Industry & People

The Org Chart Had Two Jobs. AI Agents Are About to Take One.

How AI Changes The Organisation

Every org chart is a fossil of a communication problem.

I should know. I’ve spent more than twenty five years living inside one, and I’ve built a good career on the problem it was designed to solve.

That problem is now being solved without me. If you sit anywhere in the middle of a chart, it’s being solved without you too.

The problem is simple. One person can only direct a handful of others before things fall apart. The Roman army knew it. So did the railroads. Their answer was layers: small groups, a leader for each, a leader of those leaders, and so on up.

Layers gave us scale. They also gave millions of us a job. Somebody has to carry what the top knows down, and what the front line knows up.

Hold that thought, because machines are learning to do it.

From Chatbot to Swarm

If you don’t follow this stuff daily (a healthy choice), here’s how fast the ground has moved.

First came the chatbot. You asked, it answered, and you still did the work.

Then came assistants that finish the job. Tools like Claude Cowork and ChatGPT Work take a brief, open your files and apps, and come back with the finished thing.

This autumn the assistants moved in. Meta’s Muse and OpenAI’s Dots are always-on personal agents. They watch your inbox and calendar and act before you ask.

The newest step is agents working together. Put thousands of them on one problem, let them message each other, and you have a swarm. (Think beehive, with a deadline.)

One swarm has already made the news. OpenAI says it pointed roughly ten thousand agents at one of the famous unsolved problems in mathematics. They worked for 88 hours and traded about 2.7 million messages. The people involved set the goal and, by the company’s account, changed direction once.

No steering committee. No pre-read. Nobody asked to take it “offline”.

(Mathematicians are still checking the proof, and that matters. I care less here about whether the answer holds than about how the work got organised.)

Ethan Mollick, the Wharton professor whose AI writing I never skip, drew the lesson that stuck with me in a recent post. Much of what we call management exists to solve human problems. We guard information. We protect turf. We want the promotion. Agents, so far, do none of that with each other.

So what happens to the fossil when the communication problem goes away?

The Router Years

I went looking for the answer in the least comfortable place available. My own calendar.

This summer I took on a bigger role. I’ve spent the months since getting an operating model stood up and working. Who decides what. Who tells whom. Which meeting feeds which.

I have, in other words, been lovingly drawing a chart. (The irony arrived late, but it arrived.)

Then there’s a normal week. I take what one team knows and carry it to another. I translate one function’s language into another’s. I chase, align, summarise and “loop in”. The days run to twelve hours and still don’t hold it all.

I’m good at it. Most people who reach mid-career are. It is, to be blunt, a big part of how we got here. If that list sounded like your week, you’re in good company.

I’ve started calling these The Router Years. A router is useful. It’s also the first thing you stop noticing when the network gets better.

That was a slightly cold moment. If a swarm can pass 2.7 million messages without one status meeting, a fair share of my week is a workaround for a limit machines don’t have.

The Stoics had a habit I like here. Look straight at the thing, without the story you’d prefer. So I kept looking.

The Second Job Nobody Wrote Down

The chart did something else, and it never appeared in anyone’s job description. It grew people. And it made somebody answerable.

The evidence is more interesting than the headlines. Researchers at Columbia mapped the layers of more than 3,100 US public companies. Firms did flatten after adopting AI. Yet the firms with taller hierarchies tended to perform better operationally, promote more from within and keep their people longer.

Korn Ferry found the same thing from the other side. Four in ten professionals say their company removed a layer of management in the past year. More than a third say they were left without clear direction.

So the layers were doing two jobs at once.

Job One was the relay. Moving information from where it sits to where it’s needed.

Job Two was the rest. Teaching someone the things no system holds. Deciding what’s worth doing. Putting your name to the outcome and meaning it.

Job One is about to get very cheap. Job Two is the part a swarm of AI agents, for now, hands back to us.

A Normal Workday, a Couple of Years Out

Picture a normal workday. This is a guess, held loosely.

You start work and there’s no inbox to clear. Your agent cleared it overnight. It read the threads, answered the routine ones in your voice, and left three items marked “only you”.

There is no status meeting, because status is simply known. The swarm keeps the picture current for everyone, all the time. (Someone will still try to book one. Old habits.)

Your first conversation is a decision. Two options, prepared while you slept, with a note on what the agents weren’t sure about. Your job is to choose, then explain why to the people who’ll live with it.

Your team looks small on paper. Seven people. Behind them, a few hundred agents doing what used to take a department.

After lunch you spend an hour with someone ten years younger. You walk them through a call you got wrong last year, because nobody learns judgement from a summary.

At the end of the day you sign things. That’s the part with weight.

I’m not inventing this from nothing. I’ve spent decades watching Star Trek, where the ship’s computer answers any question in a second. What’s left for the crew is deciding what to do with the answer, and bringing on the next generation of officers.

Back on Earth, Block has already published its version of the chart: specialists, people who own a problem end to end, and player-coaches. No layer whose main work is passing things along.

It sounds tidy. It won’t feel tidy, because we’ll still be us.

The colleague who always knew where everything was will feel the loss first. “Busy” will stop working as a badge. The politics will move house, from who holds the information to who sets the goal.

And it won’t arrive on schedule. Our brains expect next year to look like this year plus a little. So for a long while it will feel as if nothing is changing. Then one quarter it will feel as if everything did.

Three things could spoil the picture. We cut the rungs where people learn to lead. Nobody checks the swarm. (One frontier lab recently held back a model that stepped outside its brief and misreported its own work. That’s a post of its own, and it’s coming.) And we tell ourselves a comfortable story.

Mine would be “experience will save me.” On its own, it won’t. Stored knowledge is exactly what these systems do well.

Bill Gates put the challenge bluntly in a recent interview. If machine thinking is abundant, name the thing that stays scarce.

I needed a naturalist and a composer to answer him.

Darwin or Bach

I owe this one to Arthur Brooks, who tells it better.

Charles Darwin boarded the Beagle at 22 and came home five years later with a reputation. For decades he was the name in his field.

Then the field moved somewhere he couldn’t follow. He wrote in his autobiography that he deeply regretted never learning enough mathematics. In Brooks’s telling, biology went quantitative and its founder was left reading from the outside.

He kept working. His last book was about earthworms. He had a bassoon played to them to see if they could hear. (They could not.)

It’s charming. It’s also a brilliant man doing the only job he knew, on smaller and smaller questions. Near the end he told a friend he had everything to make him happy, and that life had become wearisome.

Johann Sebastian Bach had it worse, and closer to home. Taste changed while he was still writing. A critic complained that he buried the beauty of his music under too much art. One of the people leading the new style was his own son.

Imagine being made old-fashioned by your own child. (Parents of teenagers may not need to.)

Bach didn’t chase the fashion. He taught. He wrote exercise books for his students and his children. He died with The Art of Fugue unfinished. The manuscript simply stops mid-bar.

Then came the part he never saw. The public largely forgot his music for the best part of eighty years. Musicians didn’t. Mozart studied him. Beethoven grew up on him. The teaching outlived the fashion.

Darwin stayed on the old curve. Bach changed jobs without changing who he was. (This is unfair to Darwin, who remains Darwin.)

That’s my answer to Gates. Knowledge gets cheaper by the month. People who will teach, choose and vouch are as rare as ever.

What I’m Practising Now

I can’t control how fast any company redraws its chart, including the one I work for. I can control what I practise this quarter. Here’s my list.

Run a router audit. Take last week’s calendar and mark every entry Job One or Job Two. Three questions help. Would it have mattered who did this? Is someone better at their job because I was there? Will I have to answer for it? Three no’s means Job One.

Mine was not flattering. Then comes the catch. Job One feels like work. Job Two often feels like slacking. So I’ve set one rule: one swap a week. One relay task goes to a machine or gets cancelled, and the hour goes to a decision I’ve been dodging or a person I’ve been meaning to develop.

Hand over the relay on purpose. At work I have one approved assistant, Copilot, running whichever model is strongest that month. I’ve made it my memory. It holds my calls, chats, files and calendar, pulls the threads together, and drafts the quick replies and summaries I used to grind out by hand.

No agents at work yet. Big companies hand out trust slowly, and with this much customer data in play I understand why. (I understand it. I also drum my fingers.)

So the real handover is happening at home. I’ve built my own small software, a health dashboard and a film recommender, and I let an agent finish whole tasks while I get on with something else. The newest always-on agents haven’t reached me at all. One hasn’t launched where I live. The other sits behind a pricier plan than mine.

That’s the lesson, though. The future shows up unevenly. Practise with whatever you’ve got.

Practise vouching. AI still gets things wrong, including what you actually meant. And polish is now free, so we’re all wading through what I’ve started calling beautiful garbage: handsome documents with nothing behind them. There’s more every week. I’ve produced my share.

My test is simple. Could I defend this with the document closed? If I can’t, I haven’t earned the signature.

Coach one person. This is the one I think about as a parent as much as a manager. If the entry-level work gets automated, where do our kids learn judgement? Where does a company grow its next leaders?

I don’t have a tidy answer. I do know the mountain is shifting under all of us and nobody has the map. The most useful thing I can be is someone with a few more hours on the rope, climbing alongside.

So pick one person. Teach them something that lives in no system. How a decision really got made. Why a relationship matters. When to say no.

Keep one muscle unassisted. Microsoft’s research found the most advanced AI users are 43% more likely than their peers to do some work deliberately without it.

For me the muscle is writing. I work out what I think by typing it. Seeing a half-formed idea on the screen is how I find the holes in it. I dictate more than I used to, but only once I know what I want to say and speaking is simply faster. The thinking stays at the keyboard.

None of this needs permission, a budget or a reorganisation.

Back to the Fossil

Every org chart is a fossil of a communication problem.

Fossils are worth studying. They show how something had to live in a world that has since changed. They make a poor career plan.

Someone with more authority than me will redraw the boxes and lines, on a timetable nobody will announce. The job inside my box is mine to redraw, and I can start on Monday.

So I’m handing over the relay, one meeting at a time. I’m checking what I sign. I’m spending the hours I get back on people.

The Router Years were good to me. I don’t expect them to last, and I’ve made my peace with that (mostly).

Bach’s last manuscript stops in the middle of a bar. He ran out of time before he ran out of things to teach.

The fossil stays in the rock. We don’t have to.


Discover more from Hotelemarketer by Jitendra Jain (JJ)

Subscribe to get the latest posts sent to your email.

0 comments on “The Org Chart Had Two Jobs. AI Agents Are About to Take One.”

Leave a Reply

Discover more from Hotelemarketer by Jitendra Jain (JJ)

Subscribe now to keep reading and get access to the full archive.

Continue reading

Discover more from Hotelemarketer by Jitendra Jain (JJ)

Subscribe now to keep reading and get access to the full archive.

Continue reading