pm is a crazy job.
as a new grad, you're handed a laptop, and are the person responsible for the direction, strategy and roadmap of a product. you're accountable for both being directionally correct, and shipping fast. on day 1, you're thrown into standup where you now tell engineers, often 10 years your senior, what they should be working on next.
except you don't actually have the authority to tell them what to do. you have to persuade them that it's the right thing to do.
*power through influence, not through authority. *
this is true not just of engineering. you have to convince marketing, sales, customers, your leadership — stakeholders in every direction that your thing is the right thing to build, and advocate for what you need in order to be successful.
there's a tricky needle to thread: you have to compromise and be flexible to build relationships — but without compromising on doing the right things for the user. you have to push for what you think is right, and have the confidence to do so, while having knowledge that is limited relative to your engineering peers and counterparts.
(btw, i lead a team of PMs for developer products where the line between product and implementation is thinner than ever).
you have to keep the team accountable to shipping and deadlines, while you're not the one holding the pager if a launch is too early or causes incidents.
so yeah, pm is a crazy job. it's extremely terrifying and impostor-syndrome inducing. and there's never been a better time to do it, and here's why (yes, you guessed it: AI!).
using AI for deeply understanding
it's always been true that the best PMs are the ones that can go deep — on the problem and on the technology. there's a balance here of course in not getting so sucked into the implementation that you lose track of the experience, but still, understanding how the systems that support your product work is important. it helps you understand trade offs and gives you leverage to push back when you don't agree with a timeline, a compromise, or a technical requirement.
one of the biggest unlocks i've seen for my team since starting to use AI is in using it to gain deeper understanding of all the context PMs need access to.
AI is a great teacher. it doesn't judge. it allows you to inspect, and ask questions (no such thing as dumb ones), until you can really wrap your head around a problem (or a solution for that matter).
it has access to your codebase, so you can ask all of your questions without having to spend several hours of an engineer's time.
it has access to all of your customer notes, so you can ask it to pull all conversations relating to X, or find themes across them. it can do deep competitive research — not just reading about other products, but going through and setting them up. something that would previously take hours' worth of work.
at a time when software is abundant, building the right thing becomes even more critical. and AI can be a really helpful tool in being a partner here: it has a larger context window than most humans, and serves as a great thinking partner. this is contrary to many of the accusations of AI, that it absolves people of thinking. if anything, it sharpens their thinking.
AI for prototyping
if a picture is worth a thousand words, a prototype is worth, well, at least millions.
prototyping is more powerful than a PRD, and accelerates the development process. it allows you to come to engineering with: here's what i want to build, and how i want it to work. it also gives you something to walk through with customers early on to understand what they're looking for, and validate your direction.
using opencode + kumo-ui at cloudflare for prototyping has been a massive unlock for PMs, has accelerated the feedback loop of going from PRD to UI, and improved the overall look and feel of our dashboard.
this is useful not just for UI for for mocking CLI and API behavior as well.
AI for all the QoL backlog your team never gets around to
beyond prototyping, rather than just being able to report bugs, and minor annoyances, PMs are able to submit actual PRs.
i realize the topic of who should be contributing code, is still something that is debated. it's a conversation we're having internally, and for larger changes that are more than just about code, but about approach and architecture, you have to exercise a bit more caution.
but so many bugs and quality of life improvements are no brainers. people used to submit bug tickets that would sit on the backlog for years and never get picked up — just never made the cut for being high priority. or teams would schedule days of the week, or sprints to address them.
now instead of submitting a bug, you can actually just fix it.
AI for marketing
as the PM, no one understands the customer problems, and thus the narrative of your product better than you do. you're the one that's constantly talking to customers, and getting to refine your narrative in real-time based on reactions — seeing what's sticks and what doesn't.
PMM then became the translation layer that would take the refined narrative and turn it into artifacts — slides, landing pages, one-pagers.
the problem with this model is that there's always a bit of the message that becomes lost in translation, and it's often one step behind.
with AI, once you have the narrative down, turning it into assets is a much more efficient process. our PMs use opencode to generate slides, contribute directly to the marketing website, etc. AI is very efficient at switching between these different mediums. as someone with no background in consulting, i never mastered the art of slides. but asking opencode to build a react app for slides and feeding it the narrative produces beautiful results every time.
there's never been a better time to be a PM
the above is not even a comprehensive list of ways that PMs at cloudflare use AI. it makes life easier in so many different directions.
the hardest part about being a PM is that the work is never done. you're always thinking about your product, and there's always something more you could do to make it successful. there have always been and always will be more of those things than hours in the day.
AI hasn't changed the number of hours in a day, but it can really stretch what you can make of them. and of course, make no mistakes.
PS: if you want to be a PM on an ai-cracked team, we're hiring!
Link: http://x.com/i/article/2020232857034321920
