Workshops · AI for Insights
AI for Insights: Foundation.
Sixty minutes on using AI for research. Free, live, and ten people in the room so you get to ask me things.
60 minutesOnline10 seatsFreeNext session: 10 September
Sixty minutes on using AI for research. Free, live, and ten people in the room so you get to ask me things.
One session a month. Ten seats each time.
You typed a question about your customers into ChatGPT. What came back was 400 words of beige. Confident, well-punctuated, and useful to nobody. So you wrote a longer prompt and got a longer version of the same beige.
A lot of people stop there and file AI under "not ready yet." I understand why, but I think it's the wrong read. Almost every failure I've run into traces back to one of 3 things: the brief never told the model what job it was doing, the model was never set up to know anything about you, or the answer came back with no evidence trail and got believed anyway.
This hour deals with all 3.
I'm not going to sell you a tool. I don't own one.
AN HOUR OF WORK, NOT AN HOUR OF SLIDES.
Seven things, then your questions. No breakout rooms, no icebreakers.
What got easier. What didn't. And where the promise has run ahead of what the tools can currently do.
Listen, Ask, Analyse, Synthesise. Knowing which one you're in tells you what to ask for, and what to be suspicious of.
Enough of the mechanics to explain the odd behaviour, then the setup that makes a model useful to you in particular: custom instructions, project context, memory.
Across the general models, the qual analysis platforms, the AI interviewing tools, and the listening and scraping end. Including the ones I've stopped using. Full list goes out afterwards.
The 6 things every good brief gets right. We take a real business question and build the brief live, so you watch the weak version fail and the good one work.
Why models invent things, the failure modes that catch researchers in particular (fabricated quotes, invented sample sizes, confident misreadings of your own data), and the check you run every time.
Which tools train on what you paste, what should never go near them, and where the GDPR line sits for client and respondent data.
Gather what people already say. Reviews, forums, transcripts, the open web, at a scale no person could read.
Put questions to people, and through AI. Open-ends, interviews, follow-ups that probe instead of accepting the first answer.
Turn raw material into structure. Themes, sentiment, patterns you can stand behind and trace back to source.
Pull it together into a point of view someone can act on. The part that was always the real job.
A quick view in the session. You keep the one-pager.
The angle you're looking through. Whose view, what frame, what you're really trying to see.
What you give the model to work from, and what counts as good enough to trust.
The job to be done, stated so plainly the model can't wander off it.
The exact shape you want back. Format, length, structure.
The caveats and edge cases. The "be careful here" that separates a pro brief from a lazy one.
Where the answer must come from and where it must not, so you can trace every claim.
Insights people, marketers, founders, and small teams who need to understand their customers properly and want AI doing some of the lifting. If you can write an email, you can do this.
If you already run RAG pipelines and write your own evals, this hour will bore you. Go to Data Harvesting instead.
25 years in research and brand consulting, now rebuilding all of it around AI. Interviews, focus groups, ethnography, the analysis and the boardroom conversation afterwards. I'm still doing that work every week, which is the only reason I have anything useful to say about where AI helps with it and where it doesn't.
More about meFoundation gives you the map. These two put it to work on something you keep.
Turn the open web into evidence. Build a repeatable architecture that gathers reviews, forums, and competitor data into clean datasets, then analyse them for content, theme, and sentiment.
View workshopResearch you can talk to. Build a knowledge base that holds everything you've ever learned about your customers and answers questions from it. Your team's second brain.
View workshopTen is deliberate. I want to hear what you're stuck on and fix it live, and that stops working somewhere around a dozen people.
Booking asks 3 questions: your role, what you're working on, and the one thing you most want AI to do for you. I read every answer before the session and build the hour around what comes back. It's why 10 people get more out of this than 100 would.
Want it for your team on its own? A private session built around your category, your customers, and your questions.
Yes. No card, no upsell call afterwards. I run paid workshops and a research consultancy, and a few people who come to this end up working with me. Most don't, and that's fine.
No. If you can write an email, you can do this.
None, to attend. Bring any of the main ones if you want to follow along: ChatGPT, Claude, or Gemini. A paid plan helps but isn't essential.
Yes, and they go. One session a month, ten seats each time. If it's full I'll hold you a place in the next one.
Yes, and you'll get the link. The questions are where the value is though, so come live if you can.
Tell me and I'll move you. There are 10 seats and I'd rather someone else had yours than watch it sit empty.
Yes, and knowing exactly how to keep it that way is one of the 7 things we cover.