AI can act.
Can you prove it acted correctly?
Organisations can have policies, permissions and monitoring in place and still struggle to prove those controls actually worked. WALCROFT independently tests whether the evidence supports the assurance being placed on AI and autonomous agents.
Not “do you have controls?”
“Can you prove they worked?”
AI governance often tells leaders what should happen. Logs and dashboards tell them what happened. WALCROFT tests whether the evidence is strong enough to justify the assurance management is relying on.
Declare
Management states the control and the assurance it believes it provides.
Evidence
Relevant records, approvals, logs and outputs are assembled for review.
Test
WALCROFT challenges whether the evidence actually corroborates the claimed control.
Assure
A clear independent judgement identifies confidence, limitations and assurance gaps.
What assurance actually looks like.
Illustrative demonstration only. It shows the WALCROFT approach without exposing the underlying assessment methodology.
An autonomous agent was authorised to execute a procurement action. The final transaction value materially exceeded the intended value.
Approval controls prevent material unauthorised spend.
Partial
Evidence supports parts of the control chain but does not fully corroborate the assurance claimed.
Limited Assurance
A material assurance gap remains between control design and evidenced operation.
Make uncertainty visible.
WALCROFT does not turn weak evidence into false confidence. Where assurance cannot be substantiated, the gap is made explicit.
The objective is not to create another compliance dashboard. It is to give boards, executives, customers and assurance functions a defensible view of what can — and cannot — be relied upon.
Independent assurance for the agentic era.
Independent
Assurance is separated from the team making or operating the AI control.
Evidence-led
Judgements distinguish between what is declared, supported, sample-tested and independently verified.
Decision-useful
Material gaps are surfaced clearly so senior leaders can act rather than interpret another technical dashboard.
Help define the next standard of AI assurance.
WALCROFT is speaking with a small number of organisations using or preparing to use consequential AI and autonomous agents. Design partners can explore the assurance approach against a suitable use case and help shape what independent AI assurance needs to deliver in practice.
Design-partner participation does not imply certification, regulatory approval or endorsement.