Responsible AI
We help businesses use AI, and we use AI tools in our own work every day. That makes it especially important that we are clear about how we use it, and where its limits are.
This approach is informed by the Artificial Intelligence Playbook for the UK Government (2025), the Information Commissioner's Office guidance on AI and data protection, the Algorithmic Transparency Recording Standard (ATRS), the Data and AI Ethics Framework and the Guidelines for AI procurement. Following these principles does not mean we are endorsed by any of these bodies.
Our principles
1. People stay in control
A person reviews AI output before it is relied on, and the client or a named person remains responsible for decisions. We design workflows so that AI supports people rather than making significant decisions about individuals on its own, and we make it easy to step in, correct or switch off an AI tool. We will not use AI to make a solely automated decision that has a legal or similarly significant effect on someone, and we will not present unchecked AI output as fact.
2. Client data is never used to train AI models
We do not use client data to train or fine-tune AI models, and we choose tools and settings that stop providers using it for training wherever that option exists. We put only the minimum data needed into an AI tool, and we do not put personal or confidential client data into a tool until that tool, its plan and its settings have been confirmed and agreed with the client. We never put that data into a consumer or free-tier tool that uses prompts for training. For demos and testing we prefer synthetic, anonymised or public examples.
3. Client data is returned or deleted
The client stays in control of their data. At the end of an engagement we return or delete client information as the contract requires, keeping only what we must keep for tax, insurance or legal reasons, stored securely and then deleted.
4. We are open about where AI is used
We tell clients when AI has been used in our work, and which tools we use. For any AI tool we set up that interacts with the public or helps make decisions, we will document what it does, the data it uses, its known limits and who is accountable, using the ATRS template so that public sector clients can publish a record if they need to.
5. Lawful, fair and secure
When we advise on a client's own AI tools, the client normally remains the controller of their data unless the contract says otherwise. We help clients consider a data protection impact assessment (DPIA) before any AI use that is likely to be high risk, and we apply our information security measures to every AI tool we use. We help clients understand their own obligations under UK GDPR and the Equality Act 2010.
6. Tested for accuracy and bias
AI can be wrong, and it can reflect biases in its training data. Before a client relies on an AI workflow, we test it on realistic examples, check for errors and unfair outcomes, and agree how it will be monitored. We tell clients plainly what the tool cannot do.
7. The right tool for the job
AI is not always the answer. We recommend AI only where it genuinely helps, prefer the simplest tool that works, and say so when a non-AI solution is better.
8. Skills and honesty
We keep our knowledge up to date, train client teams to use AI safely, and never make claims about AI that we cannot back up.
How we use AI in our own business
We're an AI-native consultancy: we use AI responsibly in our own work every day, with a human accountable for every decision. We use AI tools for research, drafting, admin and preparing client work. Lee Taylor reviews and is accountable for everything we send to clients. Our information security page explains how client data is protected in the tools we use.
Questions or concerns
If you have a question or concern about how we use AI, email [email protected].