HeptaMS
What is Lean-AI
For software teams

You work in an agile way and you use agents. The lead time is the same.

Now coding agents arrive, the team writes code at a pace that was unthinkable two years ago, and at the end of the quarter the board shows the same number.

The cause

Why this happens

AI lowers the cost of production. It does not lower the cost of deciding, and it raises the number of decisions per unit of time.

That shifts the bottleneck. As long as a draft took days, production set the pace. When a draft takes minutes, the pace is set by what comes after: review, approval, alignment, clarification. The volume of work requiring review goes up; review capacity does not go up with it.

Scrum and Kanban never solved that part. They organised it. As long as production was the bottleneck, that was enough.

It also explains why rolling out tools without changing the system leads nowhere. Licences and training improve individual activities. The value stream stays as it was, and the organisation produces half-finished work faster, in front of the same queue.

Work Waiting
Without AI Lead time 100 %
With AI Lead time 89 %
Lead time of one item, split into work and waiting. The work shrinks to a quarter. The lead time barely moves.
The answer

What Lean-AI changes about it

For seventy years Lean has worked on exactly this question: not how to work faster, but where the work gets stuck.

For a software team under AI conditions, that means:

Measure waiting, not utilisation

How much of the lead time is work and how much is waiting. The answer surprises most teams.

Architecture ownership inside the team

When agents commit, architecture decisions are made inside the team, faster than any upstream review can respond.

Specify before you generate

Whatever is not decided beforehand gets decided by the agent.

Name the three responsibilities

Value, flow, architecture. Wherever they are vacant or improperly merged, the queue forms.

The evidence

Where things stand

74 per cent of companies expect to use autonomous agents at least moderately within two years. 21 per cent have a mature governance model for them today.

Deloitte, State of AI in the Enterprise, 3,235 leaders in 24 countries

Your case is not in the catalogue

A different starting point, a different mix of IT and business, a different construction site. A training programme cut to fit your organisation before the first date is set.

AI workflows running on live systems at your place? For business units