I’m going to say something that might get me kicked out of the Silicon Valley founder’s club: most product roadmaps are garbage. Especially for AI products.
There, I said it. For years, I’ve seen founders, including my past self, slave over these beautiful, color-coded Gantt charts and detailed quarterly plans. We’d present them to our boards, our teams, and feel so proud. Then, two weeks later, reality would hit. A new model drops, a competitor pivots, or a core assumption proves dead wrong. And that beautiful roadmap? It becomes a beautiful lie.
I remember one of my early startups, a B2B SaaS tool. We spent a whole quarter building a complex predictive feature. It was the centerpiece of our roadmap. We launched, and… crickets. Our customers didn’t get it, and it didn’t solve a real problem for them. We’d wasted three months and a ton of cash because we were married to a document, not to solving the problem. That failure stung, but it taught me a lesson I’ll never forget.
AI product development isn’t a straight line. It’s a chaotic, exhilarating, and often unpredictable dance between what’s possible and what’s valuable. You can’t plan it a year in advance. You need a system that breathes, that adapts, that thrives on uncertainty. That’s why I threw out the old playbook and built my own system in Notion. It’s not about predicting the future; it’s about building a machine that learns and adapts to it, fast.
The Big Lie of AI Product Roadmaps
Most advice on AI roadmapping is a rehash of traditional software development practices with "AI" sprinkled on top. It’s fundamentally broken. They tell you to define your vision, break it down into epics, and then into user stories. That works fine when you’re building a CRUD app. But for an AI product, your core assumptions are constantly in flux.
Think about it. The state-of-the-art in AI changes weekly. A model that was impossible last month is open-source today. The biggest risk isn’t shipping a feature late; it’s shipping the wrong feature entirely. You need a system that embraces this chaos, not one that tries to tame it with rigid plans.
I’ve seen so many startups in my portfolio get stuck in this trap. They hire a team of expensive data scientists, spend months building a complex model, and then realize they’ve built something nobody wants. They were so focused on the "how" – the cool new AI tech – that they forgot about the "why". Why does this matter to the user? What problem does it actually solve?
This is where my Notion template comes in. It’s designed to force you to answer that question, over and over again, at every stage of the process.
My Counterintuitive Approach: The AI Product Engine
I don’t call it a roadmap. I call it my AI Product Engine. It’s a living system in Notion designed to do three things:
- Maximize Learning: Every single thing you build should be an experiment designed to teach you something.
- Embrace Speed: Move fast, break things, and learn from the wreckage. Small, rapid iterations are your best friend.
- Stay Grounded in User Pain: Never lose sight of the problem you’re solving for your users.
It’s not a pretty document to show your investors (though they’ll love the results). It’s a messy, dynamic workspace where the real work of building a great AI product happens. Let’s break down how it works.
Part 1: The Possibility Matrix
This is where ideas are born and brutally prioritized. It’s a simple database in Notion with a few key properties:
- Idea: A one-sentence description of the feature or experiment.
- User Pain: What specific user problem does this solve? Be brutally honest. If you can’t articulate this clearly, the idea is probably a dud.
- Impact (1-5): How much will this move the needle for our users? A 5 is a game-changer; a 1 is a nice-to-have.
- Feasibility (1-5): How hard is this to build? A 5 is a weekend project; a 1 is a multi-month research project.
I learned the importance of this the hard way at MovieLaLa. We had this grand vision for a recommendation engine that could predict what movie you’d want to watch next with uncanny accuracy. We spent six months and a huge chunk of our seed round building it. The tech was incredible, but the user impact was a 2, maybe a 3. People just didn’t care that much. We could have built a much simpler version in a few weeks, learned the same lesson, and saved ourselves a ton of time and money.
Now, I force myself and the founders I invest in to be ruthless here. We only work on ideas that are a 4 or 5 on both Impact and Feasibility. Everything else goes into the backlog. This simple filter saves you from chasing shiny objects and keeps you focused on what matters.
Part 2: The Experiment Tracker
Once an idea makes it through the Possibility Matrix, it becomes an experiment. We don’t build features; we run experiments. This is a crucial mindset shift.
Each experiment gets its own page in Notion, linked to the idea in the Possibility Matrix. The page has a simple template:
- Hypothesis: What do we believe will happen if we build this? (e.g., "We believe that adding a natural language search bar will increase user engagement by 20%.")
- Success Metrics: How will we know if we’re right? Be specific. (e.g., "20% increase in DAUs performing a search action.")
- MVP Spec: What is the absolute minimum we can build to test this hypothesis? No gold-plating allowed.
- Results: What actually happened? Did we validate or invalidate our hypothesis?
- Learnings: What did we learn from this experiment?
This is where the magic happens. By tracking everything, you build a knowledge base of what works and what doesn’t for your specific product and users. This is your real IP, not your code or your models.
At RemoteTeam, which was acquired by Gusto, we used this system to build our entire product. We ran dozens of small experiments every month. Some were wild successes, others were spectacular failures. But every single one taught us something valuable. We were able to out-maneuver bigger, better-funded competitors because we could learn and adapt faster than they could.
Part 3: The Feedback Loop
Your users have all the answers. You just need to listen. My Notion template has a built-in feedback loop to make sure you’re constantly capturing and integrating user feedback.
It’s another simple database, linked to your users and your experiments. Every piece of feedback – from a support ticket, a sales call, a tweet – gets logged here. We tag it with the relevant feature or experiment and a sentiment (positive, negative, neutral).
This creates a direct line from your users to your product decisions. When you’re debating an idea in the Possibility Matrix, you can filter this database and see everything your users have ever said about that topic. It’s like having a focus group on demand, 24/7.
I can’t tell you how many times a single piece of user feedback has saved us from making a huge mistake. I remember one user who sent us a long email about how confusing our onboarding was. We were about to spend a month rebuilding it based on our own assumptions. His email showed us we were completely wrong. We spent a weekend making the simple changes he suggested, and our activation rate doubled. Doubled! All because we listened.
Why Notion? Because Your Brain is for Ideas, Not for Storage
I’ve tried everything. Jira, Asana, Trello, even a physical whiteboard. Nothing comes close to Notion for managing this kind of dynamic, learning-focused system. The flexibility of databases, the ease of creating templates, the way you can link everything together – it’s the perfect tool for the job. It’s like an external brain for your entire product team.
It’s not about the tool itself, of course. It’s about the mindset. But the right tool can make it a hell of a lot easier to put that mindset into practice.
Stop Planning, Start Learning
If you’re a founder building an AI product, I challenge you to do one thing: throw out your roadmap. I’m serious. Burn it. Instead, start building your own AI Product Engine. Focus on learning, speed, and user pain. It’s a messier, more chaotic way to build a product. But it’s also the only way to win.
I’m not selling anything here. I’m just sharing what’s worked for me, from my own startups to the 200+ companies I’ve invested in. To make it even easier, I’ve cleaned up my personal Notion template and made it public. You can duplicate it and start using it today.
Stop building beautiful lies. Start building a machine that learns. Your future self will thank you.
[Link to the Notion Template]
Frequently Asked Questions
What if I disagree with some of the advice?
Good. That means you're thinking critically, which is exactly what a good founder should do. Take what resonates, test it, and discard what doesn't work for your specific situation. No advice is universal.
How often is this guide updated?
I revisit and update my guides regularly as I learn new things and as the market evolves. The core principles tend to stay stable, but specific tactics and tools get refreshed based on what's working right now.
How should I work through this guide?
Don't try to absorb everything in one sitting. Read through once to get the big picture, then go back and work through each section as it becomes relevant to your current challenges. Bookmark it and return to it regularly.
Who is this guide designed for?
This guide is written for founders and operators who want practical, actionable advice rather than theoretical frameworks. Whether you're just starting out or scaling an existing business, the principles here apply across stages.