blankit

Pricing

One plan. Pay for the AI you actually use.

Every firm gets the whole platform — client book, renewals pipeline, document engine, chat assistant, integrations. The subscription includes a monthly allowance of AI credits; heavy AI use bills per credit, and nothing is ever blocked mid-renewal.

The plan

Free during early access

  • Unlimited clients, users, and documents
  • Every feature — nothing is gated behind a tier
  • 5,000 AI credits included every month
  • CA$0.02 per credit after that, billed on your next invoice
  • Canadian-hosted, per-firm isolated database
Start now

How credits work

A credit is a cent of AI compute.

Every AI-powered action — parsing a carrier renewal, comparing booklets, asking the assistant — uses credits in proportion to the actual computation it takes. One credit is one US cent of AI compute at list price: that is the cost basis the meter counts in, and it is why the number is stable — what you pay is quoted and invoiced in Canadian dollars. Roughly:

  • A full renewal analysis — a carrier PDF parsed, the actuarial engine run, the report built — is typically 300–500 credits.
  • A booklet comparison or claims report — usually well under 100.
  • An assistant or chatbot answer — single-digit credits.

Your usage is visible at all times — the firm-wide picture with a per-advisor breakdown for administrators, and every advisor’s own numbers on their settings page. Passing the allowance never blocks anything; it simply bills as overage.

Questions

What happens if we run out of credits?

Nothing stops. Usage past the included allowance is billed at CA$0.02 per credit on your next invoice — no interruptions, no paywalls mid-renewal.

Can we see who used what?

Yes. Administrators see the firm’s period usage, its projection, and a per-advisor breakdown under Settings → Billing & usage. Each advisor also sees their own usage on their settings page.

Is our data used to train AI?

No. Documents are processed only to answer your own request, in Canada-hosted infrastructure, and are never used to train any model. See Trust & security.