A friend forwarded me Salim Ismail‘s new book, The Organizational Singularity, which starts with why companies look the way they do.

Ismail goes back to the economist Ronald Coase, who argued in 1937 that firms exist because coordinating work through the market was expensive. So it was cheaper to bring people under one roof and put them on a payroll. The traditional org chart and its layers of managers are just the hierarchy that grew out of that, built specifically to move information and decisions between people. Ismail’s core argument is that AI is driving the cost of that coordination toward zero, which means the structure built around it starts to come apart. The company remains a legal shell someone is accountable for, but the people inside stop rubber-stamping everything as it goes through and start checking for exceptions instead.

The clients I met around Southeast Asia are working on a different problem: They are looking at Copilot licenses that some people use with enthusiasm and others never touch, a few “vibe coding” experiments; and urgent requests for help with governance before things get out of hand. In my last piece, I argued for getting AI tools directly into the hands of your employees. Reading the book made me wonder whether that goal is too small. In Indonesia, I have not heard anyone even talk about agents, let alone ask which parts of their company exist purely to pass information from desk to desk.

Ismail argues that when AI speeds up the work itself but leaves the decision layer untouched, the faster work just piles up waiting for approval. He calls this pile-up “congestion.” If an underwriter learns to prepare a file in half the time with AI, and that file still has to wait for the same chain of signatures it needed last year, the company hasn’t gained very much. You can put every single one of those couple of thousand people on steroids, and you still end up with a building full of fast people waiting on each other.

To put it in one line from the book: “Hierarchy is not a law of physics. It is a coordination technology.” One of Ismail’s strongest examples is Haier, the Chinese appliance maker that has been splitting itself into thousands of small micro-units with their own customer accountability since 2012. Haier is not Southeast Asian, but it is a lot closer to home than Silicon Valley.

Starting at the Edge

The book’s answer to where a large company should start is something it calls an Edge Twin.For a company with more than fifty people, you look for the function where the most time goes into passing work between people instead of actually making decisions. Then, you set up a team of three to five people at the edge of the organization. That team runs the function the new way - with agents doing the coordinating - and proves it works before any process moves over from the main company. I like this idea because a handful of people and a single function is an experiment I can actually imagine a director asking a board to fund.I am a strong believer in this kind of model. Give a company a small team of rebels with freedom, a token budget, and room to play, and let them try out new things. That team could take specific customer problems and use AI to solve them in hours or days. Or it could be an enablement team that figures out the tools first, then goes around the organization fixing things that are broken so people actually save time.Of course, a team like that also looks very familiar. During our years of agile consulting and team transformation work across Indonesia and Southeast Asia with Ekipa, we ran pilot teams, and the results were always mixed. The outcome usually depended on whether a senior leader actively protected the pilot. I see no reason an Edge Twin would escape that reality. In fact, it would probably need even more protection than an agile pilot did, because besides working faster than the teams around it, it quietly exposes which coordination roles the company might no longer need at all.

The Missing Pieces

I would also want to know who those five people are. In my last piece, I wrote that most companies here don't have people inside who can do the enabling. An Edge Twin needs people who are significantly further along than that. They need to work with agents every day and be able to explain to a risk team why an algorithm is allowed to make a certain call. Whether a bank's compliance and risk teams would let such a team run in the wild is another open question. The book has a detailed governance section, with every agent decision tracing back to a named human owner, and I suspect that part would interest a local bank a lot more than the grand vision does.I don't know whether companies here can skip a step.One version says you go through the slow work first: get people comfortable with AI one at a time, build the governance, learn where the work piles up, and only then redesign the organization around what you’ve learned.The other version, the one my belief in rebel teams pulls me toward, says that three years of enablement inside an unchanged structure just gives you fast people waiting on slow approvals, and the companies that started a small team working the new way will be miles ahead by then.What my years doing agile tell me is that whichever version a company picks, it will need a senior leader willing to protect the people trying it out. And looking at a book or a conference panel, I still can’t tell you how many of those leaders Southeast Asia has.

We’ll get back to you fast

Tell us what you need and our team will reach out from across Southeast Asia.

We’ll get back to you fast

Tell us what you need and our team will reach out from across Southeast Asia.

Ekipa
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.