Dream big.
This is the page for the agent who loves AI.
Your book. Your market. Your industry. Questions nobody could afford to ask before 2023. Now they take one night.
Impossible. Until it wasn’t.
Every date below unlocked something an agency could not buy before, at any price.
- Software could search.Sort, file, rate. It could not read a policy.
- ChatGPTAsk in plain English. Get an answer.
- GPT-4Reads like a professional. Passed a simulated bar exam near the top 10%.
- 100,000-token readingReads a whole policy at once. About 75,000 words. Unlocks cross-selling your book.
- 128K to 200KReads a stack of policies in one pass.
- One million tokensHundreds of documents in one read.
- Reads scans and tablesThe appetite guide PDF. The scanned loss run. Unlocks appetite matching.
- Computer useWorks a website like a person does.
- Model Context ProtocolPlugs into the systems you already run.
- Writes and runs its own codeBuilds the scraper, the database and the report. Unlocks mining the market.
- Works for hours on one jobBuilds and runs a whole system, start to finish. Unlocks your own money calculator.
Sources: ChatGPT GPT-4 100K 200K 1M Scans Computer use MCP Code Hours
Five steps. Every play.
Different data, same shape. The system does the first four. You do the fifth.
Public records. Your book. Appetite guides. Commission schedules.
Names matched. Duplicates gone. One record per business.
Every account against every market. Every prospect against your niche.
Ranked by what it’s worth to you.
Call. Re-market. Cross-sell. You decide.
Mine the market.
Who needs a trucking agent this week?
The federal motor carrier agency publishes every trucking company’s authority, insurance filings and cancellations as open data. It updates every morning.
A new authority needs a policy on file before it can haul. A cancellation notice starts a 30-day clock. The system reads all of it overnight, adds fleet size and safety record, and ranks who to call.
Possible since Feb 2025The math
Illustrative. Say 5,000 new filings a month in your states. A person spends about 5 minutes on each: look it up, check the fleet, find a phone number, judge the fit. 5,000 × 5 min ≈ 417 hours. The system reads them overnight. You spend about an hour on the top of the list.
Book × carrier appetite.
Which accounts are with the wrong market?
Every carrier sends an appetite guide: classes, states, fleet size, loss limits. Nobody reads twenty guides against eight hundred accounts.
A model reads the guides. A database holds your book. Every account gets a fit score with every market. The mismatches show up on their own.
Possible since Mar 2024| MKT A | MKT B | MKT C | MKT D | MKT E | MKT F | |
|---|---|---|---|---|---|---|
| Sal’s General Contracting | 2 | 1 | 4 | 0 | 3 | 1 |
| Ridgeline Freight | 0 | 4 | 0 | 3 | 1 | 0 |
| Bay Café Group | 3 | 0 | 1 | 1 | 4 | 2 |
| Oak Ridge Builders | 4 | 1 | 3 | 0 | 2 | 1 |
| Elm St Apartments | 1 | 0 | 2 | 2 | 0 | 4 |
The math
Illustrative. 800 accounts × 20 appetite guides = 16,000 checks. At one minute each, a person needs about 267 hours, and the guides change before they finish. The system re-checks every account whenever a guide changes. You review the handful it flags.
Cross-sell your own book.
What is every client missing?
Your book already knows. GL with no umbrella. Trucks with no cargo. Payroll up 40%, workers’ comp still on last year’s estimate.
A model reads every declarations page and application. It lines up what each client has against what an account like theirs should carry. You get a list, not a hunch.
Possible since May 2023| GL | AUTO | WC | UMB | CARGO | IM | CYBER | GAPS | |
|---|---|---|---|---|---|---|---|---|
| Ridgeline Freight | 2 | |||||||
| Bay Café Group | 2 | |||||||
| Oak Ridge Builders | 1 | |||||||
| Elm St Apartments | 2 |
The math
Illustrative. 800 accounts, about 12 minutes each to read the policies and spot what is missing: about 160 hours. The system reads them overnight. You spend a couple of hours on the gap list.
Your own money calculator.
Which market pays me the most per hour I spend?
Commission is half the story. The other half is your time: no portal, endorsements by email, claims that come back to your desk.
Put your commission schedules and your service hours in one database. Every account gets a real number: what you keep per hour, by market. Refreshed every night.
Possible since May 2025The lowest commission pays the most. Nobody sees that on a commission statement.
The math
Illustrative. Market A: $13,500 commission ÷ 60 hours = $225 an hour. Market C: $12,150 ÷ 40 = $304. Market B: $10,800 ÷ 18 = $600. Doing that by hand for 800 accounts and 20 markets means tracking every hour against every policy. Nobody keeps that spreadsheet. A system keeps it without being asked.
Same engine. Your niche.
A different niche means different data, a different market and a different book. That’s the experience part. We build it around yours.
New authorities. Cancellation notices. Fleets that just grew.
New openings. Liquor licences with no liquor liability.
Who pulled permits this week, and how big the jobs are.
Unit counts. Buildings that just changed hands.
New practices. New locations. New providers.
Revenue, staff and board size. Who needs D&O.
Big data. Clean hands.
- Public records and open data only.
- Every site’s terms and robots rules followed.
- Calls scrubbed against Do Not Call.
- Every opt-out honored, for good.
- Texts only with consent.
- Your clients’ data stays in your systems. Never sold.
Next. Bigger.
These are forecasts, not promises. Each one has a source. Under each is what it could mean for an agency.
How to read this
METR measured the length of tasks, timed by how long they take a skilled person, that AI models finish on their own about half the time. Across six years the length doubled about every seven months, and it was near an hour in early 2025. The dashed line simply extends that trend. METR has since updated its estimates, so read the dates as direction, not schedule.
Task length doubling every ~7 months. METR expects agents that finish weeks-long projects within a decade.
For an agency: a whole renewal remarket, start to finish, waiting for your sign-off.GPT-3.5-level performance cost over 280 times less in Oct 2024 than in Nov 2022.
For an agency: every account against every market, every night, for pocket change.By 2028, at least 15% of day-to-day work decisions made by agents, up from 0% in 2024.
For an agency: carriers and vendors run agents too. Submissions go agent to agent.Powerful AI, smarter than a Nobel winner in most fields, “could come as early as 2026.”
For an agency: expert coverage review on every account, not just the big ones.Built for the next model.
We don’t bet on one model. We build so that every better one makes your agency better on day one.
Every procedure lives in plain words, once. A smarter model follows the same procedures better. Nothing to rebuild.
The model is a part, not the product. When a better one ships, it slots in. We’ve already done it more than once.
One record per business. Every action on the file, with the email that caused it. That history is what the next model reads to know your agency.
As models earn trust, you decide what still needs your OK. Binding, money and anything leaving the agency stay with a licensed person.
Left behind is a choice.
Dream big. Doux it.
Bring your niche and your book. We’ll show you what they can tell you.