What is Lean AI? | Lean AI, Lean Enterprise Institute

Why now

AI is the same choice software was, moving faster.

Software was rarely designed with the value stream in mind. It optimized parts instead of enabling flow, and eroded teams' ability to improve their own work.

Adopted as a black box to automate tasks, AI will fragment the flow further and lock in the dysfunction already built into your systems. Adopted to build capability, it restores people's ability to see, improve, and change the work.

AI can make each layer adaptive again, if people understand the work, and the technology, well enough to guide it.

Tyson Heaton presenting the state of AI and lean to a room of practitioners

The principles

Nothing new. Lean, applied to AI.

The Toyota house stands on two pillars. Both are older than the computer, and both tell you where AI belongs.

flag

The precondition: leaders go first.

Nobody learns AI from the back of the room. Leaders don't need to build the enterprise's tools, but they should build for their own work, hands on. When they use AI in the open with their people, it becomes safe for everyone to experiment and share. And everyone needs a good coach, leaders included.

pan_tool Jidoka: automation with a human touch

Stop and call a person.

Sakichi Toyoda's mother farmed by day and wove by night on a hand loom, and he set out to ease her work. Over decades of looms, his machines came to stop themselves when a thread broke, before the cloth went bad. No one had to stand watch, braced for the next break. One person could run several looms, with headroom to make them better.

Ask the same of any AI system you build. Can people see what it's doing, and why? Does it stop when something's off? Does it grow their ability to fix and improve the work, or pile on stress, with the technology taking the easy work and leaving people the cleanup, the edge cases, and the breakdowns?

groups Lean AI for Teams arrow_forward

double_arrow Just-in-time

What's needed, when it's needed.

The target is work that flows steadily at the pace of customer demand. Improving a step helps when it serves that.

Put AI where it takes variation out and moves the flow, and only where people know the work well enough to spot when it drifts. Start with the improvement work itself.

account_tree Lean AI for Value Streams arrow_forward

Under both: stable, standard work is the floor, and people are the center.

Lean AI in a transformation

The Lean Transformation Framework

Transformation isn't random. It requires answering five questions. Select a part of the house to explore it.

Framework by John Shook, Lean Enterprise Institute

flag

Purpose

build

Work

manage_accounts

Management
System

school

Capability

psychology

Basic Thinking & Mindset

Lean AI doesn't change the questions. It can speed up the transformation and extend what your people can do. Leave their capability for later, and you take on debt: the business gets more complex while the people running it get less able to change it.

Your turn

Try it on your own problem.

A prompt for your own AI. It coaches you through the five questions, asks where AI helps and where it hurts, and ends with the people you need to go see. Or bring the problem to a discovery session, and we'll walk it through the five questions with you.

You are a lean coach. Help me think through a change I'm considering, using the Lean Enterprise Institute's Lean Transformation Framework. It asks five questions:

1. Purpose: What is the value-driven purpose, or the problem to solve?
2. The work: What work needs to be done to solve it, and how will it be improved?
3. Capability: What capabilities do people need to do that work?
4. Management system: What routines, measures, and leadership behaviors are needed?
5. Basic thinking: What mindsets and assumptions does this require?

How to coach me:
- Ask one question at a time. Wait for my answer.
- Don't answer for me. If I'm vague, ask follow-ups until it's specific: who, where, how much, how we'd know.
- Don't let me jump to a solution. If I name a fix before we understand the problem, ask what we'd need to see first.
- Start with purpose. If I say the goal is "to use AI" or "to roll out a tool," ask what problem it solves for the customer or the business. Keep going until there's a real problem with a measure.
- After each answer, ask where AI could help and where it could hurt. Push me on whether AI would help people do and improve the work, or just do it for them.
- Point out when one part doesn't fit another, like a new process nobody is trained to run, or a new measure leaders won't review.

When we've covered all five, give me:
- A table: each part, my answer, where AI could help, the risk.
- The weakest part of the house, and why.
- A reminder that this is a first draft, not a plan. Change happens through people. Tell me who I need to see before going further: the people who do the work, the people it affects, and the leaders who need to agree.
- My next steps:
1. Go to the gemba, where the work happens. Watch the problem happen, and ask the people doing the work what they see.
2. Start nemawashi: one-on-one conversations with each person the change touches, before any meeting or decision. Bring the problem, not the solution. Ask what I'm missing.
3. Use catchball: take what I learn back into the draft, change it, and share it again, until the people involved see it as their problem too.
- Then help me plan my first conversation: who, what I'll ask, and what I'll listen for.

The prompt draws on Tyson Heaton, "Accelerating Lean Transformation with AI", and Art Smalley, "Technology × Behavior × Management = Impact" and "Prompting for Problem Solving".

Explore the full framework at lean.org arrow_outward

Build your people's capability now. Change the process with care.

AI is the same choice software was, moving faster.

Nothing new. Lean, applied to AI.

Stop and call a person.

What's needed, when it's needed.

The Lean Transformation Framework

Try it on your own problem.

What is lean practice?

Lean Enterprise Institute

Foundational reading

Lean Tech & AI Journal

Ready to put Lean AI into practice?