AI for product decisions
Don’t be a launch bad products faster kind of team.
Practical AI workflows for product teams. Make better use of customer evidence, challenge assumptions and figure out what’s actually worth building.
I build and teach practical AI workflows that bring that evidence together, uncover important patterns, check the results and help your team make decisions you can defend.
- Lenny’s Newsletter contributor
- 3,500+ hours testing AI workflows
- 2,500+ people directly trained
Start here · Claude Code · Approximately 3 hours
Data to Decisions
Turn customer evidence into a product decision you can explain.
A free, hands-on course for Product Managers and product teams.
Work through a realistic customer dataset in Claude Code. Find and verify patterns across survey responses, customer sessions and account information. Rank problems according to evidence strength and business consequences, then record what to act on, watch or park.
You’ll finish with checked findings, a written decision and a reusable Claude Code skill.
Built by me, hosted on agentsforproduct.com. See what’s included
What you’ll build
- Verified customer findings
- An evidence-based problem ranking
- An act, watch and park decision
- A record of the reasoning behind the decision
- A reusable discovery skill
Free course files · Approximately 3 hours · No coding knowledge or API key required. A paid Claude plan is required.
See how it works
Don’t just ask AI for a summary. Make it show its work.
Watch three AI agents examine three customer segments, find patterns across the evidence and check every quote against its original source.
See what happens when you give AI a proper analysis process rather than asking it to produce a report and hoping for the best.
4:36 from a live session. Sign up below to watch the full 38 min.
The work
Put AI to work on the product questions that matter
Understand what’s happening with customers
Bring together interviews, surveys, support conversations, sales feedback and product data. Find the underlying problems, not just the most frequently repeated comments.
Separate strong evidence from convincing-sounding output
Check whether claims are supported by actual customer evidence. Find contradictions, missing context, weak samples and conclusions that go further than the data allows.
Find opportunities worth investigating
Identify patterns across customer groups, uncover overlooked problems and investigate what the evidence suggests about unmet needs and commercial opportunities.
Build workflows you can use again
Move beyond individual AI conversations. Build instructions, skills, agents and repeatable systems that can work through customer evidence consistently.
Make the next decision clearer
Turn findings into a view your team can discuss: what to act on, what to investigate further, what to watch and what the evidence doesn’t support yet.
Trusted by teams building products people use
Go deeper
Build AI workflows your product team can rely on
AI Analysis for Customer Insights
A self-paced course for getting better results from AI when analyzing interviews, surveys and customer feedback. Build a structured analysis process, verify findings and create your own Analysis Copilot.
Claude Code for Customer Insights
An advanced live course on building file-based workflows, skills and agents for customer discovery and evidence analysis. Learn to structure the work, control AI access, test outputs and develop workflows your team can reuse.
The Insights Audit System
A ready-to-run system for checking AI-generated qualitative, quantitative, mixed-methods and multi-source analysis before it informs a decision. Use TRACE, Evidence Maps and workflow evaluations to assess whether a finding is actually supported.
For product teams
Get your whole team working with AI workflows they can trust
Your team doesn’t need another presentation about what AI might do.
They need to know how to use it on their own customer evidence, recognize where it goes wrong and build workflows that support the decisions they actually have to make.
Private cohorts, practical workshops and tailored training for PMs, designers and researchers.
What people say
I gained confidence on applying AI, as well as ideas for new areas where AI could be applied. I’ve struggled with some sort of mental bias or imposter syndrome when it comes to heavy AI use, and this course alleviated that.
Caitlin’s work was actionable and we could see the impact she had on the team right away. She has a thorough understanding of startup culture and corresponding challenges.
This course was incredibly insightful. Caitlin was super committed to everyone making the most of the content during the cohort sessions, and gave mindful, in depth feedback that made the learning experience feel very personalised.
We’re all seeing results from our LLM that are miles away from what we were getting before. I feel like it would have taken us years to figure out what you showed us in hours.
Where you may have seen my work
Lenny’s Newsletter contributor
How to do AI analysis you can actually trust
Lenny’s Newsletter
Video feature
This AI expert’s method will change how you do customer research
Aakash Gupta’s podcast
Workshops
Conference workshops are my jam. A few places I’ve run them:
- Hatch Conference
- EXL Austin
- PromptUX
- Circus London
and more
Research expertise. Product context. A lot of testing.
Hi, I’m Caitlin
I’ve spent 18 years working across marketing, design and research, including leading customer insights as Head of User Research at Soundtrack, Spotify’s B2B business.
I know what happens when customer evidence has to make its way into an actual product team: contradictory feedback, missing context, pressure to make a call and the difficulty of knowing when you have enough evidence.
I’ve spent more than 3,500 hours testing AI workflows to understand where they help, where they fail and what needs to be in place before a team can trust the output.
I don’t teach generic AI productivity. I teach systems for understanding customers and making better product decisions without treating an AI-generated answer as fact.
I’ve directly trained more than 2,500 people, and more than 20,000 people have attended my demos.