AI for Executive Coach

Tip: Describe the client by role, industry, and challenge type rather than by name. That's enough detail for the AI to generate something useful, and it means nothing identifying ever leaves your hands. Treat the output as a starting list, not a script. The best question in the room is usually the one you ask in response to what they actually say, not the one you prepared the night before.

Tip: Free chatbots occasionally invent a plausible-sounding title, misattribute an author, or describe a real book inaccurately. Before you pass any recommendation to a client, check that the title and author actually exist and that the one-line description matches what the book is really about.

Tip: Fill in your actual cancellation window and communication rules rather than leaving the AI's placeholder numbers in place. This one's fully generic (no client details go into the prompt), so it's a good one to build once and reuse for every new client rather than rewriting it each time.

Tip: Pull themes from the pattern across several engagements, not from one specific client's situation this week. If a post reads like it could only be about one identifiable person, generalize it further before it goes live. That's a marketing-content prompt, so there's no session data involved here, just the shape of what leaders tend to struggle with.

Tip: Ask for the specific conflict framework you already teach if you have one (Thomas-Kilmann, or your own shorthand) rather than accepting whatever labels the AI defaults to. A guide that matches language you've used before lands better than one introducing new terms mid-session.

Tip: Describe the challenge the way the prospect framed it in the call, but leave out their company name and any detail that would let someone else identify them if this draft ever sat in a shared inbox or a tool's chat history. Nothing here is under a signed coaching agreement yet, but the same habit of describing the situation rather than the person will serve you well once it is.

Tip: Change enough (industry, company size, exact title, timeline) that neither the client nor anyone at their organization could identify them, and get the client's explicit sign-off before this goes anywhere public, even in de-identified form. A composite that feels 90% true to what happened is usually still specific enough to recognize from the inside, so err toward changing more than feels necessary.

Tip: Feed the AI the development themes only, never raw assessment scores or verbatim 360 comments, and never anything that names the client's employer. Before the plan goes anywhere near the client, check it against the actual assessment report yourself. A generated plan can sound coherent while quietly drifting from what the data actually showed, and you're the one accountable for that gap, not the chatbot.