Pricing
Priced per alert worked. Published, so you can model it before a call.
One price per triage run. Sandbox is free on test data. Team is $1,500 a month with 2,000 runs included. These are early prices, and the first design partners will help set them.
Your monthly bill, from your own numbers.
Move the sliders. The calculator picks the plan that fits. Above 10 analysts or 20,000 runs a month it hands you to Enterprise, which is priced by contract.
Estimated monthly bill
Team plan$2,700.00 / month, Team plan
Team, for the live model, more than 100 runs and more than 2 seats. Up to 10 seats.
- Base, includes 2,000 runs
- $1,500.00
- 2,000 more runs × $0.60
- $1,200.00
- Total
- $2,700.00
- Effective price per alert
- $0.68
Modeled Analyst time this could give back: $51,282 a month.
4,000 alerts × $28.85 of L1 labor × 44% time saved, out of about $115,385 a month today.
The 44% is the twelve-alert shift modeled on A day in the queue. A pilot measures your real number. Freed time is capacity, not a headcount cut.
Three plans. Start free, pilot free.
Sandbox
$0 / month
100 runs a month, rules model, test data
Try the workflow on test data with the rules model.
- 100 triage runs a month
- Rules model only
- CSV import
- Audit trail and QA
- 2 seats
Team
$1,500 / month
2,000 runs included, then $0.60 a run
For compliance teams of 2 to 25 analysts.
- 2,000 triage runs a month, then $0.60 each
- Live model: Anthropic, OpenAI, or Azure OpenAI
- REST API and CSV import
- QA sampling and autonomy controls
- 10 seats
Enterprise
Custom
Contract volume, single-tenant in your Azure subscription
For banks and larger programs with their own cloud and model governance.
- Single-tenant deployment in your Azure subscription
- Your model endpoint, your data residency
- SSO with Microsoft Entra ID
- Model risk documentation pack
- Custom typologies and SOPs
Lookback reviews. One-off review of historical alerts or transactions, such as a lookback required by a consent order. Priced per batch after a scoping call.
Questions buyers ask first.
Something missing? Email parikshit.ambhore@rice.edu. I answer pricing questions myself.
What counts as a run?
One alert worked by the agent, whatever the outcome: a recommendation, an abstention, or a policy lock that stops the alert before the model. Re-sending an alert ID that is already in the workspace returns a duplicate and is not billed. Opening, deciding, reviewing and exporting are never billed.
Is a re-run billed?
Yes. Running an alert again, for example after you change a threshold or new transactions arrive, is a new run with its own entry in the audit log. That keeps the bill and the audit log counting the same thing.
What do live model calls cost, and who pays for them?
On Team, model usage is part of the run price. An average demo alert sends and receives about 3,600 tokens, which is about $0.011 at Claude Sonnet 5.5 list prices1, so the $0.60 run price is mostly the product, not the tokens. On Enterprise the model is your own Azure OpenAI deployment and the tokens are on your Microsoft bill.
Where does our data live?
Team runs on the hosted edition; the providers behind it are listed on Security. Enterprise deploys single-tenant in your own Azure subscription, in the region you choose, with your model endpoint, so alert data stays inside your boundary. Assay trains no models on your data. Production data waits for a signed data processing agreement.
Are pilots free?
Yes. A design-partner pilot is free for its whole length: shadow mode on 30 days of your historical alerts, in your own Azure tenant, measured against your own dispositions, while your analysts work exactly as they do today. Apply for one of three slots.
Can we cancel at any time?
Yes, from the billing portal. The plan runs to the end of the period you paid for, then the workspace returns to Sandbox. Your audit log stays exportable.
Why per run and not per seat?
Alert volume is how your team already measures its workload, so the bill moves with the work the agent does. Seats are included rather than billed one by one: 2 on Sandbox, 10 on Team. At the Team rate the included runs cost $0.75 each, against about $28.85 of modeled L1 labor per alert Modeled.
What changes at L3, auto-close?
The price does not. An auto-closed alert is a run like any other and gets its own audit entry. What changes is the bar: an alert type reaches L3 only after 98% QA agreement on 2,000 reviewed alerts and written sign-off, and only closes at 0.95 confidence or more qualify. Today the sign-off is an owner or admin attestation, not an automatic check against the QA data. Nothing starts at L3.
Measure it on your own alerts before you pay for it.
A shadow-mode pilot runs on 30 days of your history, free, and reports agreement by alert type against your own dispositions.