Commercial and speciality
From broker submission to a price you can talk through
Fleet, property, liability and speciality risks don't arrive as tidy forms. Your AI client reads the submission, Ptally prices it with your rating model, and your underwriters or brokers adjust the quote in conversation. Every quote and amendment is recorded.
Documents in. A price to work with out.
The broker's pack goes into the chat. Claude fills in the model's inputs, Ptally prices and saves the quote, and the underwriter's experience credit and the sale are recorded against it.
What it takes
Experience-based pricing, without the rekeying
Commercial pricing depends on people. Ptally takes the slow parts off them and keeps the judgement with them.
Unstructured
Submissions aren't forms
Risks arrive as broker emails, vehicle schedules, statements of values, loss runs and proposal forms. Someone has to turn them into rating inputs.
Judgement
The underwriter makes the call
The model gives a consistent technical price. The underwriter or broker decides what to offer, within their authority.
Explainable
Every price has to stand up
Brokers ask why the price moved. Managers, auditors and capacity providers need to see how it was built and who changed it.
Time
Fewer risks, more work per risk
Each risk is worth the effort, but nobody wants to spend a day rekeying a 40-vehicle schedule before they can quote.
How it works
Read, price, pivot, agree
Our consultants build your rating model from your rating guide, rater spreadsheets or filings. Your team works with it from the AI tools they already use.
01 · AI client
Read
Drop the submission into Claude or ChatGPT. It reads the documents, maps what it finds to your model's inputs and tells you what's missing, so you check the inputs instead of typing them.
02 · Your rating model
Price
Ptally runs your model on those inputs and returns the technical price, with any referrals and declines. It's the same deterministic engine that prices our high-volume products.
03 · In conversation
Pivot
Ask what-ifs in plain English: a higher excess, telematics fitted, two vehicles removed, a different limit. Each one is priced by the model and compared with the last, so you can work through options with the broker.
04 · Audit trail
Agree and record
Apply your experience-based judgement as a price change with a reason, then mark the quote sold. Each step is logged with who made it and when, and the engine's original price is kept.
A full history for every risk
Every quote shows who priced it, every price change and why, and when it was sold, whether the change came from the dashboard, the API or an AI client acting for an underwriter. The engine's original price is always kept beside the agreed one.
Human in the loop
Pause for an underwriter where you choose
Refer rules mark the risks that need a second look, such as those above an underwriter's authority. A referred risk is still priced, but it waits for the right person before it is offered.
Your authority
Rules match your referral levels
Refer rules encode the risks that need a senior underwriter. In the commercial auto example, 50 vehicles or more, or five or more with no claim-free years.
Priced, then paused
The underwriter starts from a number
A referred risk still gets a technical price. The response lists each rule that fired and the values it used, so the reviewer sees why at once.
Recorded
Every decision has a name on it
The underwriter accepts the quote, changes the price or marks it not valid. Each change is logged with who made it, when and why. The engine's price is kept.
Optional
Tighten or relax as you learn
Refer rules are versioned with the rates. Add, tighten or remove one in a draft, run the tests, then approve and publish it like any other change.
Refer step in the commercial auto example
{
"id": "auto_refers",
"key": "auto_refers",
"step": "refer",
"name": "Commercial auto refers",
"def": false,
"items": [
{
"id": "large_fleet",
"name": "50+ vehicles",
"exp": "{{vehicle_count}} >= 50",
"def": false
},
{
"id": "recent_losses",
"name": "No claim-free years",
"exp": "{{fleet_years_claim_free}} == 0 && {{vehicle_count}} >= 5",
"def": false
}
]
}Response for a five-vehicle fleet, trimmed
{
"result": 7279.4,
"valid": true,
"refer": true,
"refers": [
{
"step": "auto_refers",
"name": "No claim-free years",
"fleet_years_claim_free": 0,
"vehicle_count": 5,
"exp": "{{fleet_years_claim_free}} == 0 && {{vehicle_count}} >= 5",
"result": true
}
]
}The commercial auto example with no claim-free years and five vehicles. It prices at 7,279.40 and is held for an underwriter, with the rule and the values that fired it.
Three ways to price
Commercial, annuity or commoditised?
The same engine prices all three. What changes is where the inputs come from and who makes the final call.
Commoditised, high volumePersonal lines and embedded coverFind out more →Commercial and specialityFleet, property, liability and specialityYou're hereAnnuityStandard and enhanced, priced from documentsFind out more →
InputsStructured, from forms and APIsBroker submissions, schedules and loss runsHealth questionnaires, prescriptions and pension statements
Who decidesThe model, with no one in the loopAn underwriter or broker, working in an AI clientThe model prices. Underwriters review referrals and evidence
The priceFinal, in milliseconds, the same every timeA technical price in minutes, to adjust within authorityQuoted on the basis in force, guaranteed for a period, rechecked at completion
What's recordedEvery quote, with its inputs and model versionEvery quote, amendment and model changeEvery field and its source page, the basis version and the hash at signature
Quote your next submission in minutes
Send us your rating guide or rater. We'll build the model and show you a live submission priced in Claude or ChatGPT.