Use case

Automate customer onboarding with an agent

The problem

Onboarding stalls on documents. Identity is eyeballed, supporting documents are chased by email, and a file sits incomplete for days before anyone notices what’s missing.

Hand it to an agent without verification tools and it just trusts whatever it’s handed — confidently passing a document that doesn’t actually establish identity.

How Stipple helps

Give the agent one tool. verify_identity reads each document a customer submits (its type, the fields it shows, its issue date) and a deterministic engine tallies them against the Australian 100-point check (or the AUSTRAC safe-harbour rules) — returning whether identity is established and exactly what’s still missing.

The agent orchestrates, the engine scores, and a human owns the decision — every step attributable. It reports coverage, not a verdict: it tells you the documents add up (or don’t), never that a person is “cleared”. The same pattern extends to income proof and screening as those tools land.

The call

text
=== Onboarding case: applicant (afp_100_point) ===

  [agent ] case.created        applicant
  [agent ] documents.collected passport.pdf, licence.pdf
  [agent ] tool.call           verify_identity (2 documents)
  [engine] document.scored     passport.pdf   australian_passport   counted +70 (primary)
  [engine] document.scored     licence.pdf    drivers_licence       counted +40 (primary)
  [engine] identity.tally      110/100 points, 2 primary doc(s)
  [agent ] decision            IDENTITY_VERIFIED -> proceed to screening & income checks

The identity gate becomes automatic and auditable: complete files flow straight through, incomplete ones come back with the precise gap, and every decision leaves a record a system of record can store.