In my current Stay with Tim O’Reilly interview, I spoke with Marily Nika, writer of Constructing AI-Powered Merchandise and some of the considerate voices on the intersection of AI and product administration. We talked about what it means to construct merchandise within the age of AI—and the way the function of product supervisor is being redefined in actual time. It is a topic that’s close to and expensive to me as I work with the O’Reilly crew to take the bitter lesson to coronary heart and rethink all of our processes and merchandise in mild of the brand new capabilities of AI. (For extra perspective, see additionally Drew Breunig’s critique of the bitter lesson as utilized to company AI technique.)
Marily began in AI product administration at Google again in 2013, earlier than most of us even referred to as it that. As we speak, she argues, that is not a distinct segment talent set. It’s turning into THE job. “All product managers will probably be AI product managers,” she mentioned. However she additionally warned towards what she referred to as the “shiny object lure”—utilizing AI simply to maintain up with the hype. Good PMs should keep grounded in person ache factors and product technique. AI must be used solely when it’s the very best answer. “Use instances haven’t modified,” Marily famous. “Individuals nonetheless need the identical issues. What’s modified is how we will clear up for them.”
Marily’s Speedy Prototyping Workflow
One of the thrilling elements of our dialog was listening to about Marily’s speedy prototyping workflow utilizing Perplexity for person analysis, customized GPTs for spec era in her personal voice, and v0 for UI mockups. With these instruments, she will go from thought to useful prototype in hours, not weeks. “Each week I block time on my calendar only for AI experimentation. It’s made me a a lot better PM,” she mentioned.
I hadn’t thought of limiting a search to Reddit to mine for person ache factors. That’s good.
One in every of our stay attendees requested a considerate query: Is there such a factor as “vibe PMing”? Right here’s Marily’s reply:
We additionally mentioned when to prioritize polish over pace—and whether or not AI will help us do each. “AI is a slider, not a change. It’s important to determine how a lot to make use of it at every stage,” she famous.
Marily additionally laid out three rising product supervisor archetypes:
- AI builder PMs, who work on the fashions themselves
- AI expertise PMs, who craft novel UX with these fashions
- AI-enhanced PMs, who use AI to amplify conventional product work
That’s actual meals for thought, and one thing that we’ll need to dig deeper into as we proceed to develop our O’Reilly stay coaching curriculum for AI-centered product administration.
Technique Meets Implementation
We talked a couple of theme near my coronary heart: the PM as translator between technique and implementation.
I’m very influenced by my spouse Jen Pahlka’s work on authorities transformation, as described in her e book Recoding America. In her telling, product administration is the talent of shaping not simply what to do when creating a product but in addition what not to do. Authorities is in some ways an excessive case, with mandates developed by nontechnical members of Congress and their workers, or by administrative companies, with little consideration given to the main points of how these mandates will probably be carried out, whether or not the implied implementation will work, or even when the specs are implementable! However these classes are additionally usually surprisingly related for these of us within the company world.
Two tales stick in my thoughts. The primary is a couple of PM on the Facilities for Medicare & Medicaid Providers who was confronted with a spec that she thought was unimplementable. Conflicting mandates from Congress meant that docs could be required to enroll in a program three months earlier than they’d obtain the knowledge they wanted to make that call. Altering the spec would have been subsequent to unattainable. So she made the daring choice to override it, reasoning that Congress had specified quarterly reporting as a result of they didn’t perceive that it could be attainable to create an API to supply real-time updates. The second is a couple of challenge chief who acknowledged that the challenge as specified wouldn’t work however mentioned, “In the event that they inform us to construct a concrete boat, we’ll construct a concrete boat.”
In her response to my prolonged tirade, Marily emphasised that whereas PMs don’t run day-to-day supply, they have to perceive the trade-offs between latency, value, UX, privateness, and feasibility—particularly in AI improvement. You don’t must construct a concrete boat simply because somebody advised you to.
Shared Instruments and Workforce AI Adoption
The most effective attendee questions of the hour—one which was so good that I’m utilizing it as a part of the framing of an extended put up I’m engaged on about AI for teams—was “What are some recommendations on coping with the truth that we’re presently working in groups, however in silos of particular person AI assistants?” (This query was from somebody recognized solely as DP. For some motive, a lot of our company prospects don’t need their workers to be recognized by identify or affiliation within the chat for our stay occasions, which is just too dangerous. DP, in the event you occur to learn this put up, please attain out. I’d love to speak with you extra about this concept. If you realize my identify, you realize my e mail.)
As you may see from the video excerpt, Marily fully agreed that this can be a drawback. AI use continues to be usually siloed and secretive in groups—folks afraid they’re “dishonest” through the use of it, she famous. She referred to as for groups to be open and collaborative about their AI workflows: create shared immediate libraries, use group instruments like NotebookLM, and normalize AI use with shared brokers and techniques.
It occurred to me based mostly on her response that NotebookLM might have a great begin as a platform for shared AI work by nondevelopers, as a result of it inherits lots of the collaboration options from Google Drive and the related household of Google productiveness apps. In the same manner, AI for builders depends on GitHub for many of its “groupware” capabilities.
However that highlights simply how LLMs themselves are actually weak on this space. Leaning on exterior infrastructure will not be an alternative to native options. For instance, how would possibly an LLM occasion have a bunch reminiscence, not simply person reminiscence? How would possibly it embody model management? How would possibly we share an AI workflow versus simply sending round hyperlinks to outputs, a lot as we used to ship round Phrase and Excel information earlier than 2005, when Google Docs taught us there was a greater manner.
The Rise of AI-Native PMs
In response to a different viewers query, we talked about Andreessen Horowitz’s declare that the world’s largest firm would possibly nicely be an AI healthtech firm. How would possibly somebody in healthcare get into AI product administration? Marily gave a robust reminder: You don’t should be an AI knowledgeable to get began. Now’s the time. It doesn’t matter what your job is as we speak, you may study, experiment, and construct with AI. Lean into your healthcare experience. She advised a narrative from one among her product administration stay programs on the O’Reilly platform that illustrated how one person had made the transition from a small {hardware} firm into an AI healthtech alternative at Apple.
We each agreed: We’re nonetheless early. Regardless of all of the hype in regards to the present market leaders, as we speak’s AI is barely scratching the floor. A few of as we speak’s dominant gamers might not survive. So many killer AI-native purposes haven’t been invented but. The way forward for AI continues to be up for grabs, and it’s as much as us to construct it.
Due to Marily for sharing her experience with us, and to the entire O’Reilly prospects whose questions are such an essential a part of our stay occasions, together with this one.
AI instruments are shortly shifting past chat UX to classy agent interactions. Our upcoming AI Codecon occasion, Coding for the Future Agentic World, will spotlight how builders are already utilizing brokers to construct modern and efficient AI-powered experiences. We hope you’ll be part of us on September 9 to discover the instruments, workflows, and architectures defining the subsequent period of programming. It’s free to attend.
