Journal  /  Strategy  /  The Bottleneck is moving

Faster with AI: designing teams that need less communication

AI has made building software faster, but the time saved now goes into coordination. The solution may lie in how you structure product, design and development teams.

Faster with AI: designing teams that need less communication
01

Every extra person adds communication

The cost of coordination does not grow in a straight line. In a team of n people, the number of communication lines is n(n-1)/2. A team of eight has 28 lines, a team of five has ten, and a team of two has one. Fred Brooks made a related point in The Mythical Man-Month back in 1975: adding people to a late software project makes it later.

AI makes this effect stronger. If each person can produce several times as much, a team spends a much larger share of its time agreeing on what to produce, reviewing it and handing it over. The ratio of talking to building shifts in the wrong direction.

The answer is not more meetings or better meeting tools. It is smaller teams with fewer dependencies.

02

Start with autonomous teams

A good starting point is Team Topologies by Matthew Skelton and Manuel Pais. The book focuses on how teams interact, how to reduce their cognitive load and how to improve the flow of work. It describes four team types: stream-aligned, enabling, complicated-subsystem and platform. Stream-aligned teams are the default. They own a product, service or user journey end to end, and their purpose is to deliver value with minimal hand-offs. The other three types exist to reduce the load on them.

For product leaders, the practical lesson is to organise teams around a customer journey or product area, give them what they need to deliver end to end, and make the interfaces between teams explicit. Every dependency between teams is a communication line you pay for every week.

Most product teams today are five to eight people: a product manager, a designer and a handful of engineers. That size made sense when development was the constraint. With AI, it is worth questioning. At Swat.io, we are now starting with two-person teams.

03

Then reduce the need to agree within teams

Team Topologies reduces the communication between teams. The next step is to reduce it within them. If AI lets a small team cover the ground a large one used to, the logical team size shrinks as well.

I believe the direction is two-person teams: a product person and a developer who together own a problem from insight to release. One line of communication, decisions made in a conversation rather than a meeting, and no handovers.

This does not remove the need for specialists. Platform and enabling teams still provide shared services, design systems and deep expertise. But the day-to-day work of discovering and shipping a solution sits with two people.

04

Collaboration is not the problem, waiting is

Smaller teams can sound like lonelier work, and that would be a loss. For many people, the best part of their job is working with others: solving a hard problem together, learning from a colleague who knows more, the energy of a good discussion. That is not the communication to cut.

What slows teams down is a different kind of communication: waiting for a decision, a review, a handover or a free slot in someone's calendar. People enjoy learning from each other. I doubt waiting for someone else is on anyone's wish list.

So the aim is to take dependencies out of the delivery path, not people out of each other's working lives. Pairing, demos and communities of practice stay, because that is where people learn. The meetings that exist only to get agreement before anyone can move are the ones to remove.

05

What this means for roles

Two-person teams change what each role needs to cover. Product managers and designers become more T-shaped: deep in one area, but able to do enough of the other. A product manager writes a clear brief, sketches the flow and builds a prototype with AI. A designer understands the customer problem and the business case well enough to make prioritisation calls. In some teams the two roles will merge into one product role.

On the engineering side, full-stack developers become more important. When one developer owns the whole solution, a split between front end, back end and infrastructure brings the handovers back. AI helps here: a developer who is strong in one layer can now work productively in the others, with platform teams covering the parts that need deep expertise.

The hiring profile changes with it. Breadth, judgement and the ability to own a problem end to end matter more than narrow specialisation.

This is part 2 of the series The Bottleneck is moving. Part 1 looked at human attention as the new constraint; this one looks at communication within and between teams. If you are experimenting with smaller teams, I would like to hear what works and what does not.

Next Article

Communicate product ideas faster with demo videos
Product

Communicate product ideas faster with demo videos

5 min read