The Human API: Your Only Defense Against AI's Corporate Takeover

I had a meeting last evening with my ex-chief-the man who gave me a chance in product management-and an ex-colleague. We were discussing AI in our lives as PMs, and the consensus was that AI is finally eliminating the stupid parts of our work.

One of the guys mentioned he no longer dreads pointless meetings. Before, you had to sit there like a monkey, listening to corporate staff talk because that’s how they perceive work. Now, he’s a bigger fan of these meetings because he can skip them. He lets the Microsoft Teams AI log the call, and then he just reads the transcript, the summary, and the to-dos. Let’s be honest, a huge chunk of corporate meetings are a waste of time with no real outcome. AI helps save that time.

This leads to a funny side effect. Soon, half the “participants” in a meeting will be AI notetakers like Otter.ai or the built-in Teams function, recording for people who couldn’t join. You will essentially be making your presentation for AI transcription agents, whose job is to deliver something more concise, coherent, and down-to-earth to your boss. Your task is now to speak in a more understandable way for the machine. The AI overlords joining your meetings will rely on your ability to explain things clearly; otherwise, the summary will be just as stupid as the meeting itself.

Your ability to communicate in clear terms will define the quality of the summaries that go to the people who matter. In a way, AI is already controlling our language. If your input is shitty, the output will be shitty, just like when a human leaves a meeting with nothing in their head-just a weird feeling of having wasted two hours with no clear next steps.

Soon, you might be running a presentation with only a few people-or none at all-just a bunch of AI notetakers. You’ll be presenting audio to the LLMs, which will then generate an actionable newsletter for the stakeholders. You probably won’t even need to prepare the visual part. Business stakeholders who value their time will focus on those AI summaries. So, don’t mess it up. Have a checklist, make your statements short and concise, and state the outcomes clearly at the beginning and end of the meeting. You have to adjust your language for the LLMs. Name the to-dos explicitly so the model can pick them up. Otherwise, people won’t watch the videos, and that meeting will be another lost hour of your life. Unless, of course, that’s all you do at your workplace. In that case, God help you, because you won’t be there in 5-10 years.

That’s one point. The second one is that the price of building things is dropping like crazy-for now. VCs are sponsoring OpenAI, Anthropic, and others, but soon those tools will get more expensive. Vendor lock-ins will appear, and the cost of operations for LLM tools will rise significantly. Companies will start cutting corners, using these tools only for crucial tasks. This means smaller headcounts. If you’re useless, you won’t even be there to use the LLM tools that would make your work easier.

I see many PMs, especially those from marketing, who have no idea about the unit economics of their company. They lack fundamentals. Their work is feature-delivery-based, and they track nonsense like story points delivered per sprint while revenue stays flat. Now, these same people can build more prototypes and deliver more features, digging themselves into a giant hole of open projects without understanding the business reality. The software space will be filled with lost effort.

My ex-chief made a great point: the most important role will be the “human protein API.” There will be fewer product managers, but if you are that person who can connect the dots-from the code to the balance sheet-then you’re okay. You have to understand how these tools work, increase your time to market, and learn on a daily basis. You can’t just stick to the Scrum guide anymore. The technical PM is no longer a joke; it’s a necessity. There is no such thing as a non-technical PM anymore. You’re either on that train, or you’re not there at all.

Designers are already moving ahead, building prototypes with tools like Galileo and Vercode AI. They are merging design with front-end work. Meanwhile, many PMs are getting stuck in the middle, using AI for PRD creation and ticket summaries, basically becoming glorified secretaries. The “human API” is needed because LLMs lack a value system; they don’t have a connection to the fundamentals of your specific company context. But you can’t just be a glorified secretary who manages Jira tickets-that will be 1% of your job in the future.

This conversation changed my perspective. I’ve been a doomer about how AI is being used by the masses to generate stupid images for Instagram. But there are real applications that will increase our speed and change how we work. However, you need management that understands this-not just as a way to cut headcount, but as a way to empower the right people.

When the VC funds disappear and these tools become expensive, companies will start counting their money. The useless jobs in corporate branding and the like will be gone. Only people with experience, with something in their brains, will be trusted with expensive LLM seats. It’s an exciting time, but you have to understand how much time you now need to invest daily just to make yourself less replaceable. Keep learning, keep building prototypes, because if you don’t act today, you will be gone in 5-10 years.

• This post is a corrected version of an audio transcript; it is full of mistakes, punctuation fails and grammar outrageousness, I don’t care.