Not a list of hype. A practical, opinionated breakdown of the AI tools that have actually changed how I work — and which ones I've tried and quietly stopped using.
Every month there's a new wave of "AI tools for PMs" articles. Most of them are sponsored or written by people who've used the tools for three days. This is a different kind of assessment: six months of real use, real product work, and real results — or lack of them.
The tools that stuck are the ones that changed the ratio of time-to-insight, not just the ones that sounded impressive. Let me share what actually made it into my regular workflow.
The Tools That Changed My Work
LLMs for synthesis (Claude, GPT-4)
I use language models for synthesis more than for generation. I paste in 50 customer interview transcripts and ask the model to identify the top 10 themes and the three most surprising insights. I paste in a competitor's product changelog from the last six months and ask what their strategic priorities appear to be. I paste in a draft document and ask what questions it leaves unanswered. In each case, I'm not asking AI to create the output — I'm asking it to help me process information faster so I can apply my judgment to a better-synthesized set of inputs.
AI-powered analytics interfaces
Natural language query interfaces on top of analytics data are genuinely transformative. I can ask "which features are correlated with 90-day retention?" or "which segments have the highest activation rates?" and get answers in seconds that would have taken an analyst request and a 24-hour wait. Not every platform has this, but the ones that do have changed how quickly I can move from question to answer.
AI meeting assistants (Otter.ai, Fireflies)
Automatic transcription and summary of user research interviews, stakeholder meetings, and customer calls. The specific value: I no longer have to take notes during conversations. I can be fully present in the discussion, knowing that the AI is capturing the full transcript and will produce a summary I can review afterward. The summaries aren't perfect, but they're good enough that reviewing and correcting them is faster than taking notes manually.
AI-assisted writing (for first drafts only)
I use AI to generate first drafts of user stories, acceptance criteria, and structured outlines for PRDs. Key emphasis on "first drafts" — I revise substantially before anything goes to the team. The value is that the first draft gives me something to react to, which is faster than starting from a blank page. It also surfaces edge cases and questions I might have missed.
The Tools I Stopped Using
AI roadmap generators
Multiple tools now offer to generate roadmaps from your goals or user feedback. I've tried three of them. They produce well-structured, generic roadmaps that look reasonable and contain no genuine strategy. A roadmap generated without the context of your specific business situation, team capabilities, competitive position, and organizational dynamics is a template, not a strategy. I haven't found one that's useful beyond giving junior PMs a starting structure to react to.
AI-generated personas
Generating personas from user data using AI produces statistically plausible personas that feel real but aren't grounded in actual user observation. Real personas are built from specific user quotes, specific behaviors observed in research, and specific tensions that only emerge from genuine interaction with users. AI-generated personas smooth over the specific details that make personas actually useful.
The Rule I've Developed
The tools that work are the ones where I'm using AI to handle information volume that would slow me down, so I can apply my judgment to a better synthesis. The tools that don't work are the ones where I'm using AI to make the judgment itself. Synthesis: yes. Analysis: yes. Research: yes. Strategy, prioritization, stakeholder management, user empathy: no.
The moment you're using AI to make the product judgment, you've outsourced the PM job. The moment you're using AI to make the information processing faster so your judgment lands on better inputs — that's leverage.
Figuring out how to use AI in your PM workflow?
The specific tools and workflows depend heavily on your product context and team setup. I'm happy to share what I've learned from six months of real experimentation. Let's compare notes.
Work with Yogesh →