Nobody Wants to Move the Card

Work · Attention · Part two of two

Nobody Wants to Move the Card

Kanban was invented so that nobody would have to manage anything. Software turned it into a second job. Trello hit the ceiling first. Monday.com and the rest are next, because what they bill for, human attention spent inside a board, is exactly what AI removes.

By John Rector · October 30, 2026

Nobody wants to manage tasks. Not the engineer, not the designer, not the founder, not even the project manager, who mostly wants the project done. We tolerate task management because we were told it is the price of coordination. It is worth asking where that price came from, because the original idea was the opposite: a system so simple that nobody had to manage anything at all.

What kanban actually was

In the early 1950s, Taiichi Ohno was trying to stop a Toyota machine shop from making parts nobody had asked for yet. Parts were pushed downstream whether they were needed or not, and they piled up. He took his answer from the American supermarket, where a customer takes what is needed, when it is needed, and the shelf gets restocked because something was taken. Ohno put a card on each bin of parts. When a worker used up a bin, the empty bin and its card went back upstream, and the card itself was the order. Kanban means signboard.

The genius was not the card. It was where the card lived. It rode on the work. Nobody had to remember to report that parts were running low; running low is what sent the card. The signal was a byproduct of doing the job. That is why Ohno could say kanban automated production control.

Figure 01

The same card, two systems

Toyota, 1950s

1. A part is used
2. The bin empties
3. The card rides upstream with the bin
4. A replacement is made

Attention required: none

Software board, 2010s

1. Work gets done
2. A person remembers it got done
3. Opens the app, finds the card
4. Drags it to the right column
5. The board is right, until it is not

Attention required: every time

Humans are terrible at lists

The software board asks a person to do something people are famously bad at. Working memory holds about four things at once. Most of us cannot drive home and remember six items from the grocery store. A list is a promise to pay attention later, and most of us default on it, not because we are lazy but because attention is the scarcest thing we have and a list is the least interesting place to spend it.

So every board decays. It is perfect the afternoon it is set up, clean columns, full of hope. From that moment it drifts away from the truth at exactly the rate its owners’ attention wanders. Everyone has seen a board graveyard: a beautiful workspace last touched in March.

Why Trello really failed

Trello was the best version of the board anyone built. It was simple, visual and loved. Atlassian agreed to buy it in January 2017 for about $425 million, when it had more than 19 million registered users. Trello did not die. It did something more instructive. It hit the ceiling every board hits, and the ceiling is not design. It is the double entry.

The tell came in 2025. Atlassian rebuilt Trello around a new Inbox that uses AI to pull to-dos out of email, Slack and Microsoft Teams and extract the due dates and action items for you, plus a Planner that puts them on your calendar. Read that as a confession. After nearly fourteen years, the board’s owner concluded that the problem was never organizing the cards. It was that people would not make them. Capture, the part that used to cost a person’s attention, was the product all along.

But notice what the new Trello still does. It uses AI to fill in a board that you then have to manage. Elon Musk’s five-step algorithm, as Walter Isaacson records it, warns about exactly this: a common mistake is to optimize a part or process that should not exist, and automation comes last, after you have deleted everything you can. Automating the card is step five performed on a step that should not have survived step two.

Why Monday.com will follow

Monday.com, Asana, Smartsheet and their peers did not really sell boards. They sold seats. A seat is a person logging in to update the state of the work. Strip away the branding and the whole category bills for one thing: human attention spent inside the board. That is the line on the invoice.

Asana’s founders even named the disease. They called it “work about work”: the status updates, the chasing, the searching, the coordinating. Asana’s own research put it at about 60 percent of the average knowledge worker’s day. The trouble is that a task board is made of work about work. It can make coordination cheaper. It cannot stop being coordination.

The market has started to price this. Monday.com trades roughly four-fifths below its November 2021 peak. Asana’s shares are down more than 90 percent from theirs. Smartsheet left the public market in January 2025 in an $8.4 billion take-private. Those numbers are not the argument. The argument is structural. The numbers are just what the structure looks like from Wall Street.

Monday’s answer is agents. At its Elevate conference this week, every attendee was invited to build one. That is the strongest counter to my case, and I take it seriously: if Monday becomes the place where agents do the work, the seat becomes an agent and the business survives. I do not think it works, for a simple reason. Agents do not need a board. A board is a picture of the work drawn for a human with limited attention. An agent can read the work directly: the sent email, the merged code, the signed contract, the paid invoice. Whoever holds that record holds the true state of the work. The board becomes a free view of it.

Nobody wants to manage tasks. AI is perfect at it.

Here is the turn, and it is simpler than the software industry would like. Nobody wants to manage tasks, and AI is perfect at it. Not because AI is a faster card-mover. Because it does not need cards.

An AI with access to your record does what Ohno’s card did. It knows the work happened because it can see the work happen. The email went out, so the item is done. The contract came back signed, so the next step is due. The supplier asked a question, so someone has to answer by noon. The signal rides on the work again. Task management stops being a job a person does and goes back to being a byproduct, which is what it was the first time somebody got it right.

That is the zero-attention economy in one small corner. The useful machine is the one that keeps the state of the work without asking anyone to describe it.

Five predictions

1. Monday.com is sold, merged or taken private by the end of 2030, at a value below its November 2021 peak.

   Pennies on the dollar is the likely shape, relative to what the market once believed a work-management company was worth. The buyer will want the customer list and the record, not the board.

2. No standalone work-management company regains its 2021 peak by 2030.

   Not Monday, not Asana, not a private challenger raising at those numbers. The category’s ceiling was set by human attention, and that ceiling is falling.

3. The board survives as a view, not a workplace.

   Boards will be generated on request from the record, read-only by default, the way a dashboard is. You will look at one when you want a picture. You will not update one.

4. Seats give way to outcomes.

   As fewer humans log in, pricing in the category shifts from per seat to per agent or per outcome, and vendors quietly stop leading with how many people use the product every day.

5. The survivors go niche: compliance, not coordination.

   Where a signed human record of who approved what is itself the product, in construction, regulated manufacturing, pharma and government contracting, boards live on as audit trails. Everywhere else they become a feature, sold for a fraction of what the category was once worth.

Where I am in this

I came up in supply chains. E2open, the company Paul Stern and I started, was built to move signals between companies’ systems so that partners could act on what was actually happening. That world taught me early that the best signal is the one nobody had to type. Today my working life runs on two short instructions to my AI: You do it. You decide. I describe the state change I want; it keeps track of what moved. I wrote a book about this, The Zero-Attention Economy, and the task board is the cleanest case I know of a product whose value was measured in the attention it consumed.

The ledger

Already true

Trello’s owner rebuilt it around AI that captures to-dos from email and chat. Monday.com and Asana trade far below their 2021 peaks, and Smartsheet has gone private. AI can already read an inbox, a code repository or a ledger and say what changed.

What has to happen

Companies have to let AI read the record, the email, documents, code, finance and calendars, with permission to say what moved and increasingly to act on it. That is a trust question more than a technical one, and it is why I argue every business needs a policy for people who bring their own AI to work.

Where I am probably wrong

Lists do real psychological work. Research by E. J. Masicampo and Roy Baumeister found that making a specific plan for an unfinished goal quiets the mind’s nagging about it. If what people really buy from a board is relief rather than coordination, they may keep paying for the feeling long after a machine knows the state of the work better than they do. And enterprise software dies slowly: a vendor that becomes the compliance record, or the place agents are governed, can outlive its founding idea by a decade.

Go back to the stand-up. Eight people, one of them dragging rectangles, everyone waiting. Ohno solved this problem seventy years ago by putting the signal on the work. We are about to solve it again the same way.

Nobody will move the card. The work will.