + 社労士

Technical

Platform OAuth app configured (Zoom live; Teams and Meet staged)
Zoom Marketplace listing (or Server-to-Server OAuth for single-tenant install). Microsoft Azure AD app registration + tenant admin consent. Google Workspace OAuth + domain-wide delegation. Production webhook secrets set. Today this is the code stub at /technology.html §4 + /api/webhooks/zoom; at Day 60 it is production-wired for the pilot tenant.
Co-founder
First real transcript ingested end-to-end
One of the pilot's own meetings flows through the production webhook path, gets analyzed by the detector pipeline, produces a Manager Mirror. Dog-food before launch. This is also the moment keigo-detector quality gets a reality-check on non-seed data.
Co-founder
Diarization-confidence gate tuned on pilot audio
JP hybrid-room transcripts produce more short turns + more overlap than EN. The MIN_DIAR_CONFIDENCE floor may need raising for JP. Validate against 5-10 pilot meetings before opening the Manager Mirror to the pilot's managers.
Co-founder
JP-native UI copy pass
Native speaker (Co-founder) + one additional native reviewer pass every user-facing string. Specifically check: honorific register in system messages, 敬語 usage in the Manager Mirror observational language, 遠回し phrasing in pattern-summary narratives. Machine translation is not acceptable for any pilot-visible text.
Co-founder + native reviewer
Day-60 exit gate — continue? Pilot LOI signed + 就業規則 amendment filed with 労働基準監督署 + first real transcript analyzed + JP copy pass complete + Manager Mirror available (but not yet visible) to the pilot's managers. If any of these are missing, extend Phase 2 by 2 weeks rather than pushing a half-ready deployment live.

04Phase 3 — Days 61-90 (outcome + reference + expand)

Days 61-90 Managers see mirrors + outcome report + reference-customer release

Deployment

Managers onboarded to Manager Mirror (Lane A only)
Each manager gets a 1:1 walkthrough (human, not email). Emphasize: this is private, this is for self-correction, this is not a scorecard. Anti-laundering rule explained verbally. Budget 30 min per manager; with 5-10 managers at a 300-person co, this is 3-5 hours of founder time.
Justine + Co-founder
Weekly pilot check-ins (sponsor + CHRO + 代表者)
30 min, Fridays. Three-buyer framework in practice: economic sponsor tracks ROI, operational gatekeeper tracks legal comfort, 代表者 tracks worker-side reception. If any of the three stops showing up, pilot is at risk.
Both

Outcome measurement

Day-75 mid-point review
Early signal report. Grade ladder shown. Honest about what the data does and doesn't say at 15 weeks of lookback vs the 90-day baseline target. Revise the Day-90 measurement plan if reality differs from expectation.
Both
Day-90 outcome report
Lane-A effectiveness: did any managers' patterns measurably shift after receiving the Mirror? Lane-B trigger rate: how many patterns persisted past the 90-day window and escalated to governed-remediation? 代表者 sentiment: unprompted feedback from the worker-rep lens. No claims about harassment-prevention — the measured outcome is structural pattern change.
Both

Commercial

Full 1-year contract OR extended pilot decision
Pilot CEO makes a go/no-go call. If go: full contract at Professional or Enterprise tier. If no-go: honest post-mortem, preserve the relationship for re-engagement at month 12. Never pressure.
Justine
Reference-customer release
Case-study page (anonymized metrics + named company if agreed). Logo rights for the Kashi website. One quote from the CEO. One quote from the CHRO. One neutral line from the 代表者 (this third quote is the hardest and the most valuable). Schedule 3-5 reference calls into Q2 outbound.
Both
Pipeline seeded from reference
5-10 next-pilot conversations triggered by the reference. Goal is not signed Q2 pilots; goal is qualified conversations so the Day-180 pipeline can close 2-3 new pilots.
Both
Day-90 exit gate — what "win" looks like Pilot outcome report delivered + reference release signed + 5+ qualified next-pilot conversations booked + one go/no-go decision from the pilot company. Anything less is not failure; it is information about what Phase 4 (Days 91-180) needs to recover.

05Budget

Planning numbers, not quotes. Actual costs vary with advisor seniority, audit firm choice, and pilot concession depth. All figures are 90-day cash outlay in JPY.

CategoryItem90-day costNotes
Advisors 社労士 retainer ¥150-300k ¥50-100k/month × 3 months
Advisors 弁護士 (employment-law) on call ¥200-500k Hourly; only called in for pilot contract review + escalations
Certifications SOC 2 Type II (auditor + evidence tool) ¥3-6M Kickoff fee + first 3 months of Vanta/Drata; full cost realized over 6 months
Certifications ISO 27001 scoping + gap assessment ¥1-2M Gap assessment + policy-writing support; full cert later
Infrastructure Supabase Tokyo region + Vercel ¥100-300k Existing tier likely sufficient; pilot traffic small
Infrastructure Monitoring + security tooling ¥200-500k Sentry + Datadog or equivalent + security scanner
Sales Pilot concession (revenue opportunity cost) ¥2-4M Free or 50%-off pilot at Professional tier × 90 days; treat as CAC
Sales Travel, pilot site visits, consultation attendance ¥300-800k In-person is not optional for consultation meetings in JP
Reserve Contingency (unexpected legal/tech issue) ¥1-2M Keep uncommitted
Total ¥8-14M Founders' salaries not included
Funding posture. This is seed-level outlay, not Series A. Of this, ~¥5-9M is one-time-but-durable investment (certifications, JP infra, template library). The recurring cost is ~¥300k/month for advisors + infra.

06Team RACI (who does what)

Small founding team means everyone does everything. The table below names the primary owner of each workstream — where the buck stops when that thing is late or wrong.

WorkstreamPrimary ownerSecondary / advisor
Legal & compliance (就業規則 / APPI / 労使協議) Justine 社労士 + Co-founder (drafting)
Certification (SOC 2 + ISO 27001) Justine Audit firm + tooling vendor
Pilot sales + commercial terms Justine Co-founder (co-pitch)
Platform integration (Zoom webhook; Teams + Meet staged) Co-founder Justine (review)
Detector pipeline + JP diarization tuning Co-founder Justine (eval harness)
JP-native UX copy + pattern-narrative language Co-founder External native reviewer
Pilot onboarding + manager 1:1 walkthroughs Both
Outcome measurement + Day-90 report Co-founder Justine (narrative)
Reference-customer release + case study Justine Co-founder
Phase-4 pipeline (Days 91-180) Both

07JP-specific risk register

The 11-risk register in business.html §9 covers product + market risk globally. These are the launch-specific risks that only apply during the Japan entry window.

Severity — high
代表者 blocks deployment
Even with CEO signature, the 従業員代表者 can delay or refuse consultation consent. Pilot stalls in Phase 2 with no clear path forward.
Mitigation: consult 代表者 before LOI signing, not after. Treat as a first-class buyer, not a compliance checkbox. 社労士 co-attends the first consultation meeting.
Severity — high
Manager-mirror leak
A manager shares their Mirror screenshot. It circulates. Brand is destroyed before pilot ends.
Mitigation: watermark the Mirror with the viewer's own email. Legal terms explicit: screenshots shared externally void the pilot. Never mandate Mirror viewing; self-select only.
Severity — high
JP diarization quality floor
Hybrid-room JP audio diarization is weaker than headset-grade. Some pilot meetings fail the confidence gate; pilot perceives the product as flaky.
Mitigation: pre-screen the pilot's meeting setup in Phase 1. If they use hybrid rooms with ceiling mics heavily, recommend a headset-only subset for the first 30 days of ingestion.
Severity — medium
Enterprise procurement timeline
CFO / legal / security review extends beyond 90 days. Pilot starts late, Day-90 outcome window doesn't close on time.
Mitigation: bias pilot selection toward founder-led companies where the CEO can compress procurement. Expect 6-12 months for fully-institutional buyers.
Severity — medium
Incumbent acquisition play
Wevox / Atrae / Archaic notice the pilot traction and offer acquisition at a low price, pre-emptively capping upside.
Mitigation: avoid public signals until the reference customer is live. Non-dilutive funding (funding-pathways memo) keeps optionality open without a pressured raise.
Severity — medium
Content-classifier buyer confusion
Buyer expects Kashi to detect "bad messages" (Archaic / FRONTEO model). When told Kashi's default lane does not read content, asks "then how do you detect harassment?" — loops the sales cycle back to the education phase.
Mitigation: lead with the 70%-miss statistic on content classifiers. Keigo-asymmetry detector is the never-available-elsewhere demo moment. Position structural signals as stronger because they catch what content misses.
Severity — low (but watch)
Vendor-consolidation preference
Enterprise IT says "we already have Wevox; adding Kashi is tool-bloat." Procurement blocks even if CEO wants it.
Mitigation: position as complementary: Wevox measures engagement, Kashi measures structural dynamics. Bundle pricing if needed for specific accounts. Never sell as a Wevox replacement.
Severity — low
Cross-cultural UI dissonance
JP-native reviewers flag that the product "feels foreign" even after translation pass. Adoption is slower than expected.
Mitigation: pattern-narrative tone study with 5-10 JP users in Phase 2. Adjust observational language per SAMHSA-style guidelines translated through the 遠回し convention.

08What the runbook does not cover