Tweet by ritakozlov

August 5, 2025

last week i asked @cloudflaredev PMs how they use AI day to day this was one of the best meetings i’ve been part of recently — there was so much creativity, way beyond the use cases you would think! here's what i learned: everyone assumes that AI for PMs is really in writing PRDs, but to my surprise, that was a use case not a single person had brought up in fact, towards the end of the meeting i asked: what’s something you tried using AI for, and found it wasn’t a good fit? writing PRDs was one of the responses don't offload your PRD (thinking) to the LLM the reason for that is that writing a PRD is a thinking process — it’s the process of defining the problem and trying to get at the root of what it is that you’re trying to solve for. you can’t outsource this part, and trying to shortcut it by asking AI to take a stab at writing a whole PRD only creates more work down the line while you shouldn’t have AI write the PRD for you, AI can be great at helping you write the PRD. here are few concrete examples in which PMs were using AI throughout the PRD writing process: - refining personas and what they need one way in which a PM mentioned they had used AI was to create very elaborate personas that they could then run the requirements through and see whether it satisfied them. if the personas have differing needs, you can then use AI to have them dialogue through their differences, and find solutions that address both (h/t @korinne_dev ) - defining states and edge cases; generating mermaid diagrams one of the hardest things to think through are all the states a user can end up in, and what should happen in each scenario. PMs used AI for: asking it to come up with all the edge cases that weren’t spelled out in the PRD that a user could hit thinking through solutions for the edge cases mapping out mermaid diagrams to help define user flows and journeys and visualize all the states one could end up in (h/t @yomnashousha) - competitive research a big part of PRDs is doing competitive research and understanding what state of the art looks like. LLM of choice + web search can expedite this process significantly, and provided the right prompt, can even give you pros / cons of different choices. prototyping this is the top response of what PMs are using AI for (!!) many PMs use tools like @lovabledev and @boltdotnew to prototype UIs — dropping in a screenshot of an existing UI and providing functionality in the prompt gets you a pretty long way. bringing prototypes to conversations with engineering makes it much easier to reason about behavior and expectations. (h/t @aninibread @yomnashousha ) some PMs have also started using Claude Code for similar prototypes with CLI (and even shipped a few minor features / changes!) (h/t @thomasgauvin ) building example apps with MCPs a big part of being a PM on the developer platform is writing many apps and examples to test out the developer experience AND provide better examples to users. previously, this could quickly become a major undertaking to get something up and running. today with AI you can ship examples relatively quickly to help refine the quality of the code, MCP servers (especially the cloudflare docs server) is great for making sure you’re using the latest information. another great hack was using gitmcp (go to claude, and just replace https://t.co/nF4BZgm9ZV with https://t.co/kF3e8EeQIP in the repo name) and add it as a tool. you can immediately use this as an MCP server that has direct access to a repo (h/t @dinasaur_404 ) summarizing looooong threads this was my favorite example of the day. one of the most stressful experiences as a PM is finding yourself in a 3 hour meeting block while engineering continues discussion in a chat thread. it can take hours to comb through the conversation, follow along, and understand the proposals. a very clever idea here was to use AI to not only summarize the conversation, but generate diagrams of each of the proposed solutions in the thread and map them to the name of the person who proposed them. so you can get a clear view of what’s on the table, what the differences between the proposals are, and jump into the conversation or make an informed decision. (h/t @nevikashah ) PS: come join our team this is an exciting time to be a PM, especially on dev tools! if the above sounds fun, we're hiring!

Author
ritakozlov
Date
August 5, 2025
Open on X