How I Use AI

AI prepares and I decide. It helps me read conversations, draft and edit, run admin tasks and build systems, and it gets only the access each task needs. It never sends, publishes or decides for me. I record calls only with your permission, I check every article's sources, and you can ask me to delete your information.

You are asked first. Fathom or Granola records and transcribes the call, and AI helps me read the transcript. I read the parts that matter and share what I found with you. I keep the recording and transcript until you ask me to delete them, and then I delete them everywhere I hold them.

1. I ask first. Calls are recorded only with your permission, and you know when a call is being recorded.
2. It is transcribed. Fathom or Granola records the call and makes the transcript.
3. AI helps me read it. Claude and NotebookLM build a summary, action items and quotes.
4. I read it. I quote you only from the transcript, never from an AI summary, and only with permission I captured at the time.
5. I share what I found. The people who give me their time get the findings back.
6. I keep it until you say. Recordings and transcripts stay until you ask me to delete them. When you ask, I delete them from every tool that holds them and tell you when it is done.

I keep recordings, transcripts and notes until you ask me to delete them. I do not run an automatic deletion timer, so I will not promise one. When you ask, I delete your information from every tool that holds it, and I tell you when it is done. I answer within 30 days.

• Recordings and transcripts. Kept until you ask me to delete them. They sit in Fathom or Granola, and copies of transcripts and notes may sit in Google Drive and NotebookLM.
• Notes and summaries. Kept until you ask. Notes with names removed may stay, so the patterns stay useful.
• Contact details and Scorecard answers. Kept while you are an active lead or client, and for a reasonable period after, to meet legal and business obligations. I delete them when you ask. The Privacy Policy has the detail.
• Email history. Kept in GoHighLevel. Unsubscribing stops the emails, and the record stays until you ask me to delete it.
• Testimonials. A testimonial stays published until the person asks me to remove it.
• Tool backups. After I delete something, a tool may keep its own backup for a short time. The tools I use clear theirs within about 30 days.

AI helps me read many conversations at once, draft and edit, make covers and diagrams, run admin tasks like scheduling and filing, and build systems. In each case it prepares the work and I review it. Nothing goes out under my name unread.

• Reading many conversations at once. AI shows me which words people repeat across dozens of calls. I read the calls that matter myself.
• Drafting and editing. AI drafts in my voice and I rewrite. The ideas, the stories and the results come from my work and my clients.
• Images. AI helps me edit my own photographs and make covers, diagrams and infographics. Where an image is AI-generated, I say so.
• Admin. Scheduled tasks help me schedule calls, sort my inbox, archive, organize files and move information between tools. They delete nothing, and they stop at "prepare."
• Research. AI watches what people ask search engines and AI assistants about my work, and helps me find sources.
• Building. AI helps me design and build the funnel, the pages and the automations I use. I test them before they go live.

AI never sends a message, publishes a page, writes a testimonial or invents a quote. It never decides for me, and it never touches photographs of my clients. It drafts, sorts and prepares. I review and polish everything before anyone sees it.

• It never sends anything. It drafts, and I review and polish every message. My email system sends the ones I have set to go out automatically, and I say upfront when a message is automated.
• It never publishes. Every article, post and page is published by me.
• It never writes a testimonial or invents a quote. Testimonials are clients' own recorded words, shared with their permission.
• It never touches photographs of my clients.
• It never gets your personal details from me unless I have checked where the tool stores them, whether I can delete them, and whether it trains on them.
• Automated emails are drafted with AI and polished by me. They say upfront that they are automated, and they never claim a personal memory I do not have.

Each AI tool and automation gets only the access its job needs. My inbox sorter can label and archive, and it cannot delete. My drafting tools can create drafts, and they cannot send. I decide what each one can see, and I look at those limits again when I review my automations.

• Scope. A tool sees the folders, accounts and tools its job needs, and nothing else.
• Actions. I give each one the smallest set of actions that does the job: label but not delete, draft but not send.
• Start narrow. I begin with less access and add more only when a task needs it.

For every article I open the sources myself and check that what I wrote says what they say. AI can sound sure and be wrong, so I never take its word for it. The more a piece matters, the more of it I read. Once a month I re-read one live automation from start to finish.

The more a result matters, the more I read. Summaries get spot-checks. Anything with my name on it gets read in full.

I learned this by catching myself. When my automations kept running well, I looked at them less. That is how oversight fades: the better a system performs, the less often anyone checks it. So I keep a rhythm, and some work I still do by hand, so my own judgement stays sharp.

Once a year I also check that my AI-assisted marketing still sounds like me and reaches the people I mean to reach.

Fathom and Granola record calls. Claude and NotebookLM help me read transcripts and notes. Google Drive holds notes and transcripts. GoHighLevel, Convex and Airtable hold contact details and Scorecard answers. I do not sell this information, and I do not put it into tools that are not on this list.

This list changes as my work changes. When it does, I update this page. The Privacy Policy covers contact details and Scorecard answers in full.

Yes, and I hold it to the same rules. Before I recommend an AI tool, I check where it stores information, whether you can delete it, and whether it trains on what you give it, and I tell you what I found. When I build for you, I ask what your clients' information is and where it goes.

• Where does the tool store information?
• Can you delete it?
• Does it train on what you give it?
• What does its agreement say, and how long does it keep your data?

I am building my own device with a private model on it, so my most sensitive material does not have to leave my desk. It is not in use yet. When it is, this page will say so.

You reach me, a person. AI can sound sure and be wrong, which is why I check its work. If something it prepared for me was wrong, tell me at [email protected]. I will correct it and tell you what changed.

You can ask what information I hold, ask me to correct it, or ask me to delete it. You can also ask me to leave your call out of my research, before or after the call. Email [email protected] with the subject "Data Request." I answer within 30 days.

I studied responsible AI use in the AI Leadership Accelerator at the London School of Economics, and these are the practices I wrote into my own work. If you coach or consult, you need a page like this too. These are the five questions I answer before I let AI near anything. You are welcome to copy them.

1. What does AI prepare, and what stays mine to decide? Mine: sending, publishing, quoting and anything that could not be undone.
2. Whose information goes in, and did they say yes? Write down when you asked.
3. Where does it go, and can I delete it? Check storage, deletion and whether the tool trains on what you give it.
4. Who checks the output, and how closely? Match the checking to the stakes.
5. When will I look again? Put a date on it, because a system that runs well is easy to stop checking.

If you hold an ICF credential, its Code of Ethics expects you to disclose AI tools that touch identifiable client data. Check the current text, and make your page and your coaching agreement say the same thing.

Answer them in plain language, publish the page, and review it twice a year.

• October 14, 2026: first version.