Case file — 4E209C46
The idea
“Plaintiff personal injury law firms work on contingency (no win no fee). Their biggest operational cost: processing medical records. For every case, paralegals spend 10-30 hours manually reading through hundreds of pages of medical records, building treatment chronologies, calculating special damages (total medical bills), flagging gaps in treatment that defense counsel will exploit, and identifying missing records to request. At $30-45/paralegal hour, this costs $300-1,350 per case. General legal AI (Harvey, Clio Duo) doesn't understand PI-specific workflows: they can't chronologize treatment across providers, calculate damages in the format insurers demand, or flag the specific gaps that experienced defense counsel look for. We build an AI that ingests medical records and bills, produces a treatment chronology, calculates special damages, and flags gaps/missing records in under 10 minutes per case. Price: $1K-3K/month for PI firms handling 100+ cases.”
The panel
Live data confirms the pain: paralegals spend 2–6 hours sorting medical records per case, and PI firms handling 40+ cases face 200+ hours monthly—exactly your target. The market research identifies four competing platforms (InQuery, Supio, DigitalOwl, Wisedocs) already solving record sorting and extraction, though specifics on their PI-specific capabilities, pricing, or funding are absent from the data provided. Reddit signal shows QME/AME specialists built similar automation—pulling clinical info, organizing by date, structuring summaries in under 10 minutes. They flagged handwritten accuracy and source traceability as real friction points you'll inherit. Red flag: The live data doesn't show whether existing competitors already handle PI-specific damage calculation and gap-flagging—the exact differentiators you claim. If they do, you're repackaging; if they don't, verify this via direct calls to 10 PI firms before building. Strength: PI contingency economics create genuine urgency. Firms bleeding $300–1,350 per case on paralegal time have immediate ROI math. This is sticky if you nail workflow fit.
You're massively underestimating the medical records ingestion problem. PDFs come in wildly inconsistent formats—handwritten notes, scanned faxes, different EHR systems, abbreviations that vary by hospital. OCR + NLP will fail silently on 15-20% of records, producing confidently wrong chronologies. You'll need domain-specific training data (thousands of labeled records) and a human-in-the-loop validation layer that destroys your unit economics before you have one paying customer. Build-vs-buy trap: Don't build your own damages calculator. State insurance commissions have specific rules; you need a configurable rules engine tied to state-by-state regs and insurer formats. Licensing a damages platform exists—use it. No moat. Once one legal AI vendor (Harvey, LexisNexis) adds PI-specific templates and gap-detection heuristics, they'll own this for $200/month bundled. Your only moat is operational: relationships with firms and proof you don't produce garbage output. That takes 18 months and 50 paying customers minimum. One real win: flagging missing records is genuinely valuable and technically tractable—pattern-matching against state-mandated discovery lists plus statistical anomalies (no imaging for spinal injury) works.
The fatal unit economics problem: You're pricing $1K-3K/month on contingency. PI firms don't pay until they win. Your SaaS model assumes recurring revenue; theirs assumes zero cash outflow until settlement. You'll face brutal adoption friction—they'll demand per-case pricing or revenue-share, which collapses your margins. At $300-1,350 saved per case, a firm handling 100 cases/year sees maybe $50K annual benefit. Monthly SaaS pricing doesn't map to their cash cycle. Pricing assumption that fails: You're anchoring to paralegal replacement cost ($300-1,350/case), but PI firms won't pay 75% of that savings to you monthly—they'll negotiate to 20-30% of the actual case value they recover, or demand a 12-month pilot at $500/month. Contingency economics are brutal for vendors. Runway math: Zero traction + no paying customers = 12-18 months before you need revenue. You're burning before you've solved the pricing model. What works: The problem is real, quantified, and repeatable across thousands of firms. If you shift to revenue-share or per-case pricing tied to settlement outcomes, you align incentives with their cash flow and become defensible.
Timing verdict: Late, but with a narrow opening. PI firms have been wrestling with this problem for 15+ years—the pain is real but ingrained. What makes now viable: LLMs finally handle multi-document synthesis reliably enough, and PI firms are actively piloting legal AI after Harvey's 2023 launch. But you're entering a window that closes in 18 months as Harvey, Westlaw, and LexisNexis layer PI-specific modules into their platforms. Macro trend: Regulatory pressure on contingency economics. Rising defense costs and insurer pushback on settlement values are squeezing PI firm margins harder than ever. Automation that cuts paralegal hours directly hits the P&L—this resonates now in ways it didn't five years ago. Opportunity window: Closing. Big legal tech already owns the distribution channel. Your only path is horizontal—sell direct to mid-market firms before they're locked into an enterprise platform. After 2027, you're a feature, not a business. Genuine timing advantage: Defense counsel is already weaponizing AI. PI plaintiffs' attorneys feel threatened by AI-assisted defense strategies. They'll adopt your tool partly defensively—matching capability parity with opposing counsel. That urgency is now.
Competitors found during analysis
Live dataInQuery
Medical records sorting, indexing, extraction
Supio
Medical records sorting, indexing, extraction
DigitalOwl
Medical records sorting, indexing, extraction
Cause of death
The medical records ingestion problem will eat you alive before you have a product
You're pitching "under 10 minutes per case" but the panel's tech assessment says 15–20% of medical records will produce silently wrong outputs — handwritten notes, scanned faxes, inconsistent EHR formats, hospital-specific abbreviations. In PI litigation, a confidently wrong chronology isn't just a bug; it's malpractice exposure for the firm. You'll need thousands of labeled training records, a human-in-the-loop validation layer, and domain-specific OCR tuning. That validation layer destroys your "10 minutes per case" promise and your unit economics simultaneously. You can't half-solve this — 85% accuracy in medical record chronology is worse than a paralegal, because the paralegal knows when she's unsure.
Your pricing model is structurally misaligned with how PI firms spend money
PI firms operate on contingency. They spend nothing until they win. Your $1K–3K/month SaaS subscription asks them to commit recurring cash outflow against an uncertain future recovery. The CFO panel nailed this: they'll demand per-case pricing or revenue-share tied to settlements, and both of those models collapse your margins and make revenue forecasting nearly impossible. You're anchoring to paralegal replacement cost, but firms won't pay 75% of savings to a vendor — they'll negotiate you down to 20–30%, which at $60–$400 per case means you need massive volume to make the math work. You haven't solved the pricing model, and it's not a detail — it's the business.
Four competitors already exist and the enterprise platforms are coming
InQuery, Supio, DigitalOwl, and Wisedocs are already in the medical records extraction space. The market agent couldn't confirm whether they already handle PI-specific damage calculation and gap-flagging — which means you can't confirm your differentiation exists until you've called those competitors' customers. Meanwhile, Harvey, LexisNexis, and Westlaw own the distribution channel to every law firm in America. When they ship PI-specific templates — and the timing panel says that window closes around 2027 — you become a feature they bundle for $200/month. Your only moat is operational trust built over 18 months with 50+ paying customers, and you're starting from zero.
⚠ Blind spot
Malpractice liability will be your real go-to-market blocker. PI attorneys are personally liable for every claim they file. A paralegal who misreads a record can be corrected; an AI that produces a confident-looking chronology with a subtle error — a wrong date, a misattributed provider, an omitted treatment — could lead to an undervalued demand letter or a missed statute of limitations on a records request. The first time that happens, your customer doesn't just churn — they tell every PI attorney at every state bar conference. You're not selling software; you're asking attorneys to trust their contingency fee (often 33–40% of a six-figure settlement) to an AI they can't cross-examine. The human-in-the-loop validation isn't a nice-to-have; it's a regulatory and reputational requirement that fundamentally changes your cost structure, your speed promise, and your positioning.
What would need to be true
You can achieve 95%+ accuracy on missing records detection within 6 months using publicly available clinical logic rules and a training set of 500+ real PI case files — without requiring a human-in-the-loop for every output.
At least 2 of the 4 existing competitors (InQuery, Supio, DigitalOwl, Wisedocs) do NOT already offer PI-specific gap-flagging and damages calculation — confirming your differentiation is real and not assumed. You must verify this with 10 direct customer calls before writing code.
Mid-market PI firms (10–50 attorneys) will adopt a per-case priced tool without requiring integration into their existing case management system — because if they demand Clio/Filevine integration before buying, your sales cycle just went from 2 weeks to 6 months, and you don't survive the runway math.
Recommended intervention
Stop trying to be the full-stack "AI paralegal" and become the missing records detection engine — and only that. The tech panel confirmed this is technically tractable: pattern-matching against state discovery requirements, clinical logic rules (spinal injury without imaging = flag), and temporal gap analysis across providers. This is the one feature where (a) the accuracy bar is lower (flagging for human review, not producing final work product), (b) the liability exposure is minimal (you're catching omissions, not making affirmative claims), (c) the value is immediately demonstrable (one recovered missing record can swing a case value by $50K+), and (d) no competitor has nailed it yet. Price it per-case at $50–$150 — aligned with PI cash flow — and use it as a wedge into 200 firms in 12 months. Once you have trust and data from those firms, then you expand into chronology and damages. You earn the right to be the full platform by proving you don't produce garbage on the simplest, highest-value task first.
Intervention unlocking
5seconds
No account needed. One email, no follow-ups.
Want your idea examined? Free triage or full panel →