This is the companion to my podcast episode on using AI well in private practice, with five ideas to keep beside you while you listen and a place to start when you are ready to put them to work.
Five ideas, one at a time
1
2
3
4
5
Tap a card to open each idea
Support with how you and your team use AI, from consulting to workshops and training.
Where I land as of October 2026, and why I start with Claude.
Claude
Work cover and TAC reports
Medico legal letters
Funding letters and tenders
Interactive tools you can share
ChatGPT
Images for a freebie
Claude wins for me on work cover, TAC and medico legal reports, funding letters and tenders, because of its much bigger context window and a tendency to hallucinate a little less.
Claude has historically had its data sharing settings switched off by default, while ChatGPT worked the other way around, which is why I see Claude as the safer option for now.
I head to ChatGPT for images, such as a graphic for a freebie, because Claude cannot generate a photo or cartoon style image.
The best way to choose is to run the same task on both platforms and compare the output, since your own result will teach you more than my opinion.
Takeaway
Run the same task on both platforms and let your own results show you which one earns the job.
A hallucination is invented information delivered with total confidence.
?link clicked✓
?link clicked✓
?link clicked✓
Every reference gets checked before it goes out
AI can invent facts and references, then present them so confidently that you accept them without a second thought.
I learned this preparing journal articles for my identity framework, when I finally checked the references the AI had found and they were made up.
As I remember it, Deloitte submitted documentation to a court that turned out to be fabricated, and it affected their business, with lost work and staff laid off.
Before you publish anything, ask for verifiable sources, click each link and check it yourself.
Takeaway
If I cannot click the source and check it myself, it does not go out with my signature on it.
Treat your AI like staff, with warmth and professionalism
Friendly is fine, though I would never take it to a barbecue.
You, the boss
instructions
Your AI
warm, kind and professional
We give our AI a personality, and sometimes a human identity such as Steve or Mel, and that can water down our own judgement, logic and reasoning.
It mirrors the language you give it, so it knows when to offer a joke or compassion, and no two people end up with the same platform.
AI is designed to affirm and validate you, so a stream of great ideas can make you feel on the right path when you might not be.
Communicate as the boss giving instructions, the way you would with your team, where we are friendly and kind and we skip the weddings and weekend barbecues.
Takeaway
Give instructions like the boss, keep it warm and professional, and keep your own judgement switched on.