Anthropic recently released a playbook for startup founders — a practical guide on how startups have fundamentally changed in the AI era. After reading it, I realized it's not only for startup founders, but also to the day-to-day work of AI product managers (AI PMs). In many ways, AI PMs are already operating like early-stage startup founders.
Here are the three takeaways that resonated most with me and sharpened how I think about my own career.
"Same job, new rules." — The Founders Playbook
1. AI lowers the cost of building products, but it doesn't create real demand
Nowadays, with the help of AI, it’s possible to develop and launch products quickly. But that doesn’t mean that these products will actually be used. In other words, just because a product is well-designed, technically advanced, or extremely attractive, doesn’t mean it will be popular in the market.
As the Founders' Playbook notes: "42% of startups failed because they built something nobody wanted. AI will only increase this proportion, not reduce it."
Building a product faster does not make it more likely to succeed in the market. Value has to be proven by the market — it can't be assumed.
This point might sound counterintuitive, but it's actually practical. A product can be technically impressive and still fail to resonate with users.
Business-wise, this also reminds me of classic startup frameworks from The Four Steps to the Epiphany and The Lean Startup.
- Value Hypothesis - Problem-Solution Fit, MVP
- Growth Hypothesis - Product-Value Fit
- Demand Hypothesis - Product-Market Fit, PMF
- System Hypothesis - Product System Fit
These frameworks help explain why even a well-designed product can fail to find users. The product team probably skips the first two steps and jumps to the third one, PMF, instead. They never validate whether the product's solution actually fits users' needs. In some cases, even the problem itself may not be a real problem at all for the users.
I once worked on an AI video product that offered nearly every major model category, from text-to-video to image-to-video models. Everything was out there, but the product ROI was extremely low. At one point, I even doubted whether it had any paid subscribers.
After reviewing the product from multiple angles, including usage data, GMV, user interviews, and feature analysis, I concluded that there was almost no real user demand. The team had built a product with no viable market. Before I took over this project, the situation had persisted for more than a year. Based on my calculation, it couldn't cover the overall investment finacially.
Validating user demand is the first step, and it cannot be skipped. If a product can't pass it, then don't move to the second step. At that point, the team should consider stopping further investment and cutting its losses.
2. The founder role is becoming more about orchestrating agents
For decades, many founders came from technical backgrounds and could build products themselves. They often operated as hands-on individual contributors. AI is changing that. Founders are increasingly becoming orchestrators of agents rather than doing the work themselves.
As the playbook puts it, the founder becomes less of an individual contributor and more of an orchestrator of agents.
It’s important to determine what actions are worth taking, what should be done first, and which tasks can be automated by AI. On the other hand, some tasks should be handled by humans. A founder's judgment can either raise the team's ceiling or become its limiting factor.
Without judgment and business insight, a founder will inevitably limit both the team and the product.
For tasks AI can handle well, they can delegate to AI. For tasks outside of their scope, they should learn to let go and assign them to experts in the team.
The less ego a founder brings to the role, the more likely both the team and the product are to succeed.
The same applies to those of us working as AI PMs. When building products, it's important to clearly define our own scope, product scope, decision-making principles, and control the process appropriately.
3. The winners will not just build faster. They will validate harder, scope tighter, and systematize earlier.
AI increases the efficiency of our daily work, which allows us to build much faster. If we have clear direction, we can leverage AI to enhance the value of our products. However, if we move in the wrong direction, the product can fail faster than before.
The playbook's point is that speed alone is not enough; what matters is whether direction, scope, and process are established early.
If a team consumes substantial resources early in product development without getting concrete results from the market, it's too risky to keep investing, with little chance of meaningful returns.
If a product lacks clarity in both business boundaries and product scope, it weakens the product's core value proposition and makes it harder for users to form a clear mental model of what the product is for.
If a founder is the only person who makes the final decision, as a result, both the team and product can drift out of control. In such cases, when products have bugs, no one knows who's responsible for what or even how to diagnose the problems. When features need optimization, ownership is unclear. The team grows chaotic — everyone is exhausted, yet nothing gets done.
AI has lowered the barriers for startups, making it easier for people to enter the startup world. However, the bar for winning is higher than ever. With AI, it’s the best time for startup founders, entrepreneurs, OPCs (One Person Company), builders, and creators. Everyone starts from the same baseline — which makes this both the greatest opportunity and the most unforgiving time to build.
"The bottlenecks are no longer what you can build, but what you choose to build." — The Founders Playbook