Make AI decisions your organisation can explain.

What we work on

AI use and accountability

Establish where AI is used, what it is intended to do and who owns the decision to use it. We help distinguish experiments from approved uses and identify the changes that should trigger a fresh review.

This can include an AI inventory, ownership model, intake process and decision record that fits existing product, procurement and risk processes.

Risk and impact assessment

Examine how a system could fail, who could be affected and what would make the consequences more serious. We connect those findings to design choices, operating limits, testing and escalation.

Where personal data is involved, we work with the relevant privacy owner on the information needed for data-protection assessment. A single generic AI risk score is rarely enough to explain what needs attention.

EU AI Act readiness

Map the system’s intended purpose, the organisation’s role and its connection to the EU market. From there, we help identify relevant obligations and the evidence needed to support implementation.

The work can include a role and applicability assessment, a gap review, technical-documentation planning and an implementation roadmap. Ambiguous legal questions are identified for specialist legal input rather than hidden in a checklist.

Physical AI governance and safety

For AI that senses, decides or acts in the physical world, connect governance with the operating environment and the consequences of failure. We help define intended use, operating limits, safety responsibilities and the evidence needed before a trial or wider deployment.

The work can include scenario and failure analysis, human-intervention requirements, evaluation planning and an evidence register for specialist safety review. For teams exploring NVIDIA Halos, we can support an initial applicability review and organise the integration questions, documentation and evidence needed for discussions with the relevant programme or assessment provider.

This connects with our embedded engineering work on NVIDIA and Hailo platforms. Platform selection, AI performance and system safety are examined together; formal inspection or certification is carried out through the applicable assessment route.

Policies and operating controls

Write policies people can use: which uses are permitted, what information may be entered, who reviews an exception and what must happen before a system goes live.

Supporting procedures can cover supplier selection, evaluation, human oversight, incident handling, changes and retirement. We build on existing controls where they already work.

For AI agents, we also examine what the system can access and do: tool permissions, approval points, records of actions and a way to stop or recover from an unwanted action. These controls belong in the workflow as well as the policy.

Supplier and customer assurance

Turn supplier claims into questions that can be answered with evidence. We review documentation, deployment arrangements, contractual dependencies and the practical implications of service changes.

For product teams, we help organise the evidence behind customer assurance responses so that answers remain consistent with the actual system.

Management systems and audit preparation

Support the development and improvement of AI governance within management systems, including ISO/IEC 42001 and its connections with information security under ISO/IEC 27001. The aim is a system of work that the organisation can maintain.

We can help with scope, gap assessment, responsibilities, documentation, evidence organisation and preparation for external assessment. Certification decisions remain with the certification body.