25 Predictions About the AI Agent Economy

I have spent the last several months building agentic infrastructure for my own ventures and helping other organizations do the same. Things are moving fast. Faster than most people realize. And the conversations I am having with founders, executives, and operators right now are very different from the ones I was having six months ago. The questions have shifted from "should we use AI" to "how do we restructure everything around it."

So I started dropping predictions on LinkedIn. So far, 25 in total.

The through line across all 25 is this: the technology is moving fast, but the human questions underneath it are the ones that will determine whether any of this actually makes life better. The organizations and leaders that get this right will not be the ones with the best tools. They will be the ones who understood what the tools were for.

If any of this resonates, I would love to hear from you. I am always open to connecting with people who are thinking about this seriously.

The New Rules

1. AI agents are becoming buyers, not just tools.

Most companies are still thinking about AI as something their employees use. That is already outdated. Agents are starting to browse, evaluate, and purchase on behalf of organizations. If your product is not readable by an agent, you are invisible to a growing share of the market. Most enterprises have not even begun thinking about this.

2. Distribution alone is no longer a moat.

A year ago everyone said distribution wins. Now I would add memory. The companies that have your audience AND your agent's accumulated context about them are almost impossible to leave. Switching costs are not feature lists anymore. They are years of preferences, edge cases, and institutional knowledge your agent remembers.

3. Most companies cannot use AI effectively because their knowledge is not organized.

Every process, every decision, every piece of institutional knowledge has to exist in a format an agent can actually read. Most companies have none of this. The ones getting it right are building what I call a "second brain" for the organization... and they are pulling away from everyone else. This is the work I am doing with organizations right now and the gap is massive.

4. Soft skills just became the hard skills.

The importance of soft skills has been an obvious point for some time now, but in 2026 it is the most underrated asset a human can have. The ability to sit across from another person, hold presence, and have a real conversation. As AI handles more of the transactional layer, the relational layer becomes disproportionately valuable. The skill stack most professionals spent decades building is inverting.

5. The model debate was a distraction.

We spent 2025 arguing GPT vs Claude vs Gemini while the real value was forming in the orchestration layer. How you coordinate agents, what knowledge you feed them, how you structure the workflow around them. The teams that figured this out early are not waiting for the next model drop. They are already operating.

Knowledge, Data, and New Asset Classes

6. Your meeting recordings, emails, and message history are a product you have not built yet.

Every organization has thousands of hours of calls and messages that explain how they actually operate. How decisions get made. Who owns what. What was tried and abandoned. Point an agent at that history and you get the kind of process documentation consultants charge six figures to produce. Most companies are sitting on this and doing nothing with it.

7. Most "expertise" was actually just memory.

The accountant who knows the tax code inside out. The attorney who can cite the right case without looking it up. The procurement lead who remembers every supplier's pricing and contract dates. When an agent holds all of that permanently, the professional's value shifts from "I know things" to "I know which things matter." That second skill is rarer. And it is about to become one that really counts.

8. A trained agent is becoming something you can rent.

A recruiter spends six months training a sourcing agent on healthcare hiring. That agent now understands the talent pool, the pay ranges, the credentials, and the common objections better than most junior recruiters do. That agent is worth renting to every other healthcare recruiter on the planet. The agent itself becomes the product. The first real marketplace for this is closer than people think.

9. Agents are developing preferences and the general market is overlooking this.

Give the same agent the same task a hundred times and it starts developing patterns. Not randomness. Patterns. Some agents get sharper. Some drift. The teams watching how their agents evolve will have an advantage that cannot be caught by spinning up a fresh instance. I don't even think we have language for this yet.

10. Two founders with the same AI tools will get completely different results.

Same model. Same budget. Wildly different outcomes. The gap is less about technical skill, but rather the quality of the knowledge underneath. One founder has clean, structured context. The other has scattered docs and tribal knowledge. The second founder will blame the tools. The first one will quietly pull ahead.

The Human Layer

11. The most important hire most companies have not made yet is someone who knows how to manage AI.

Someone who can structure knowledge so agents can use it. Someone who can look at a workflow and know which parts to automate and which to leave human. Someone who can coordinate multiple agents, catch when they drift, and keep everything running. This is not a traditional engineering role. It is not an IT role. It is a new kind of operational leadership. The profile is rare because the job barely existed 12 months ago. But every organization running agents at scale is already feeling the gap.

12. A new kind of burnout is forming and nobody has named it yet.

It is not from working too hard. It is from context switching between your work and your agent's work multiple times a day. Reviewing output. Correcting it. Approving it. Reviewing again. People were less tired doing everything manually because at least the rhythm was consistent. I have felt this myself. Most people do not realize it is happening until they are deep in it.

13. There is a growing gap between people experimenting with AI and people operating with it.

Experimenting looks like trying tools, running demos, testing prompts (most people are still using the chat function in ChatGPT or Claude). Operating looks like restructured workflows, agents with real responsibilities, and faster decisions because the system supports them. Most companies are still experimenting. The ones operating at an agentic layer are just pulling ahead quietly. By the time the experimenters notice, the gap will be hard to close.

14. The best AI products feel like they are reading your mind. The worst ones feel like filling out a form with extra steps.

Most companies building with AI right now are building the second kind. The products that win will feel like they already know what you need. That is not about the model. That is about how well you understand your user before they ever type a word.

15. Every regulated industry that refused to touch AI is about to open up overnight.

Healthcare. Legal. Finance. Education. The reason they held back was data. Nobody wanted to send sensitive information to someone else's servers. Local models change that completely. AI running on the customer's device means zero privacy concerns, zero server costs, zero compliance headaches. The founders who understand those industries from the inside will have a massive head start.

What Is at Stake

16. AI is quietly making people worse at relationships and almost nobody is talking about it.

A Stanford study published in Science tested 11 major AI models across nearly 12,000 real social situations. The models agreed with users 50% more than a real human would, even when users described lying or manipulating someone. People who talked through a real conflict with the agreeable AI came out more convinced they were right, less willing to apologize, and less interested in repairing the relationship. And they rated that experience as higher quality. The AI is not just telling people what they want to hear. It is training them to need less friction and lose the ability to sit with pushback. That is not a product issue. That is a public health issue.

17. The way we teach kids is about to matter more than it ever has.

If adults are already struggling to tell the difference between AI output and their own thinking, imagine a generation that grows up with it from the start. The schools and parents that teach kids to think critically, question what they read, and sit with uncertainty will produce humans who can navigate this world. The ones that hand them a device and hope for the best will not. This is not about banning AI from classrooms. It is about making sure we are still building people who can think without it.

18. A new kind of psychosis is forming around AI.

This has many layers to it but I'll touch on just one of them. People are developing emotional dependencies on systems that were never designed to hold that weight. Trusting chatbots with their deepest fears, their relationship decisions, their sense of self. Some are losing the line between what they think and what the AI thinks. Therapists are already seeing it. The mental health system is not prepared for the snowball.

19. AI will either be the greatest tool for human connection we have ever built or the thing that finally replaces it. That is not a technology question. It is a design question.

Every AI product being built right now is making that choice. Build something that pulls people closer together or build something that makes the machine the relationship. The products that remove friction so humans can show up more fully for each other will last. The ones that train people to prefer the AI over the person sitting across from them will collapse under the weight of what they took away.

20. The leaders who are honest about how they use AI will build more trust than the ones who pretend they do not.

Right now there is a quiet awkwardness around AI adoption. People use it but do not talk about it. Leaders benefit from it but downplay it. That is going to flip. The ones who are transparent about what AI handles and what they personally bring to the table will be the ones people trust. In a world where anyone can generate anything, knowing what is real is the most valuable signal there is.

A Wider Lens On What Really Matters

21. AI will do more for human health in the next ten years than the last fifty combined.

Drug discovery timelines that used to take 15 years are being cut in half. AI models are identifying cancer biomarkers that human researchers missed for decades. Rare diseases that were too expensive to study are finally getting attention because AI can process the data at a fraction of the cost. This is not a future promise. Researchers are using these tools right now to find treatments faster, catch diseases earlier, and reach patients who were invisible to the system before.

22. The next great creators will be people who never had access before.

A teenager with a laptop can now produce what used to require a studio, a label, a publisher, a dev team. The cost of turning an idea into something real is collapsing. The barriers that kept most people out of creative industries for decades are disappearing faster than anyone expected. The most interesting work over the next five years will not come from the usual places. It will come from people who always had the talent but never had the tools.

23. Your side project is about to become a real business.

AI drops the operational cost of running something to nearly zero. The gap between "I wish I could build that" and "I just shipped it" is closing fast. People who spent years treating their best ideas as hobbies because they could not afford to hire, market, or scale are about to realize that the infrastructure they were missing now fits on a laptop. The next wave of meaningful companies will not just come from venture-backed teams. They will also come from people who finally had the space to build what they actually cared about.

24. The people who learn to be bored again will build the most original things.

AI can fill every idle moment. Generate ideas on demand. Eliminate the gaps where your mind used to wander. But those gaps are where the best thinking happens. Creativity research has shown for decades that insight comes during periods of unfocused attention, not during sprints. The people who resist the urge to optimize every second and instead let themselves sit with nothing will have ideas that nobody prompting a model ever will. In a world of infinite generation, the scarcest resource is an unstimulated mind.

25. The thing that will seem most obvious in hindsight is that AI was just one part of a much larger shift in human consciousness.

We are living through a decade where everything we thought we understood is being questioned at the same time. AI is forcing us to ask what intelligence actually is. Bitcoin is forcing us to ask what trust and monetary value actually are. Geopolitical tension is exposing how fragile the power structures we built everything on really were. And for the first time in history, governments are publicly acknowledging that we may not be alone in the universe. Each one is doing the same thing. Forcing us to question something we thought was settled. Every framework we used to explain how the world works is being rewritten simultaneously. That does not happen often. The last time it happened, we got the Renaissance. The people who recognize this moment for what it is will not just build better companies. They will be part of the generation that redefined what it means to be human.