Team
$79/user/mo
For one desk getting its first pipeline into production.
- Up to 10 users
- 3 pipelines
- All 39 modules
- 77 starter rules, versioned
- Unlimited Test Bench runs
- Email, API and folder intake
- Hash-chained decision records
SureThink/Pricing
Pricing · billed annually
Priced per person,
not per document.
Per-document pricing makes you ration the thing you bought. You end up deciding which files are worth reading properly, which is exactly the decision the software was supposed to remove.
Growth most firms
$149/user/mo
For several desks, several pipelines, and an audit function that asks questions.
- Everything in Team
- Unlimited users and pipelines
- Four-eyes and authority maps per pipeline
- Observe mode
- Replay with pinned versions
- AI budgets and usage console
- SSO and TOTP enforcement
Enterprise
Let’s talk
For firms whose risk committee needs to read the architecture before the contract.
- Everything in Growth
- Named solutions engineer
- Custom rule-pack authoring
- Security review and DPA
- Residency and retention commitments
- Migration from an existing process
Annual billing. Monthly is available at a 20% premium — we would rather say that than print a monthly price and hide the annual discount in a footnote.
AI usage · metered separately
What the model calls
actually cost.
Model calls cost money. Bury that in a seat price and you are either overcharging the light users or quietly throttling the heavy ones. So it is metered, itemised, and capped by you.
surety_intake$122.90
kyc_screen$54.10
claims_fnol$29.80
legal_intake$5.60
1,942 calls · 2 failed and are counted
avg $0.11 per case · median 4m 12s
0 calls unpriced
Test Bench runs excluded — they never call a model
How it’s counted
- Per call, both ways. A call that failed still cost money, so it still appears. Spend you cannot see is spend you cannot control.
- Priced when it happens. The rate is frozen at write time. A later change to the rate card never rewrites what an old case cost.
- Unpriced is never zero. A model with no rate-card entry reports as unknown. You find out by seeing a gap, not by being quietly wrong.
- Capped per pipeline, per day. At the ceiling the pipeline holds and the queue says why, rather than running up a bill overnight.
Typical is about eleven cents a case on a 60-page surety submission. We will estimate yours from a real file in the working session rather than quoting a range that means nothing.
On every plan
What never sits behind a tier.
The accountability machinery is not an upsell. A four-person team carries the same risk per decision as a four-hundred-person one.
| Capability | Team | Growth | Enterprise |
|---|---|---|---|
| Hash-chained decision records | Included | Included | Included |
| Human approval stations | Included | Included | Included |
| Sourced-confidence grading | Included | Included | Included |
| Test Bench (zero spend) | Unlimited | Unlimited | Unlimited |
| Canadian data residency | Included | Included | Included |
| Pipelines | 3 | Unlimited | Unlimited |
| Four-eyes & authority maps | Single map | Per pipeline | Per pipeline |
| Observe mode | — | Included | Included |
| Replay with pinned versions | — | Included | Included |
| SSO | — | Included | Included |
| Custom rule-pack authoring | — | — | Included |
Before you ask
The questions serious buyers ask.
What happens if we cancel?
You export everything — documents, structured output, decision records with their chains — in a format that stands on its own. The records are evidence about your business, not a feature of our subscription, and holding them hostage would be indefensible.
Do we need engineering to set this up?
Not for the pipeline itself. The builder is for the person who owns the process. You need engineering only to connect a downstream system by webhook, and that is typically an afternoon.
What if the pipeline is wrong?
It is designed to be wrong in one direction: it stops rather than guesses. When it does misread something, your reviewer corrects it and the file keeps both figures with her name on the change — so the error becomes evidence rather than a silent bad number.
How long until it’s doing real work?
About a week to a first pipeline running on live files in observe mode, and a further two to four weeks before most teams let it act unsupervised on routine cases. We would be suspicious of anyone promising faster.
Can we keep our own rules?
You should. The 77 starter rules are a starting point for contract surety specifically; every firm edits them. Rules are versioned, so changing one does not rewrite the cases decided under the old version.
Is our data used to train models?
No. Not ours, not a vendor’s, not in aggregate, not anonymised. There is no setting for this because there is no other mode.
Start with one workflow
Start with the one workflow
you have to answer for.
Thirty minutes on your own process. We'll estimate your real cost per case from a real file rather than quoting you a range that means nothing.
- A schematic of your workflow, yours to keep
- The rule pack it would screen against
- Your authority map, role by role
- A sample decision record and a cost per file