SecondChair — AI second-chair reporter for small court-reporting agencies

1. One-liner

AI second-chair reporter that lets small court-reporting agencies cover the depositions they currently turn away.

2. Trend signal — why now?

Court reporting is a $2.7B US service market in structural collapse. Three signals converged in 2025:

• Workforce cliff. NCRA reported average member age ~56 as of June 2025, with 50% retirement-eligible. Stenography school enrollment down 74%, half of programs closed. ~23,000 stenographers left in the US, down 21% in a decade.
• Gap is now public. California Judicial Council disclosed that 74% of unlimited civil/family/probate hearings in Q2 2025 occurred with no verbatim record (255,334 of 344,845). CA needs 458 additional reporters to meet current demand. NCSC’s August 2025 workforce findings: 70%+ of legal end-users report shortages, 76% cite scheduling difficulties.
• Tech unlock landed. OpenAI gpt-realtime hit GA Aug 28 2025 with native SIP and image input + 20% price drop. Speech-to-speech speaker-attributed transcription is now economic at $0.50–1/hr of audio. Verbit, Steno, and AAERT-certified digital reporters have built credibility — but tools target enterprise (Verbit) or vertically-integrated agencies (Steno, $49M Series C March 2026, building its own reporters).
• Money on the table. Per-page rates climbed to $3–7 standard, $6.50–7.50 in major markets, plus $250–400 appearance fees. Rough drafts are billable at $2–3.50/page. Every depo a small agency declines is $400–1,200 of margin walking out the door.

Provenance:

• Signal 1 (demand): NCSC + CA Judicial Council disclose 74% of civil hearings have no verbatim record; 70%+ of legal users report shortages — https://courts.ca.gov/news-reference/research-data/shortage-court-reporters-california — 2025-08
• Signal 2 (feasibility): OpenAI gpt-realtime GA + SIP/image + 20% price cut — https://openai.com/index/introducing-gpt-realtime/ — 2025-08-28
• Signal 3 (economic): Steno raises $49M Series C for AI-enabled court reporting; AAERT certifications gaining state-level acceptance — https://siliconangle.com/2026/03/26/steno-raises-49m-change-court-reporting-ai-enabled-transcript-analysis/ — 2026-03-26
• Reinforcing signal: stenoimperium.com chronicles disappearing reporters in rural counties — https://stenoimperium.com/2025/11/02/they-dont-know-we-need-them-the-growing-silence-around-the-disappearing-court-reporter/ — 2025-11-02
• Category: Tech-unlock (with regulatory tailwind from state-level digital-reporter acceptance)

3. The opportunity

Two ecosystems exist today and neither serves the small agency:

• Stenograph / Eclipse / Case CATalyst sell desktop tooling to the individual stenographer. Brilliant for steno — useless when the steno doesn’t exist.
• Verbit / Veritext / Steno / Esquire are the giants. They sell services to law firms — and they’re the ones eating the small agency’s lunch by becoming the default backstop when the corner agency can’t cover.

The 800–1,200 small agencies in the middle (5–25 reporters, $1–10M revenue) get squeezed. They can’t hire stenographers — there aren’t any. They can’t afford Verbit-tier infrastructure. And they’ve watched Veritext/Steno rep their attorneys, eat their accounts, and absorb their reporters with signing bonuses.

What’s missing: a tool the small agency owner runs, that turns one human (a junior digital reporter or a notary-trained AAERT CER) plus an AI second chair into the equivalent capacity of a senior stenographer. Cover the overflow. Bill the rough draft. Hand off to a scopist for certified transcript. Keep the account.

That’s a 10× UX shift versus paying $200K all-in for a stenographer who’s also retiring.

4. Target market

• Primary customer: Small court-reporting agency owner. Title: Owner / President / Director of Operations. Company size: 5–25 reporters, 200–2,000 jobs/yr, $1–10M annual revenue. Geography: US, with concentrated demand in CA, FL, TX, NY, GA, IL, OH, NJ — states with mixed digital/steno acceptance and worst shortages. Most are 1–2 generation family businesses.
• Why they buy: “I turned down 14 depos last week.” Every declined job is lost margin AND a relationship handed to a competitor. They are literally watching their book shrink. Quote pattern from industry reporting: 76% of legal users cite scheduling delays as their #1 complaint — small agencies are the ones telling them no.
• Rough TAM reasoning: ~1,000 small US court-reporting agencies (Yelp/Google maps + AAERT directory triangulation). At $1,000–2,500/mo blended ACV → $12–30M ARR if half are reachable. Adjacent: solo freelancers running their own brand (~3,000–5,000 of them) at lower price tier. Adjacent #2: county/municipal court systems (longer sales but bigger contract).
• Why now for them: State-by-state legislative shifts in 2024–2025 expanded digital reporter acceptance (TN, NC, GA, FL, CA pilots). AAERT CER certification is a real path now ($275 exam, $125/yr membership, online schools). The bottleneck moved from “is this allowed?” to “is the tooling good enough to keep my agency competitive?” Today the answer is finally yes.

5. Product sketch (MVP)

• One-click join for Zoom, Webex, Teams, RingCentral depos + an iOS app for in-person hearings (uses an external lavalier mic per speaker via 3.5mm/USB-C splitter).
• Speaker-attributed live transcript with 2–4 second latency, automatic objection/exhibit/ruling tagging, and inline read-back-on-demand for attorneys.
• Rough-draft generator — produces a billable rough draft within 15 minutes of proceedings ending, with confidence scoring per line and exhibit thumbnails inline.
• Scopist hand-off — exports clean draft into Case CATalyst / Eclipse formats so the agency’s existing scopists/proofreaders certify the final transcript without retooling.
• Agency dashboard — schedule view of which jobs are AI-covered vs steno-covered, billing per page generated, scopist throughput, attorney-facing branded portal.
• Concierge backstop — first 90 days, agency operations team can call a SecondChair human reviewer (us) for any job we flag <90% confidence. Customer never explains AI to attorneys; they explain “our digital reporter.”
• Compliance bundle — auto-generates AAERT-aligned chain-of-custody records, audio archive, signed certification statement template for the certifying transcriber.

6. AI angle — what’s load-bearing

Strip the AI and there’s no product. The wedge is:

• Speaker diarization on legal speech with overlapping crosstalk — gpt-realtime + a domain-tuned diarizer hits agency-acceptable rough-draft quality.
• Real-time tagging of objections, exhibits, off-the-record breaks — a legal-domain agent that runs on the live transcript every 3–5 seconds.
• Rough-draft polish — small fine-tuned model converts raw ASR into Case CATalyst-formatted draft (Q&A blocks, page/line numbers, parenthetical descriptions).
• Confidence-routing — the system routes ambiguous segments to human scopists rather than over-promising. The AI’s ability to say I don’t know is a feature, not a bug.

This is exactly what just became possible in 2025–2026. Two years ago the rough draft was unusable; today it beats a junior digital reporter on consistency.

7. Localization angle

N/A — this is a US-only play (state-by-state legal acceptance gates international expansion). Court systems in UK, Canada, Australia have analogous shortages but different procedural rules and different transcript formats. Phase 2 only after US lock-in.

8. Business model — path to $1M–$5M ARR

• Pricing: Tiered by agency size.
  • Solo / freelancer: $299/mo for 20 hrs of covered audio + $15/hr overage.
  • Agency Standard: $999/mo for 100 hrs + $10/hr overage. Includes 5 reporter seats.
  • Agency Pro: $2,499/mo for 300 hrs + $8/hr overage. Includes 15 seats, scopist routing, branded attorney portal, priority concierge.
• ACV: Blended ~$18,000–24,000 across small agencies. Solo tier ~$3,600.
• $1M ARR math: 60 agencies on Pro ($150K) + 200 agencies on Standard ($240K) + 2,000 solos on $300/mo ($720K) = $1.11M. Realistic at month 12–15.
• $5M ARR math: 200 Pro + 800 Standard + 5,000 solos = $5.0M. Requires US-wide brand inside court-reporting trade — hit by month 30 if the 2025 state-acceptance trend continues.
• Expansion path: (1) per-page billing — take 5–10% of agency’s rough-draft revenue as transactional layer. (2) certified transcript marketplace — offer scopist supply when agency runs short. (3) lawyer-facing transcript-search SKU sold through agency to their attorney clients (a la Steno’s Transcript Genius but co-branded).

9. Go-to-market wedge — first 100 customers

• Scrape the AAERT certification directory + state court-reporting boards → ~3,500 agency contacts. Personalize a 90-second Loom of “your last 10 declined jobs, recovered” and email the owner. Conversion target: 3–5% reply, 1% close = 30–50 customers from 3,500 emails.
• NCRA renegades + AAERT trade events. AAERT national conference (annual, ~600 attendees), CSRA, NVRA. Sponsor a $2K booth, demo the second-chair live on a mock depo. Agencies sign on-site. Target: 15–25 agencies per event.
• Stenonymous + court-reporter Facebook groups — the trade press is already screaming about the shortage. Pay industry bloggers (Stenonymous has a known audience of 10K+ reporters/agencies) for a sponsored case study with a named agency. Lead-gen through trade press is huge in this niche.
• Channel partner: scopist co-ops. ~500 freelance scopist groups exist. They want more work. Offer them 80% of polish revenue to refer their agency clients. They become unpaid sales force.
• Competitive switch from Veritext/Steno overflow. The junior reporter Veritext drops because the venue is “too small” is a perfect lead. Run a “we cover what they won’t” referral on PACER / state court calendars. Niche but pointed.

If 100 customers in 6 months isn’t realistic from the above — the idea isn’t ready. From the math, it is.

10. Build complexity — justification

Medium. Off-the-shelf: gpt-realtime / Gemini Live for ASR+diarization, Zoom/Teams/Webex SDKs for capture, standard SaaS stack. Custom: legal-domain post-processing, Case CATalyst export format (RTF-CRE), confidence routing UI, scopist workflow. iOS capture app needed for in-person. Realistic v1 in 4–5 months for a 2-person team (1 full-stack, 1 ML/audio). The hard part is not the model — it’s earning agency-owner trust on real depos. That’s domain operations, not engineering.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Certified transcript still produced by human (CER+CET or CSR). AI is workflow, not the legal record. State-by-state acceptance trending favorable.
Ethical — no harm / dark patterns	✅	Replaces missing capacity, not existing reporters. Concierge backstop ensures quality floor.
Market exists (evidence above)	✅	$2.7B services market, NCSC + CA + NCRA all confirm shortage in writing. Steno’s $49M Series C confirms VC bet.
1–5 person team can build this	✅	2–3 people, 4–5 months to MVP.
Launchable with <$50K / ₹40L	✅	~$10–15K — API credits, AAERT exam fees for in-house QC, conference sponsorship.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	18/20	Agency owners are losing 20–40% of incoming bookable work. Hair-on-fire — they’d pay this week.
Demand evidence	15	14/15	State-government published statistics, $49M VC round on adjacent thesis, AAERT certification growth. Skeptic-proof.
Build feasibility	15	11/15	Realtime ASR + legal-domain post-processing + Case CATalyst export is non-trivial but not research-grade. 4–5 months.
Distribution clarity	15	12/15	Named directories (AAERT, NCRA, state boards), trade conferences, trade press. Not 13/15 because conversion math is still estimate.
Revenue mechanics	15	13/15	Pricing benchmarked against ReporterBase, Stenograph, Verbit. Per-page transactional upside well-trodden.
Time to first revenue	10	8/10	Agency owners pre-pay pilots. Realistic first dollar in week 6–8 from a friendly agency.
Defensibility	10	6/10	Workflow lock-in (scopist integrations, agency portal), accumulating audio for fine-tuning, brand trust in a tight-knit industry. Not deep moat — Steno could clone if they pivot from full-stack.
Total	100	82/100	Adjusting down to 80 for confidence — score below.

Final adjusted total: 80/100.

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Best-fit founder: a technical operator who can ship realtime audio + LLM workflows, paired with someone who’s run or worked at a court-reporting agency (or has a close family member who has). Without the domain partner, the agency-owner trust gap is too wide.

Key assumptions to validate (3–5)

1. Assumption: Small agencies will hand a real depo to an AI second-chair with a junior human supervising. How to test: Sign 3 agencies for paid pilots in CA, TX, FL. Run 30 depos, measure: rough-draft acceptance rate, scopist hours per page vs steno baseline, billed per-page recovery.
2. Assumption: Rough drafts at gpt-realtime+post-processing quality are billable at $2–3/page without attorney complaint. How to test: Blind A/B with 5 attorneys — give them rough drafts from human stenographer vs SecondChair and measure complaint rate, redo requests, and billing-dispute rate.
3. Assumption: Agency owners will pay $1,000–2,500/mo SaaS, not just per-job. How to test: Pricing-page split test + 30 outreach calls. Measure intent-to-pay at each tier.
4. Assumption: Scopist supply is sufficient at the price points implied. How to test: Survey 50 freelance scopists on capacity + rate at $0.30/page polish for AI-prepped drafts.
5. Assumption: State legal acceptance keeps trending favorable through 2026–2027. How to test: Monthly tracker of state Supreme Court rules + AAERT advocacy filings. If 3 major states reverse course, kill criterion triggers.

Risk flags

1. Regulatory whiplash. NCRA is actively lobbying against digital reporting in CA and other steno-heavy states. A high-profile transcript error in a published case could trigger a rule change.
2. Steno / Veritext moves down-market. They have the brand and capital. If Steno launches “Steno for Agencies” as a tools-only SKU, our wedge narrows. Counter: their incentive is the opposite (they want to own the agency, not enable it).
3. Per-page billing dispute. First time an attorney finds a material AI error in a billed page, the agency pulls the product. The concierge backstop is mandatory in year one to avoid this.
4. Workforce model pushback. Industry trade press (Stenonymous and similar) is loudly skeptical of AI. Go-to-market needs to position as “fill the bookings the steno can’t take” — not “replace the steno.”

14. Structured verdict

Score:                  80/100
Verdict:                STRONG GO
Confidence:             Medium
Best-fit builder:       Technical founder + court-reporting domain partner (operator/agency owner)
Time to revenue:        6–8 weeks from MVP
Capital to launch:      $10–15K
Top 3 assumptions to validate first:
  1. Agency owners will pilot AI second-chair on real depos — 3 paid pilots in CA/TX/FL, 30 jobs, 90% rough-draft acceptance
  2. Scopist supply at $0.30/page polish exists — survey 50 scopists for capacity
  3. State legal acceptance trend continues — monthly tracker of state Supreme Court rules
Kill criteria:
  - Abandon if <2 of 3 paid pilots renew at month 3
  - Abandon if attorney complaint rate on AI-generated rough drafts >8%
  - Abandon if 3+ major states (CA, NY, IL, FL, TX) reverse digital-reporter acceptance in 12 months
  - Abandon if Steno or Veritext launches a tools-only agency SKU before our 100th customer

15. Next step — 1-week validation sprint

• Day 1: Pull AAERT directory + state court-reporting boards. Build a list of 200 small agency owners (5–25 reporters). Personalize cold email referencing their public job-board postings and their state’s shortage data.
• Day 2: Send the 200 emails. Subject: “How many depos did you turn down last month?” Goal: 8–12 booked discovery calls.
• Day 3–4: Run discovery calls. Specifically ask: “If I gave you a tool that let one CER cover the same load as a senior stenographer, what would you pay?” Capture verbatim.
• Day 5: Decide go/no-go on this measurable outcome:
  • GO if ≥4 agencies say they’d pay >$1,000/mo AND ≥2 commit to a paid pilot at a per-job rate.
  • NO-GO if fewer. The signal is real but the wedge into agencies isn’t, and we should restage the idea.

Falsifiable. Either I can sell paid pilots in five days of cold outreach or I can’t.