Services — Daisy Leung, Evidence-first AI Workflow Architect

Three bounded ways to make AI work inspectable

Start with the decision that needs clearer evidence. Each offer leaves a practical map, explicit limits, and a next step your team can inspect.

AI Workflow Reliance Sprint

Turn a useful but hard-to-review AI workflow into an inspectable path for one consequential decision.

Best fit: Your team has an AI workflow that produces output, but the claims, evidence, assumptions, and approval point are difficult to review.
Working outcome: A bounded workflow map and verification plan that makes the next reliance decision easier to inspect.

This is not a guarantee that every output is correct.

Your team keeps decision authority for consequential use.

Decide what an agent should do, what it may access, and where a person must remain in control.

Best fit: A proposed agent is accumulating tools, knowledge, state, or automation without a clear boundary for what it is actually meant to own.
Working outcome: A build-ready capability architecture with explicit tool, knowledge, state, orchestration, and approval boundaries.

This offer defines an evidence-led architecture and decision.

Implementation is included only when separately agreed.

Understand whether an AI-built or AI-modified repository supports the claims being made about it.

Best fit: A repository has changed quickly, but review findings, decision ownership, and the evidence needed before remediation are unclear.
Working outcome: Validated findings routed to the right decision owner, with a verification plan and maintainable next-step baseline.

A review supports scoped decisions from the evidence exercised.

It is not a universal security certification.

A useful first conversation

Start with non-sensitive context. We can identify the claims, review burden, decision owner, and evidence that would change the next step.