Lean Six Sigma 30 Years Later: The Discipline AI Needs

Lean Six Sigma may look unfashionable, but it builds exactly the disciplines AI needs: process mapping, simplification, data quality, and statistical thinking.

Nearly nine in ten organizations now use AI somewhere in the business. Only about a third have begun scaling it across the enterprise. Fewer still can point to a meaningful impact on earnings.

The explanation is becoming familiar: AI must be embedded in the process, not bolted on as another technology layer. That is correct. It is also not new.

I watched GE confront the same underlying challenge nearly thirty years ago, before anyone called it digital transformation. The technology has changed dramatically. The operating discipline required to turn it into business value has not.

What AI over a broken process looks like

Most companies begin by using AI to automate the process they already have. A chatbot sits on top of a confusing policy. A copilot is inserted into a workflow with seven approvals. An agent is asked to manage exceptions that nobody has properly documented.

The process moves faster, but it carries the same waste, rework loops, unclear decision rights, and broken handoffs it had before, now executed at machine speed. AI does not repair a messy process. It amplifies it.

Lean is built for precisely this problem. It makes visible the steps that add no value for the customer, the errors and exceptions that keep looping back through the organization, the approvals that exist without genuine accountability, and the queues where work waits instead of moving.

Once those problems are visible, the organization can redesign the workflow around what AI makes possible. That is fundamentally different from dropping AI into the workflow it already has.

What I learned as a GE Black Belt

I joined GE in 1999 as a Black Belt. My first project was not glamorous: reorganizing a 50-person call center. I categorized the calls, mapped how they actually flowed, measured where time and quality were being lost, simplified the work, and introduced controls to make the improvements last.

At the time, it felt like a practical process-improvement assignment. In retrospect, it gave me the operating toolkit I would use in every integration, turnaround, and transformation I led afterward.

GE had launched Six Sigma company-wide three years earlier. Its first-year economics were hardly inspiring: the company reportedly invested $200 million and generated $170 million in savings.

Many organizations would have quietly reduced their ambitions. Jack Welch did the opposite. GE intensified the program and, crucially, connected management incentives to measurable Six Sigma results, not to training hours, participation, or the number of projects launched.

In 1997, GE reportedly invested approximately $400 million and generated around $700 million in benefits. Over the program’s first five years, the company attributed approximately $12 billion in cumulative savings to Six Sigma.

The tools mattered, but the management system mattered more.

I have carried that lesson into every transformation since: operating discipline does not survive on enthusiasm. It survives when leaders establish clear ownership, measure outcomes, and align incentives with results.

“Destroy Your Business”

The part of GE’s story that often gets overlooked is that Six Sigma was never only about defects and cost. Shortly after I joined, I became involved in an initiative with a name only Jack Welch could have approved: “Destroy Your Business.”

Welch was watching the rise of Amazon, then still known primarily as an online bookstore, and did not want GE businesses to wait for digital competitors to redefine their markets. Each business was challenged to attack its own model first: to reconsider how it sold, served customers, and operated in the emerging internet economy as though it were the disruptor.

Lean Six Sigma became an important instrument for doing that work because digital transformation required us to rethink processes, not merely put them online. You cannot improve a customer journey you have never mapped. You cannot digitize a process effectively by copying every approval, workaround, and organizational boundary from the old way of working.

The process must first be understood. Then it can be simplified, automated where appropriate, and rebuilt around the needs of the customer. Beyond the reported savings, that is an important part of GE’s operating legacy. Lean Six Sigma gave thousands of teams a shared way to take work apart and rebuild it for a new channel.

Thirty years later, many companies are reaching the same conclusion about AI, usually after their pilots begin to stall.

How the method evolved

By the time I became a Master Black Belt in 2005, the methodology had evolved. Watching that evolution from inside GE taught me as much as the individual tools did.

First, process improvement became a leadership responsibility.

The people leading the work were increasingly experienced operating executives, leaders with credibility inside the business and enough organizational weight to remove obstacles and accelerate decisions. The method was no longer treated primarily as the domain of statistical specialists or a separate improvement function.

Second, the emphasis became more practical.

The sophisticated analytical tools remained available where they were needed, but everyday improvement increasingly focused on flow, waste, speed, and simplicity. The objective was not to demonstrate mastery of the methodology. It was to improve the performance of the business.

Third, GE pushed the capability into line management.

Green Belts were managers close to the people doing the work. They learned the methodology while leading real improvement projects in the processes they owned. That created an adoption engine: hundreds of managers capable of translating a central discipline into local action.

Change imposed by outside experts may be tolerated. Change led by a manager who understands the work, and designed with the people who perform it, is far more likely to be adopted.

Simplify the method. Put operating leaders behind it. Build the capability close to the work.

Whatever label a transformation program carries, I have followed those three principles ever since. They are especially relevant to AI, where technology often advances faster than the organization’s ability to redesign work around it.

Three questions every AI program should answer

Lean, Six Sigma, and GE’s Change Acceleration Process address three issues that repeatedly prevent AI programs from creating value.

1. Lean: Are we redesigning the right process?

Map the process as it actually operates today, not as the procedure manual says it should operate.

Identify the queues, handoffs, rework, workarounds, and unnecessary approvals. Then redesign the complete flow around what AI now makes possible.

Much of the value will come from removing steps, combining responsibilities, and eliminating delays, not from accelerating one task inside a process that should no longer exist in its current form.

2. Six Sigma: Can the data support the decision?

Every AI initiative needs a baseline, a target, and an accountable owner. Those must be defined before implementation, not reconstructed afterward to declare the project a success.

The organization must also decide who owns performance after go-live: the thresholds, monitoring, exceptions, and response plan.

An improvement without control is a temporary event.

3. Change Acceleration Process: Will people adopt the new way of working?

AI changes tasks, decision rights, roles, and sometimes headcount. Employees usually understand that before management addresses it directly.

The teams expected to use the new process should therefore help design it. They understand the operational exceptions that a central AI team will rarely see.

Exclude them and the organization will get workarounds, selective adoption, and quiet resistance, often presented as a request for more training.

The capability may already be inside the business

Many organizations already employ Black Belts, Lean leaders, operational excellence teams, and experienced process owners. They may simply sit far from the AI program, under titles that are no longer fashionable.

Used well, these people can provide the bridge between AI specialists and the business. They can help select the right problems, redesign the processes, validate the data, define controls, and bring the organization through the change.

AI expertise does not replace process expertise. It depends on it.

The operating discipline behind AI value

Thirty years later, the pattern has not changed. The companies that scale AI successfully will treat it as an operating model transformation first and a technology implementation second. They will redesign the process, establish a trustworthy data foundation, assign ownership, measure results, and bring the organization with them.

That is the work I still do today: mapping how the business really operates, redesigning processes around what technology now makes possible, building the necessary data foundation, and ensuring the organization adopts the new way of working, not merely pilots it.

If an AI program has plenty of use cases but little impact on earnings, the missing element is often not a better algorithm. It is operating discipline.