About the Opportunities
The teardowns reconstruct how applied-AI companies did build their products. Opportunities looks the other way: where could an AI agent fix high-value operational pain that nobody has built for yet? The unit of value is an evidenced opportunity thesis — is the pain real, how much money moves through it, how crowded is the space, and what would an agent actually have to do.
The pattern worth hunting is consistent: high-dollar, deadline- or regulation-driven workflows that still run on paper, fax, portals, and tribal knowledge — arcane and unglamorous enough that no AI-native company has shown up. The arcaneness is not a bug; it is the moat. These pages are written abstractly — about an AI agent operating over documents and multi-party workflows, not any one product — so the analysis stands on its own.
How a candidate is judged
Section titled “How a candidate is judged”Three axes, scored before any build is imagined:
| Axis | The question |
|---|---|
| Profit | How much money moves through the workflow? Deal size × volume, and is there a model that bills from money found/saved (contingency) rather than a budget line? |
| Pain | How acute and urgent is it — does it have a hard deadline, a penalty, or a recurring fire-drill? |
| Whitespace | How uncrowded? ★★★ = near-empty; ★☆☆ = entrenched incumbents already here. |
Evidence discipline
Section titled “Evidence discipline”Same bar as the teardowns — these pages make new external claims (market size, how a workflow is done today, who the incumbents are), so each claim carries a tier:
- VERIFIED — stated/shown on a cited public page.
- INFERRED — reasoned from a cited public signal; confidence flagged.
- SPECULATIVE — best-practice assumption, explicitly labeled and hedged.
Supporting quotes for each cited claim go in a per-opportunity evidence map (evidence/opp-<slug>-evidence-map.md), tracked in the repo so citations stay verifiable as pages change.
The map
Section titled “The map”Each row is a candidate workflow and the page that works it through. Lens marks how far it sits from a generic document/e-signature platform: Adjacent (reuses the same primitives) or Drastic (a true vertical bet, different domain and data).
Profit / Pain / Whitespace below are seed estimates to be confirmed in each deep dive.
| Opportunity | The mess in one line | Profit | Pain | Whitespace | Lens |
|---|---|---|---|---|---|
| Probate & estate settlement | Every death triggers a months-long, multi-party paper ordeal across courts, banks, and the IRS | Very High | Acute | ★★ | Drastic |
| Commercial property-tax appeals | Over-assessed properties overpay tax because appeals are deadline + comps drudgery | High | High | ★★★ | Drastic |
| Unclaimed property / escheatment | Every company is legally exposed to state escheatment but most don’t know it | Med–High | Latent→urgent | ★★★ | Drastic |
| Energy interconnection & permitting | Renewable projects stall for years in interconnection queues and AHJ permitting | High | High | ★★★ | Drastic |
| UCC lien perfection & monitoring | Lenders lose collateral priority on botched, deadline-driven filings | Med–High | High | ★★ | Drastic |
| Government grants & permitting | Paper-era approval chains with no status visibility and strict accessibility rules | Med–High | Med | ★★★ | Adjacent |
| Insurance subrogation recovery | Carriers leave billions unrecovered because adjusters can’t chase every file | High | High | ★★☆ | Drastic |
| Government RFP discovery & bid response | SMBs skip trillions in public contracts because responding is a compliance nightmare | Very High | High | ★★☆ | Drastic |
| Title search & curative | Every property sale needs manual record abstraction and judgment-heavy curative work | Very High | High | ★★☆ | Drastic |
| Tax resolution / IRS-notice response | Tens of millions of notices a year; taxpayers panic and overpay | Med–High | Acute | ★★☆ | Drastic |
| Customs brokerage & trade compliance | Every cross-border shipment needs arcane HS classification and entry filing | High | High | ★★☆ | Drastic |
| Workers-comp & disability claims | Multi-month, multi-party claims bleed admin cost and leakage | High | High | ★★☆ | Drastic |
| Pharma regulatory submission (eCTD) | Each submission gates billions in drug revenue and is assembled by hand | Very High | High | ★★☆ | Drastic |
| M&A due-diligence deal rooms | Associates hand-read thousands of contracts under deadline | High | High | ★★☆ | Adjacent |
| Mortgage / loan-doc closing | A multi-day, multi-party document-and-signature marathon | High | High | ★★☆ | Adjacent |
| Commercial lease execution | Redline ping-pong plus manual signature chasing across many parties | Med–High | Med | ★★☆ | Adjacent |
| Construction change-orders & lien waivers | Approvals run on paper and email; disputes over what was approved | Med–High | Med–High | ★★☆ | Adjacent |
| Procurement intake & approval routing | Vendor contracts get signed, then forgotten | Med–High | Med | ★★☆ | Adjacent |
| Insurance brokerage / policy binding | Brokers re-key the same data across carrier portals | High | Med–High | ★★☆ | Adjacent |
| Healthcare prior-authorization & intake | Auth packets bounce between provider and payer for weeks | Very High | High | ★★☆ | Adjacent |
| Immigration / visa case assembly | 200-page petitions assembled and checked by hand | Med–High | Med–High | ★★☆ | Adjacent |
| Clinical-trial consent & regulatory docs | Consent versioning across sites, audit-critical | Med–High | High | ★★☆ | Adjacent |
| Estate planning / wills & trusts | Manual drafting plus witnessing and notary logistics, per-state rules | Med | Med | ★★☆ | Adjacent |
| Vertical back-office ops agent | Humans copy-paste all day between legacy portals | High | High | ★★☆ | Drastic |
| Durable-agent / AI-action-audit infra | Applied-AI teams lack a durable, auditable layer for agent actions | High | Med | ★★☆ | Adjacent |
| Contract lifecycle management | Signed contracts are unread promises; renewals and obligations slip | High | Med–High | ★☆☆ | Adjacent |
| Accounts payable / invoice-to-pay | Invoice → PO match → approve → pay, with no audit trail | High | Med–High | ★☆☆ | Adjacent |
| HR / employee-lifecycle paperwork | Onboarding/offboarding paperwork chaos across HRIS, payroll, IT | Med | Med | ★☆☆ | Adjacent |
| Compliance attestation & audit | Annual sign-off fire-drills with evidence scattered across teams | Med | High | ★☆☆ | Adjacent |
| Remote online notarization | High-stakes documents need identity-verified, witnessed signing | Med | Med | ★☆☆ | Adjacent |
Research roadmap
Section titled “Research roadmap”Each entry below is a seed thesis — a starting hypothesis, not yet researched to evidence standard. Deep dives run roughly in priority order; a candidate can be promoted, demoted, or killed once the evidence is in.
Wave 1 — blue ocean + provable ROI (highest whitespace, and a model that bills from money found/saved): probate-estate-settlement · property-tax-appeals · unclaimed-property-escheatment · energy-interconnection · ucc-lien-perfection · insurance-subrogation
Wave 2 — big TAM, reachable buyers: government-rfp-response · title-search-curative · tax-resolution · customs-trade-compliance · workers-comp-claims · government-grants-permitting
Wave 3 — adjacent, instructive (crowded but close to the core primitives; useful for contrast): contract-lifecycle-management · accounts-payable · procurement-intake · commercial-lease-execution · ma-due-diligence · mortgage-closing · construction-change-orders · hr-employee-lifecycle · compliance-attestation · remote-online-notarization · durable-agent-infrastructure
Wave 4 — gated, high-moat (research when domain access appears): healthcare-prior-auth · insurance-brokerage · clinical-trial-consent · pharma-regulatory-submission · immigration-case-assembly · estate-planning · vertical-back-office-ops