Watchlist
The map scores specific workflows against profit, pain, and whitespace. The Watchlist sits one level up: the industries where agent-shaped work clusters, before any single workflow is picked out and researched. It’s the wide net — every sector with enough repetitive, system-spanning, expensive human labor to be worth a closer look — and a way to see at a glance which clusters already have worked pages and which are still untouched.
Narrowed specifically to AI agents — systems that do work: navigate software, make decisions, communicate, follow multi-step workflows — the field is smaller than “AI in general.” The opportunities concentrate where an industry has lots of human operators, repetitive workflows, many systems and screens, heavy email/phone/chat/forms, and expensive labor. The sharpest signal is a job where a person spends all day:
reading emails · making phone calls · logging into portals · moving data between systems · reviewing documents · coordinating people
That’s the profile the teardowns keep surfacing — e.g. agents aimed at healthcare back-office operations rather than another chatbot, replacing work done by large teams.
Primary industries
Section titled “Primary industries”Roughly in order of near-term agent-company potential. On the map links the workflows already worked as scored opportunities; the un-linked workflows in each row are open candidates not yet on the map.
| # | Industry | Agent-shaped workflows | On the map |
|---|---|---|---|
| 1 | Healthcare | scheduling · patient intake · billing · insurance verification · referrals · prior authorization | prior-auth · clinical-trial consent · pharma submission |
| 2 | Insurance | claims processing · underwriting support · policy servicing · customer support | subrogation · workers-comp claims · brokerage / binding |
| 3 | Financial services | loan processing · onboarding (KYC/AML) · compliance · document review | mortgage closing · tax resolution · AP / invoice-to-pay · UCC liens |
| 4 | Real estate | lead qualification · transaction coordination · property management | title search & curative · lease execution · property-tax appeals |
| 5 | Legal | intake · discovery · contract review · document workflows | probate & estate · CLM · M&A due diligence · immigration assembly · estate planning · notarization |
| 6 | Recruiting / HR | sourcing · screening · scheduling · onboarding | employee lifecycle |
| 7 | Customer support | ticket resolution · escalation · account servicing | — (open) |
| 8 | Sales | prospecting · qualification · CRM updates · follow-up | — (open) |
| 9 | Logistics | dispatch · shipment tracking · exception handling | customs & trade compliance |
| 10 | Construction | permits · bids · subcontractor coordination · compliance | change orders & lien waivers · grants & permitting · RFP response |
| 11 | Accounting | bookkeeping · AP/AR · reconciliation · audit prep | AP / invoice-to-pay · compliance attestation · escheatment |
| 12 | Government | forms · citizen services · case management | grants & permitting · RFP discovery & bid |
| 13 | Procurement | vendor communication · purchase requests · approvals | intake & approval routing |
| 14 | Telecom | service provisioning · support · account management | — (open) |
| 15 | Travel | booking · changes · customer support | — (open) |
Second tier
Section titled “Second tier”Real agent surface, but thinner on the operator-heavy + expensive-labor profile, or harder to reach as a buyer:
Manufacturing · Energy (interconnection & permitting) · Agriculture · Hospitality · Education · Nonprofits · Property management · Facilities management
The wider vertical universe (agent-plausible additions)
Stepping back from “which industries are agent-shaped first” to “the full set of software verticals,” the universe is far larger — but most of it (semiconductors, space, robotics, gaming, GIS/mapping, weather, data centers, smart cities) is product/hardware/deep-tech, not the operator-heavy, document-and-portal back office an agent replaces. Filtering the broader list to entries that do fit that profile and aren’t already above:
- Courts & justice / public safety — case management, court filings, records requests. Strong fit: deadline-driven, forms-and-portals, paper-era.
- Care verticals (senior care · dental · veterinary · mental health · home health) — intake, scheduling, billing, prior-auth; healthcare’s back-office pattern repeated per setting.
- Automotive — dealership F&I paperwork, title & registration, warranty claims.
- Retail / e-commerce ops — vendor onboarding, returns/chargebacks, marketplace compliance.
- Field service management — dispatch, scheduling, work-order and invoice handling.
- Media / advertising — rights clearance, contract and royalty ops, ad-trafficking.
These are seeds-of-seeds: noted so the net is on record, not yet filtered to the operator-heavy + expensive-labor bar the ranked rows above clear.
Horizontal plays
Section titled “Horizontal plays”Not an industry — infrastructure or a pattern that cuts across all of the above:
- Vertical back-office ops agent — the generic “human copy-pastes between legacy portals all day” job, productized per vertical.
- Durable-agent / action-audit infra — the durable, auditable execution layer the agents themselves need.
Where the next wave concentrates
Section titled “Where the next wave concentrates”If the goal is where the next wave of agent companies gets built, the weight sits in a handful of sectors — large back-office teams, document-and-portal-heavy work, expensive labor:
Healthcare · Insurance · Legal · Financial services · Logistics · Construction · Recruiting / HR · Accounting
Those eight plausibly hold 50%+ of near-term agent-company opportunity. The pattern is consistent: the prize is replacing the workflows large back-office teams run today, not adding another chatbot on top of them.
A useful split for sizing the field: horizontal technology categories (AI, cloud, dev tools, analytics, automation) are maybe 20–30% of the software universe; industry verticals are the other 70–80%. Which is why the question that matters here isn’t “what software should I build?” but:
Which industry has expensive problems and terrible software?
That reframe is the Opportunities thesis — the arcaneness is the moat.
This page is a hunting ground, not a thesis — industries with enough signal to warrant a look. A candidate graduates from here to a seed thesis, then to a full deep dive, as the evidence comes in. Nothing on the Watchlist is yet researched to evidence standard.