AI Policy
Last updated: 11 September 2026
Why this page exists
I help businesses use AI. It would be strange not to be clear about how I use it myself, and what I won't do with it. Most AI consultants don't publish this. I think that's the wrong way round — if you're going to trust me with how your business handles its customers, you should be able to see where I draw my own lines.
Where I use AI in my own work
- Drafting and thinking. I use AI tools to draft, research, structure and pressure-test my own writing and materials. Everything published under my name has been read, edited and approved by me.
- Research. For market and competitor work, checked against original sources rather than taken on trust.
- Building systems. The tools and automations I build for clients often use AI as a component — for drafting replies, sorting enquiries, answering routine questions.
Which tools I use
I ask clients to expect a straight answer about which tools their data touches, so here are mine.
- Claude (Anthropic) — drafting, research, analysis, and the reasoning behind the Business Leak Check
- ChatGPT (OpenAI) — occasional drafting and comparison
- Lovable — building this site and the tools on it
- Supabase — storing what the forms collect
- MailerLite and Resend — email
- Microsoft 365 — email and documents
- GoHighLevel — the platform I use to build client automation systems
Client work may use different tools where the job needs them. You'll be told which before anything is built, and it goes in writing.
What always has a person in it
- Anything I send you, or publish, or put my name to.
- Every recommendation about your business. A diagnostic can narrow the question; it doesn't make the decision, and neither does a model.
- Any judgement about whether something should be automated at all.
What I won't do
I won't put your customers' personal information into a general-purpose AI tool without agreeing it with you first, in writing, and only where the tool's terms allow it.
I won't build anything that makes an automated decision about a person — who gets a service, what they're charged, whether they're a good customer — without a person reviewing it.
I won't build an AI that pretends to be human. Any chat or voice agent I set up says clearly that it's automated, and can hand over to a real person when someone asks or when it's out of its depth.
I won't write fake reviews, testimonials, case studies or results, with AI or otherwise. If I quote a client, they said it and they've approved it.
I won't automate something that's working. Adding AI to a process that already runs reliably adds cost and fragility for no gain, and I'll tell you when that's the situation even though it means selling you less.
Your data in systems I build
If I build something that uses AI on your business's data, I'll tell you before we start: which tool it uses, what data goes into it, where that data is held, whether it could be used to train the model, and how to turn it off. If I can't answer those questions about a tool, I won't use it on your business.
You own your data and your systems. If we stop working together, you keep both, and you get an export.
When AI shouldn't be used at all
Some things shouldn't be automated regardless of what's technically possible — moments where someone needs a person, decisions with real consequences for someone's wellbeing, and anything where being handled by a machine would itself be the problem. Working out where that line sits in your business is part of the job, not an afterthought.
Before this, I worked in HR and then in private practice as a therapist. Confidentiality, consent and knowing when something needs a person aren't new questions to me — they were the job.
Questions
claire@builtforher.ai — ask me anything about this, including about a specific tool you're considering.