AI governance advisory
Choosing a tool, defining uses, preparing a strategy, setting boundaries for adoption, or bringing order to practices that are already in place.
Three decision stages
Understand the situation
Start from the organisation’s needs, its existing uses and the tools under consideration.
Separate what is worth developing, what needs more caution, and what probably does not need AI at all.
Make choices
Compare tools and use cases. Test what is worth testing. Set limits.
Decide where AI belongs, where human review still needs to remain, and which uses should be set aside.
Turn choices into governance
Translate decisions into rules, responsibilities and usable processes.
The result may be a few simple principles or a broader framework for managing AI adoption across the organisation.
Governance through Coherence
Governance through Coherence asks an additional question: whether an organisation’s actual uses of AI remain consistent with its mission, its commitments and its relationship with its publics. It’s the approach I’ve developed to examine these decisions.
It complements legal and regulatory analysis without replacing it: a use can be fully permitted and still be difficult for the organisation itself to justify.
I use coherence analysis to examine that space between what is permitted, what is technically possible and what the organisation actually chooses to do. The analysis draws where relevant on the EU AI Act, the NIST AI Risk Management Framework, ISO/IEC 42001 and applicable rules on data, rights or transparency.
Ways of working
-
AI workshop
A half-day working session on one precise question, chosen from a catalogue. Each workshop ends with a document the organisation keeps.
View the workshop catalogue
Understand the situation
Workshop You leave with Inventory of uses A register of uses and decisions Understanding generative AI One page of team rules Make choices
Workshop You leave with Data, rights and external tools (ChatGPT, Copilot, Gemini) A list of what is allowed, forbidden or needs approval Choosing and testing tools A shortlist, a test protocol and selection criteria Sensitive material and difficult cases A test protocol and a fallback for when a tool refuses or distorts content AI, creation and images Rules on disclosing AI use and on commissioning artists Turn choices into governance
Workshop You leave with Who decides? For leadership teams and boards Decision roles and a policy outline Example
A museum wants to allow ChatGPT, Copilot or Gemini for certain tasks, but teams have not agreed which data may be entered, which uses are acceptable or when human review must remain mandatory.
Formats
Within one organisation, on site or remote in two 90-minute sessions, or at a network’s meeting.
-
AI governance diagnostic
Map current uses, tools, risks and responsibilities, then identify the priority decisions.
Example
Several teams in a foundation already use AI assistants to summarise files, prepare content or work with internal information. Leadership does not know exactly which tools are being used, what data enters them or who checks the outputs.
Deliverable
A map of actual practices, the main areas of risk and a prioritised set of decisions to take.
-
AI strategy & governance framework
Decide which uses and tools to develop, then define the conditions for their adoption.
Example
A cultural institution is considering several tools for communications, ticketing, documentary research and translation. It needs to decide which ones to test or purchase, what data they may receive and what conditions should apply before wider deployment.
Deliverable
A roadmap covering priority use cases, tools to test or rule out, selection criteria, adoption conditions and responsibilities.
-
Continuing support
Address new decisions as tools and uses evolve.
Formats
A one-off question about a specific use, answered in a two- to three-page written note; an annual review of decisions and the register; or regular support for larger organisations.
Example
An organisation has adopted its first AI policy. A few months later, new questions arise around automatic note-taking, image generation, candidate analysis and AI features added to software already in use.
Deliverable
Documented, consistent decisions over time without rebuilding the governance framework for every new case.
-
Research, briefings & network work
Build an evidence base for a decision, policy, network or sector.
Example
A network of cultural venues wants to understand how its members already use AI, which problems recur most often and how other organisations or countries are responding.
Deliverable
A structured survey, comparative analysis and a briefing that can be used by the network’s leadership, members or partners.
At a network event
A talk, a panel or a workshop open to attendees, on how organisations decide their uses of AI.
Process
- 1Initial conversation
Understand the question, the context and the organisation’s starting point.
- 2Scope
Define what will be examined, the outputs and the timetable.
- 3Work
Interviews, document review, workshops, use-case analysis or research.
- 4Delivery
The agreed outputs may combine written recommendations, documented decisions and a working session with the people involved.
- 5Follow-up
Where useful, a defined point to return to implementation or address new questions.
Discuss a project
Workshop, diagnostic, governance strategy, commissioned research, a talk at a professional gathering or a more specific question.
Advisory work here concerns AI governance, uses and decision processes. It is not legal advice and does not certify compliance.