Independent evidence assurance

Independent review of what your AI and agent systems can actually evidence.

A bounded independent review that states what the evidence shows, what it does not show, and what cannot be determined from the available record. Delivered as a written finding you can hand to someone outside your team.

Opens your email app. If that does not work, email wbrissey@gmail.com. There is no web form processor and no online checkout on this site.

What this is not

Not an observability dashboard. A dashboard shows spans; it does not produce an independent exportable finding.
Not generic AI strategy consulting. The unit of work is your evidence, not a roadmap.
Not an IAM product. We review what an agent can actually do, not what identities exist.
Not penetration testing. Different question, different method, different report.
Not an AI governance maturity score. A number on a scale is not evidence.
Not consensus voting or model averaging. Averaging destroys the disagreement signal this work exists to preserve.

A — D

Four families. One review discipline.

How it works

01

Fit check

A short conversation about the situation and the record. If it is not a fit, that is said plainly.

02

Evidence intake

You provide the exported record. Scope and boundaries are agreed in writing before review starts.

03

Bounded review

Independent examination inside the agreed scope. Gaps in the record are recorded as gaps.

04

Findings and delivery

An exportable finding with explicit limitations, released by a human reviewer.

05

Next-step proposal

Where the finding warrants further work, a scoped proposal. Where it does not, none.

Read the full process

Trust and boundaries

Trust here comes from scope discipline, not from logos.

No customer counts, certifications, or case-study metrics are published, because none have been externally verified. What is committed is how a review is bounded and what it refuses to claim.

01Independent. The review is not performed by the team that built or operates the system.
02Bounded. Scope is fixed before the review begins and stated on the face of the finding.
03Evidence-preserving. Disagreement and ambiguity are recorded, not resolved away.
04Explicit limitations. Every finding states what could not be determined from the record.
05Human release. A person adjudicates and releases the finding. No automated verdicts.

Who does the work

William Edgar Brissey

Independent specialist, US sole proprietor. No company is formed for these engagements. If the reviewer previously built or operates the system under review, the engagement is declined — independence is the point. Professional-liability status is stated in writing before any engagement is signed, not assumed.

Current public state of this site. Offers and prices below are the accepted scope grid. Reply by email is the live contact path. There is no self-serve checkout tile on this homepage. After a written Statement of Work, payment instruction follows. A Contra one-time link exists for founding Band A incident reconstruction; guest checkout without a Contra account has been verified. A real charge settling, and payout after funds arrive, have not been exercised. Other offers use a direct invoice whose collection path is also not proved end-to-end. This site is publicly deployed at evidence-assurance.netlify.app.