Iro AIWhat you'll be able to do
- Cluster a pile of raw user feedback into named, counted themes
- Draft a PRD or spec from your notes in minutes, not hours
- Run a competitive teardown that surfaces real gaps, not marketing copy
- Turn a messy analysis into a stakeholder update people actually read
- Catch when AI is inventing a theme or a competitor feature
Inside the path
A focused set of five-minute lessons. Each one ends with a hands-on exercise, not a quiz you can guess.
Feedback to themes 6 min
The prompt pattern that clusters raw reviews, tickets, and interviews into counted, named themes.
Draft the PRD 6 min
Turn your notes and constraints into a first-draft PRD you edit: problem, goals, scope, risks.
Tear down a competitor 5 min
Get past marketing copy to real feature gaps, positioning, and where you can win.
Prioritize with a framework 5 min
Use AI to score options against RICE or your own criteria, then make the call yourself.
The stakeholder update 5 min
Turn a week of noise into a crisp update: decisions, progress, risks, and asks.
Try a sample exercise
This is the kind of card you'd practice inside Iro: you do the thinking, then get feedback.
◆ Sample exercise · Prompt practice
You've just closed a round of 30 user interviews plus a stack of support tickets and app-store reviews. You want AI to help you find the themes so you know what to build next, but "summarize this feedback" gives you one vague paragraph you can't prioritize from.
Your task: Pick the prompt that turns a pile of raw feedback into something a PM can act on.
- "Summarize this user feedback."
- "Act as a product analyst. Below are 30 pieces of raw user feedback. Group them into distinct themes, and for each theme give a one-line description, how many pieces mention it, one example quote, and whether it's a bug, a friction point, or a feature request. Flag any theme where the signal is thin or you're inferring intent."
- "Based on this feedback, what should we build next?"
- "Read this feedback and write the PRD for our next feature."
See why the second prompt wins
The winning prompt sets a role (product analyst), gives the model the raw material (30 real pieces of feedback), and asks for a structured output that a PM can prioritize from: themes with a frequency count, an example quote, and a type. Crucially it asks the model to flag thin signal and inferred intent, so a single loud review doesn't masquerade as a trend. The other options either hand the model your decision ("what should we build next?") or skip straight past validation to a PRD. In Iro you'd write your own version and get feedback on grouping, counts, and where you let the model guess.
Why AI fits the PM job so well
Most of a PM's day is synthesis: reading feedback, weighing data, reconciling stakeholder opinions, and turning all of it into a decision and a document. That's exactly the work AI accelerates: it can read a hundred reviews faster than you, cluster them, and draft the doc. What it can't do is decide what matters, and it will happily invent a theme or a competitor feature if you let it.
So the model does the reading and the first draft; you do the prioritization and the verification. A good PM prompt gives the model the raw material (the actual feedback, your notes, the constraints), asks for a specific structure, and tells it to flag anything it's inferring, so a thin signal never quietly becomes a roadmap item.
Where AI helps a PM — and where it shouldn't
- Synthesis: cluster feedback into themes, summarize interviews, pull patterns out of survey data.
- First drafts: PRDs, specs, user stories, release notes, stakeholder updates.
- Research: competitive teardowns, market context, sharper questions for your next round of interviews.
- Pressure-testing: red-team a spec, list edge cases, poke holes in your own prioritization.
What stays yours: the roadmap, the trade-offs, and the final read on what a theme actually means for the product. AI drafts and clusters; you decide.