AI Governance Policy

A significant part of our work involves advising charity leaders on responsible AI adoption. Publishing our own AI governance policy is not a formality - it is a demonstration of the standard we hold ourselves to and expect of the organisations we work with.

We believe that any organisation using AI tools in client-facing work has an obligation to be transparent about what those tools are, how they are used, and what safeguards are in place. This document is our attempt to meet that obligation.

We currently use the following AI tools in the course of our work:

Claude (Anthropic)
Claude is our primary AI assistant. We use it to support research, drafting, analysis, strategic thinking, and workflow development. Anthropic operates under a published usage policy and privacy notice. We use Claude via claude.ai, which means conversations may be used to improve Anthropic's models unless privacy settings are configured otherwise. We do not input personally identifiable client data into Claude.

Perplexity AI
We use Perplexity for research tasks, particularly for finding and synthesising current information, regulatory developments, and sector-specific data. Perplexity's responses are grounded in cited sources, which we verify independently before relying on them in client work.

Fathom.ai
We use Fathom to generate transcripts and summaries of video calls, including client discovery sessions and strategy conversations. Fathom operates with participant awareness - we notify all call participants that Fathom is active before a session begins, and we do not use Fathom on calls where a client has asked that recording not take place. Transcripts are stored within Fathom's secure environment and are not shared externally.

Research and background analysis
We use AI tools to support research into sector trends, regulatory developments, and best practice. All AI-generated research is verified against primary sources before being used to inform client recommendations.

Drafting and document production
We use AI to assist with drafting reports, frameworks, and communications. All outputs are reviewed, edited, and approved by a human before being shared with clients. We do not use AI to generate content that is presented as our own thinking without applying our professional judgement to it.

Strategic analysis
We use AI to help structure and stress-test strategic thinking, explore options, and identify gaps in analysis. The conclusions we reach and the recommendations we make are our own, informed by our professional experience and knowledge of the client's context.

Meeting summaries
We use Fathom to produce summaries of client calls. These summaries are used to support our own note-taking and to ensure accuracy in follow-up documentation. They are not shared with third parties.

We do not input personally identifiable information about clients, their service users, their staff, or their donors into AI tools without explicit agreement and appropriate safeguards in place.

We do not present AI-generated content as professional advice without applying human review and professional judgement.

We do not use AI tools to make consequential decisions about clients or their organisations without transparent human oversight of those decisions.

We do not use AI-generated content in grant applications or funding submissions on behalf of clients without disclosing that AI has been used and ensuring the content authentically represents the client's voice and evidence.

Our use of AI tools intersects with our data protection obligations. Our approach:

We treat AI tools as data processors where they handle personal data, and we assess their privacy terms and data handling practices before use.

We configure AI tools to minimise data retention where such settings are available.

We do not use general-purpose public AI tools to process special category data or sensitive personal information.

Where a client engagement involves AI tools processing data on the client's behalf, this is disclosed and agreed in the engagement terms.

Transparency. We are open about when and how we use AI, both with clients and publicly through this policy.

Human accountability. A human reviews all AI outputs before they are used to inform advice, recommendations, or client communications. AI supports our work - it does not replace our professional judgement.

Proportionality. We use AI tools for tasks where they genuinely add value. We do not use AI for its own sake, and we are honest when a task is better done without it.

Client primacy. Where a client has concerns about AI use in their engagement, we take those concerns seriously and adapt our approach accordingly.

Continuous review. The AI landscape is changing rapidly. We review our tool usage and this policy at least every six months to ensure our practices remain appropriate.

This policy applies to all work undertaken by Kingsbury Consulting Ltd, including work delivered by contractors or virtual assistants acting on our behalf. Contractors are expected to adhere to these principles and to disclose any AI tool use that falls outside them.

This policy was last reviewed in April 2026. Given the pace of change in AI development and regulation, we review it every six months or whenever we adopt a new AI tool or materially change how we use an existing one.

Questions about our AI practices can be directed to michael@kingsbury.consulting.