Some thoughts on new AI tools, the product process, and building the wrong things faster.
The message is constant these days: AI for product management opens up so many new possibilities. We can move faster, track things better, get to production quicker, simplify communication, and remove a lot of the friction that slows teams down. PRDs and prototypes whipped together in seconds. Holy smokes.
But if we’re being honest- there’s a lot of open questions to how we’re handling these major changes too. I’ve been using AI tools for a while now as a tool in my product tool kit and I’ve been thinking about how much the typical product process has changed. Last week, I had the opportunity to attend ProductCon in New York, where the nuances, challenges, and benefits of AI as a tool in the product toolkit were the major focus. It was encouraging to hear some of these concerns addressed, and like many in the product space who have been heavily using AI tools… I have some thoughts.*
AI has helped us move forward more efficiently and has made us faster, but it can also help us build the wrong things faster. And that’s the part I keep thinking about.
Things go undocumented. Teams think they’re aligned but they’re not. Everyone blankly nods in meetings (aka “corporate bobbleheading.”) Then you realize the new experience that just launched isn't what anyone thought it would be and oh, by the way… Customer complaints are on fire.
Historically, engineering capacity, design bandwidth, and documentation effort acted as constraints. Annoying constraints? Yes. But often useful because they inevitably forced prioritization. Now, these constraints disappear. You can prototype in hours, generate code, and move things to production way faster.
Awesome. Except now it’s also REALLY EASY to build things nobody actually needs. You whip together that button from a customer request but now you could have an unwanted and unneeded product experience for potentially thousands of other users. Shiny designs plus ready to ship code plus speed do not automatically equal value. If you don’t really know what problem you’re solving, AI just helps you get to failure faster.
I’ve found AI incredibly useful in my day-to-day work from cleaning up PRDs and planning discovery to developing KPIs and building interactive prototypes- but specifically after thoughtful product discovery and research. Unsurprisingly, in Lenny Rachitsky's AI and Productivity Survey prototyping ranked as the second most common use of AI among product managers, behind writing PRDs. After ProductCON this year, it’s clear that so many AI tools and companies are on fire to address this market need and rapidly expanding into AI prototyping.
For me, AI as a product tool has helped with moving faster and communicating ideas more clearly, while removing some of the friction that used to slow down turning ideas into something tangible. But as products come to life faster than ever, there are a few things I've been thinking about:
We need more access to faster research tools and usability testing:
With the ability to move faster to production, usability testing and user interviews will become more important than ever. Synthesizing survey results or setting up quick usability tests can be done much faster and shouldn’t be skipped. I’m keeping my eye on usability and testing tools that support essential product discovery, like Dialogue AI. AI tools that can help with recruiting participants, surveying, prototype usability testing, and collective insights could help us make sure we’re validating before shipping- instead of relying on assumption. As a product manager, it’s a DREAM to have these parts of the discovery process become standard. Faster to production shouldn’t eliminate discovery, testing, and validation- and we can use new AI tools to support this.
We need to pay attention to Product Bloat:
With an increase of product pushed to code faster, there’s an even bigger concern for multiple features leading to the same outcome or creating an overwhelming experience for users. Legacy products have always struggled with this, and now new products will face similar challenges. Excessive product bloat can lead to more tech debt, large maintenance needs, and confusing user experiences. Product feature tracking, maintaining consistent language, and thinking about the larger holistic experience across the platform is going to be more of an essential than ever before. Good documentation will be critical for supporting decisions on what to build. The goal shouldn’t be more features. It should always be better user experiences.
Garbage in = Garbage out:
Moving faster to building products can unfortunately mean loosening product and requirement documentation expectations. But confusing or under-composed prompts can lead to outputs that are incomplete and messy, like PRDs with oddly written descriptions, or prototypes with confusing user flows. If you feed AI tools weak inputs, bad assumptions, or the wrong problem to solve, then congratulations! You now have a very polished piece of garbage. Strong prompts built from product research and a solid discovery process can reduce the garbage-output risk. It’s a reminder that successful outcomes will always start with good product thinking, and it doesn’t hurt to share your prompts with your team to make sure you’re thinking the same thing.
The Beige-ification of Design:
With prototypes and design tools available to everyone, designs can become incredibly homogenous. We’re going to see eerily similar looking designs across different companies and their products. Everything starts to look imperfectly perfect and exactly the same, kind of like watching a historical drama and everyone has perfect teeth. (“Wait… Did they have expert dentistry during the French revolution?”) AI tools very much lack the soul of user-centric design and that’s ok- it’s a reminder that humans so badly need to be involved. While it’s exciting to find ways to bring ideas to life for complex problems, we increasingly will see a need for tools that support more customization and a need for people behind the tools who can bring emotion and connection to their designs.
I’ve been digging more into prototyping tools, like Miro’s new Protyping suite and Claude Design and checking out their capabilities that support design customizations. I’ve really enjoyed using Claude Design to help communicate the product vision, and it’s been a helpful tool for talking through ideas. Lovable and UXPilot have been other interesting tools to try out, but I’m going to guess that over time, the AI prototyping tools that prioritize stronger design customization, connected design systems, and an ability for true team collaboration are going to become more popular. Additionally, many of these AI prototyping tools can connect to your codebase and in the right context, support products seamlessly moving from idea to live. I’m especially interested in tools that not just connect prototyping directly into a codebase, but also match to a design system, like Alloy AI, and how this will transform how product and engineering work together in the future.
I’m also imagining a world where if it’s easier to generate a quick table or button, there can be a bigger focus on the greater overall user experience and the holistic design. With stronger prototyping tools, we could possibly see a shift to a focus on building stronger design systems. I’d love to see more AI prototyping lead to a serendipitous shift to more considerate user flows, stronger design guidelines, improved usability, and smarter, more intuitive experiences that make a user smile or feel more understood. When there’s more of a user centric design focus, this makes all the difference for the success of a product and hey- maybe we’ll see a few more fun Easter eggs scattered in the UI too. Products that stand out in this evolving AI product world will now be the ones that find opportunities to be the hot neon pink in a sea of beige.
Feature sunsetting and when to say goodbye:
Analysis paralysis often comes up when discussing feature sunsetting, and there's a common organizational bias toward keeping features alive long after they've stopped providing value. I once worked on a team that received a passionate email from a user upset about a feature planned to be removed. What made the decision easier? That user was the only person who had used the feature in the previous three years and hadn't used it in the last month.
As AI speeds up the path from idea to production, feature cleanup becomes just as important as feature creation. Not every feature deserves to live forever. Part of any sunsetting evaluation should ask two simple questions: Is this still creating value? And is the maintenance cost worth it? If the answer is no, teams need to be empowered with the confidence to simplify, consolidate with other features, or push it off the cliff. (Insert cartoon splat sound here)
With sunsetting processes, data tools for things like Usage tracking, product analytics, and feature adoption metrics can become more critical. Tools like Fullstory, Hotjar and Datadog’s Product analytics (and a development infrastructure to support these tools) can help teams make more informed decisions about what deserves continued investment and what doesn't.
At the end of the day, building faster is only half the equation. We also need to get better at letting go too.
AI doesn’t replace good (human) product thinking
As teams are already looking at AI-influenced PRDs and documentation and thinking, Do we still need product managers? Call me crazy but product management was never supposed to be about just writing Jira tickets. With all the new capabilities and access AI has afforded us in product, the thing I keep coming back to is this: the fundamentals still matter.
You still need strong use cases and customer understanding.
You still need value propositions and a clear “Why.”
You still need Jobs to Be Done.
You still need to understand the core problem. AI doesn’t remove that work. If anything, it makes it more important.
The point of product management is not just documentation.
It’s the vision.
It’s understanding customer problems. Connecting strategy to execution. Making tradeoffs. Keeping teams focused on building the right thing.
Product managers serve as both user and business advocates, and should be able to communicate ideas in a dozen different ways: research, prototypes, user flows, presentations, customer insights, technical discussions, even jumping into code when it makes sense and in the right context. A common theme from ProductCon this past year was that product leaders are seeing all roles across Product and Engineering becoming more collaborative and versatile. Roles are shifting and people are taking on tasks that might not have been a part of their toolkit before.
That doesn’t make PMs or Designers or Engineers less valuable.
It changes where the value is.
If AI systems are connected to strong product documentation, research repositories, standards, and codebases, then yeah, things can be faster… But if user-centric design systems exist, if product thinking exists, if usability testing and user research exists, if user advocacy exists, then AI truly can become a powerful and impactful accelerator.
If none of that exists?
We just move fast in the wrong direction. Without the product discovery process, AI is just a faster train without a station.
To be honest, I hesitated to write this article for many reasons. Will these thoughts one day become as cringey like the Today show about the internet? There are also some deeply troubling and problematic issues with AI at a larger level. (See below*) We don’t know what the future holds just yet for how product processes evolve, and the tools we’re talking about today might be replaced tomorrow. However, I have hope that in a year from now, the core concepts about product-thinking, discovery, and testing ideas will still be valuable.
And I think that’s what product management becomes in our ever-evolving AI world: not the people writing the work, but the people making sure the work is worth doing in the first place.
**** This article specifically focused on AI in Product only. My thoughts here didn’t dive into some of the larger general concerns of AI and I would be tap dancing around an elephant in the room if I didn’t include this:
I want AI to be regulated and I believe it can be.
I want stronger data privacy laws that include personal image and facial recognition protections
I want LLMs and all AI driven tools to obtain required consent and compensate for their sources.
AI has no business in the creative arts, including but not limited to writing, visual, theatric, and music.
I am disgusted by the environmental impact of AI data centers, specifically on clean water, and this is a major issue for the future of AI usage
AI must have stronger guardrails to prevent unethical and illegal outcome, specifically with image creation. Regulations and Laws need to adapt now.
Layoffs in the name of AI are short-sighted and self-destructive.