Skip to content
, Home

Expertise

AI & Agent
Quality

I help teams verify what their AI workflows actually achieve, and decide when one is ready to release.

Discuss your AI workflowSee the offers

01Scope

What it covers.

  • Agent evaluations

    Scenarios with expected outcomes, built from real use and known failures.

  • Workflow and handoff testing

    What happens between tools, agents and approval steps, where failures hide.

  • Deterministic checks

    Checks on the completed action and the resulting system state, not on how the answer reads.

  • Evaluation harnesses

    A repeatable run that classifies failures by cause, so each class points at an owner and a fix.

  • Release criteria

    Results reported per release, in a form product and engineering can act on.

02Offers

How we can
work on it.

Each offer starts with a problem you need to solve. We agree the scope, outputs and measures of success around your priorities.

  • AI workflow evaluation

    Check that your AI actually gets the job done.

    When it applies: An agent can give a convincing response without completing the task. Failures can also hide between tools, agents and approval steps.

    See the full offer: AI workflow evaluation

03Evidence

Evaluation work
and writing.

A case study from my employment, the dimension it maps to, and what I have published on the subject.

  1. Case study 01

    Making AI evaluations trustworthy at an AI-native language-technology platform.

    I reviewed the platform’s AI evaluations to make sure they tested the feature, not just the judges’ criteria. Improving unstable tests and assertions raised the pass rate from just over half to three in four within a month, and new dashboards made the evals’ effectiveness, time and cost, and the success rates of live user journeys, visible.

    Context
    An AI-native language-technology platform · 2026 · Head of Customer Trust, AI workflow quality and release confidence
    Result
    Within a month, the evals’ pass rate rose from just over half to three in four, and the failures that remained were known product issues rather than faults in the tests.

    AI evaluation, Test design, Test reliability, Observability, Customer trust

Related

04More expertise

The other areas.

Need credible verification
of an AI workflow?

Describe the workflow, what it is meant to achieve and where confidence is thin.

Discuss your AI workflow