Constraint 01
Identity
Verification hands off, steps up and expires. Each transition is a chance for the agent to lose everything it accumulated.
Your customers are already sending agents. The question is whether the journey lets them finish.
BFSI journeys carry constraints consumer web does not: identity, consent, one-time passwords, mandates, disclosures, audit. Every one of those is a place an agent loses state and returns nothing. Webzero measures each journey with the Webzero Agent Readiness Index, then rebuilds what stops completion.
The measurement is common across BFSI. What changes by segment is which journeys carry the volume, which break first, and what a failed run costs you.
Transactional journeys raise the stakes on every property the index measures. An ambiguous control in a lending form costs a lead. An ambiguous control in a mandate or a merchant payout costs trust, and possibly a reversal.
An action taken twice, or the wrong action taken once.
Non-idempotent endpoints plus retry-happy agents is the combination that turns a failed run into an incident rather than a metric.
Authorisation, mandate consent and audit trails have to hold whether the actor is a person or an agent acting for one.
Actions must be scoped, confirmation-bound and idempotent before an agent is allowed anywhere near them.
Platforms inherit the readiness of their partners. A partner-hosted step is still your failed journey.
Verification hands off, steps up and expires. Each transition is a chance for the agent to lose everything it accumulated.
Disclosures and authorisations must be provably given. An agent that infers its way past a consent gate is a compliance event.
A retried transaction is not a retried page view. Actions need to be scoped, confirmation-bound and idempotent.
Regulated businesses cannot act on a vendor claim. Every reading ships with reproducible run logs and trajectories.
A credit score journey inside a large diversified financial services group. Same probe, same task, run twice.
Simulation runs in a permitted environment against synthetic identities. Nothing touches a real customer record, and every run leaves a trajectory log your engineers can reproduce.
Pick two journeys that carry real volume, agree the permitted environment and the synthetic identities.
A held-still fleet runs the task library against the journeys, end to end, with every trajectory logged.
A w0 score across five pillars, with the specific breaks named and ranked by what they cost.
Remove what stops completion, re-run the identical probe to prove the movement, then watch for drift.
We measure them, show you where agents lose the thread, and give you a w0 score with the run evidence behind it.