A builder lays track where no line exists. A manager runs the timetable. Now Agentic AI pulls up the tracks and jumbles the timetables in every company at once.
Nothing yet to manage
When you start with a fresh idea there is nothing yet to manage. That is the whole difference between a builder and a manager. That puts companies with many managers but few or no builders at a disadvantage.
In 2001 we started sketching local.ch. Google Maps would not exist for another four years. There was no model of how a digital local directory should work, what people would search for, or what they would expect to see.
In late 2007 we started Memonic. No online notebook was publicly available yet, Evernote’s first public version came out in June 2008, and nobody knew how people would actually use one. It was unchartered territory.
So what next? The Agentic AI change will be profound. Don’t believe it? Simply follow the money. Every value chain will eventually be agenticly reconfigured because you can produce the same output in less time, at better quality for less cost. (Btw: I did not say with less people – but more on that in another post). Agentic work will change the core operating pattern of any company.
Track and timetable
How to think about this? I think of it in railway terms, which will surprise nobody who knows me. A builder lays track where no line exists. A manager runs the timetable once it does. Switzerland is proud of its timetable for good reason. The train leaves at 7.32 and it leaves at 7.32 every day, and nobody wants a builder improvising the 7.32 to Bern.
But a timetable is worth nothing on a line that has not been built. James March described the same split in 1991 as exploration and exploitation, and showed that organizations drift toward exploitation on their own, because its returns are nearer and surer. Builders explore. Managers exploit. Neither is better. They sit on opposite sides of a table.
Crossing the line
Agentic work resets work for every company at once. When agents take over the steps a company’s processes are made of, the current human driven operating model loses its template. The value chain gets rearranged, and an established company faces the same empty map local.ch faced in 2001. You first have to find and define your new human-agent driven operating model.
Now the problem is that in most companies the people in charge were selected for exploitation side of the table. And that is a problem: As Benson, Li and Shue have shown promotion picks past output, not the skill of the next job. Westphal and Stern surveyed 760 directors and found that flattery and favours toward peers were the strongest predictor of being appointed to more boards.
No flaw. It is what the selection system asks for, and it produces excellent timetable keepers. An agentic world asks for something else. Insight into how the work actually gets done. Nuance about where an agent can be trusted and where it cannot. And the willingness to take a risk without a precedent to point to.
You can see the mismatch in the numbers. MIT’s NANDA project found that 95 percent of organizations saw no measurable return from generative AI. Of those that evaluated enterprise systems, 20 percent reached a pilot and 5 percent reached production, and large enterprises ran the most pilots and converted the fewest. The pilot, the steering committee and the process for the process are timetable tools. They are excellent when the track exists. They are slow and off when it does not.
The order matters. The track comes first.
Combine that with that fact that building got cheap, and that changes who is scarce. When everyone can build, the judgment what to build gets scarcer. It comes from having built many times, from knowing what an empty map feels like, not from having allocated headcount to manage.
So look at who leads your agentic program. Ask one question. When did this person last build something where no model existed. If the answer is never, the fix is not another steering committee. It is a builder next to them, with the authority to lay track before anyone writes the timetable.
Every company’s tracks are being pulled up at the same time. The best timetable in the world will not tell you where to lay the new ones.
