Predicting how AI will change the PM role
As a fellow PM (product manager) in the AI space, I’ve dived into a concrete understanding of how the PM role will be affected by the impact of AI.
GenAI and its various applications have taken life by storm over the past few years. We have corporations leveraging their power to multiple degrees; we have Tobi lütke of Shopify requesting his employees to use AI instead of hiring wherever possible; we have teams of all sizes building AI-based products or internal tools to increase valuations and keep competitiveness. It’s been non-stop action since GenAI models flooded the market.
Now that being said, the traditional PM (product manager) role in most product teams hasn’t changed drastically in terms of goals or expectations. PMs are still hired to lead product teams in achieving business success, no matter where they are. I’d wager that smaller organizations and teams set ever higher demands on PMs, who likely wear multiple hats and have a larger stake in the pie. So my first strong opinion to kick off this discussion:
PMs aren’t going anywhere. They’ll become more important than ever.
That being said, my prediction is that the PM role will be changing. And that’s not unique to just the PM position; the tech industry as a whole is seeing a whole recomposition of job responsibilities. Developers are using AI tools for coding, diagnostics, QA, and debugging. Managers are using AI agents for various workflows and decision-making processes. PMs will have to adapt to this rapidly changing landscape, so what are these shifts when it comes to one’s daily operations? And, subsequently, what does the future hold as we continue this AI adoption and subsequent transformation across product teams?
In this blog, I answer these questions with my own intuitive reasons and derived experiences. So here are my three bold and educated predictions.
Fewer PRDs, more productivity
PRDs (product requirement documents) have become less popular over the last few years, as they have traditionally been plagued by flaws that have remained constant. Not everyone reads the full extent of a PM’s documents due to time limitations (or perhaps some PMs are just not writing good drafts), and things become outdated or lost in time and space.
So many AI agents and LLMs are able to draft requirements and thus go deep into competitive analyses, strategy explanations, and user stories. Why spend hours writing them when we can rely on bots to do the tedious work with an integrated internet search as a reference? GPT-5 and Claude Opus or Sonnet 4 each have their respective strengths — Claude is known to produce powerful LLMs that code well (enabling products like Cursor) and perform scientific research, while Gemini 2.5 Flash is known for its text-to-image performances. It’s pretty safe to say that AI can not only write PRDs or one-pagers (or specs, as some others describe them), but it can also drive workflows, do internet search-based analyses, and perform other impressive activities. PMs can start focusing more on roadmap maintenance, building customer relationships, and user testing.
PMs can now shift away from spending hours writing PRDs and one-pagers every week to spending more time working with sales teams, talking to customers, and understanding their market more deeply. Or perhaps, if they’re in a technical team, they can spend more time with their engineers in product development. This is a significant shift in daily operations — and hopefully drives greater PM productivity.
AI will do lots of the research and drafting; PMs will just do more thinking and tinkering.
Deployment Strategists and Product Engineers are Becoming Popular
Some who observe the job market will notice the developing popularity of the Deployment Strategist and the Product Engineer title — especially in tech-savvy regions such as Silicon Valley and New York City.
These titles are derived from the traditional Product Manager role — they include many overlapping elements, but obviously with focused areas or activities tailored to the organization’s product strategy and business requirements. For example, Deployment Strategists at companies like Palantir play the product manager role but in a more customer-exposed environment that involves complex data modelling. They’re translating requirements, integrating customer data, and impacting the product interface in an agile way.
Likewise, Product Engineers are essentially AI engineers who can overlap roles with product managers. They’re not only contributing to the code base with the help of AI tooling (and building an AI tool themselves), but are also prioritizing work and driving cross-functional support. Their main expectation is to take customer feedback and requirements into their own hands and prototype solutions accordingly. These roles are typically more commonly found at startups nowadays that try to have a hybrid product manager who can contribute to the AI code base — effectively killing two birds with one person.
Another interesting term in Silicon Valley is the concept of a ‘forward-deployed angel’ — essentially an angel investor for an early-stage startup that also contributes to product development. We’ll get into that in a separate blog.
Use of multi-modal tools for creating PoCs
Multi-modal simply means the ability for an AI model to generate multiple forms of output, such as images, videos, and websites. It’s revolutionized the way we produce, lowering the barrier to entry for website-building, art creation, and prototyping. There are a multitude of tools available that allow anyone who can simply type a prompt to create outputs that seemed impossible just years ago. It’s empowered a whole society, driving everyone to the prompt-input screen to start generating content.
I personally think that every PM should learn how to use vibe-coding or other AI technologies like Loveable, Zapier, Cursor, and V0 to:
Prototype ideas or PoCs (Proof of Concepts) for engineers or designers faster and with higher fidelity. I’d try this with Loveable or V0.
Contribute to internal dashboards or internal tools that can indirectly support the product or the team building it. After generating an idea or a dashboard with Loveable, PMs could leverage Cursor - a VS Code Wrapper - to build internal data telemetry and ‘productize’ the tool to make it functional within an actual team. For example, I could vibe-code a dashboard that delivers weekly updates about our competitors to understand where we stand against them in various features or areas of business.
PMs have never been more empowered to build and prototype themselves in the age of AI. Whether it’s internal tooling or driving a PoC to explain product requirements or to simply build side projects that perform (and thus automate) certain tasks, activities, or workflows, every PM should learn how to leverage these tools to their advantage. It only goes up from there.
Now, that being said, outside of areas that make obvious sense, when it comes to blogs, newsletters, and journal entries, I’m still a proponent of writing without too much AI support. All my articles have always been 100% written by yours truly; never with AI. Why? We’ll get into the importance of preserving writing skills and the ability to dive deep into subjects without AI help in another article.
About Me
My name is Kasey Fu. I’m the co-founder of PM Hive and work full-time as a Sr. AI PM. I host the PM Hive Podcast, write for various publications, and keep a newsletter with over 3000 subscribers. Consider subscribing to the PM Hive newsletter today!