Agent Acceptance Test · 10 business days

Before your agent goes live,
get a defensible go / no-go.

The Agent Acceptance Test stress-tests an AI agent before production and returns an evidence-readiness dossier for General Counsel, CISO, and underwriter review. You walk into your risk committee with the one thing it actually wants: proof the agent can be defended, in writing, before it touches a customer.

Not a readiness assessment. Not advice. A clear go, conditional-go, or no-go, backed by adversarial scenario results and a record built to hold up under challenge.

Request your Acceptance Test →See what gets tested →
What it gets you

Clear the internal gate
before it blocks you.

A regulated AI agent does not ship because the engineering works. It ships when the risk committee approves it, and that committee will not approve a regulated agent without evidence it can be defended. That requirement is where most agents stall.

The Agent Acceptance Test produces that evidence in ten business days. You get a faster approval, a cleared internal gate, and a go / no-go your General Counsel, CISO, and underwriter can each read and sign. The defensible answer becomes an asset you own, not a question you keep deferring.

10 days
Intake to go / no-go memo
3 readers
GC, CISO, and underwriter, one dossier
1 decision
Go, conditional-go, or no-go, in writing
What the test covers

Five components.
One production decision.

01

Adversarial scenarios

Between 75 and 150 scenarios run against the agent the way opposing counsel would read it: edge inputs, prompt manipulation, conflicting instructions, and the unhappy paths your demo never showed. Each result is recorded.

02

Policy-gate mapping

Every action the agent can take is mapped to a gate: allow, require human review, escalate, or deny. The map shows your risk committee exactly where authority stops and a human decision begins.

03

Escalation and human-in-the-loop review

We trace where the agent should hand control back to a person and confirm it actually does. The escalation matrix is the document a regulator asks for first.

04

Evidence-gap report

A line-by-line read of what the agent can prove about its own behavior today versus what a court-ready record requires. Every gap is named, with the remediation that closes it.

05

Go / conditional-go / no-go memo

One decision, in writing, with the evidence behind it. Go means ship. Conditional-go names the gates to close first. No-go names what makes the agent indefensible today. No hedging.

The deliverable

The Evidence Readiness Dossier.
Not a slide deck.

The dossier collects the go / conditional-go / no-go memo, the evidence-gap report, and a remediation roadmap into a single record your risk committee can act on. It can be delivered under outside-counsel privilege so the findings stay protected while you close the gaps.

Agent responsibility boundary
Workflow risk map
75 to 150 adversarial scenarios
Policy-gate map
Tool-call failure analysis
Human escalation matrix
Evidence-gap report
Remediation roadmap
Go / conditional-go / no-go memo
Proof

See what sealed evidence
actually looks like.

The dossier names what your agent must seal. This is one of those records: hash-chained, Merkle-rooted, and anchored to the public Sigstore Rekor transparency log. Anyone can verify it without trusting Verdict. It is built to support self-authentication and is aligned with FRE 902(14).

Live sealed evidence record
ser_210481d244494de36ff253f21efc31ee
See a sealed, independently-verifiable evidence record →
Pricing

Three ways in.
One defensible answer.

Single agent

Agent Acceptance Test

$7,50010 business days

One agent, one workflow, one production decision. Built for teams that need a defensible answer before pilot, purchase, launch, or expansion.

  • Adversarial scenario suite
  • Policy-gate map
  • Evidence-gap report
  • Go / conditional-go / no-go memo
GC + CISO + underwriter

Regulated Agent Acceptance Dossier

$15,00010 to 15 business days

For agents touching legal, insurance, finance, healthcare admin, claims, eligibility, or customer rights. The full dossier, packaged for a risk committee and an underwriter.

  • Expanded compliance map
  • Human escalation matrix
  • Evidence-readiness dossier
  • Board-ready go / no-go memo
After launch

Evidence Readiness Retainer

$3,000/monthmonthly drift watch

Your agent's behavior changes every time a model, prompt, policy, tool, or workflow changes. The retainer re-tests on each change and keeps the go / no-go current.

  • Monthly regression
  • Drift memo
  • New edge cases
  • Updated remediation list
The engagement

Ten days.
Intake to readout.

Days 1–2

Intake

You hand over the agent, its tools, its workflow, and the production decision in front of it. We map the responsibility boundary and the workflow risk map. No build work. No access to your customers.

Days 3–7

Adversarial run

The scenario suite runs. Policy gates, tool-call failures, and escalation paths are exercised and recorded. Findings accumulate against a fixed taxonomy, not a consultant's opinion.

Days 8–9

Dossier

Results compile into the evidence-readiness dossier: the evidence-gap report, the remediation roadmap, and the go / conditional-go / no-go memo. Delivered under outside-counsel privilege on request.

Day 10

Readout

A working session walks your General Counsel, CISO, and underwriter through the dossier and the decision behind it. You leave with the evidence to clear your internal gate and a remediation roadmap if any gate stays open.

Walk into the committee
with the answer already in hand.

Regulated AI agents touching legal, insurance, finance, healthcare admin, claims, eligibility, or customer rights. Tell us which agent needs a defensible go / no-go and we will reply with the testing scope and the material list.

Request your Acceptance Test →See a sealed evidence record →