Operational Excellence Beyond Lean: What Modern Organisations Need

Lean and Six Sigma reshaped manufacturing and service operations over several decades. Their core insights remain as valid as ever: eliminate waste, reduce variation, and define value from the customer's point of view. But the toolkit was built for a world of clipboards, stopwatches and periodic audits. Organisations that rely on it alone are solving today's problems with yesterday's instruments.

Operational excellence beyond Lean does not mean abandoning Lean. It means keeping the discipline and adding three things the original practitioners did not have: continuous process data, automation that can handle judgement as well as rules, and a serious approach to designing work with the people who do it. This article sets out what that looks like and where to begin.

What Lean still gets right

It is fashionable to declare old methods dead. That would be a mistake here. Value-stream mapping still exposes the queues and handoffs where most delay hides. Root-cause analysis still beats patching symptoms. Standard work still makes improvement possible, because you cannot improve what is done differently every time.

Above all, Lean is a management system, not a set of workshops. Daily attention to performance, problems raised without blame, and improvement owned by the people doing the work: these are cultural assets that no software provides. Organisations that skip them and buy tools instead get expensive visibility into problems nobody fixes.

What has changed since Lean was codified

Process data is continuous. Where practitioners once relied on sampling and manual observation, ERP and workflow systems, sensors and process mining tools now show how work actually flows, across every case and not a sample. The gap between the documented process and the real one becomes visible, along with the rework loops and workarounds that no workshop would surface.

Automation can handle judgement. Earlier automation was rule-based and brittle: it broke when a form layout changed. AI-assisted tools can now read unstructured documents, classify requests, draft responses and, in agentic designs, carry out multi-step tasks under supervision. That widens the range of work that can be automated, and raises new questions about control.

Work has moved. A large share of operational work is now knowledge work, carried out in systems and inboxes, often by hybrid teams. Waste in these settings is harder to see than a pile of inventory. It looks like waiting for approvals, re-keying between systems, chasing information and attending meetings to find out status.

Employee experience is an operational variable. Processes designed without the people who run them get worked around. Disengagement and turnover are operational costs, and co-design with frontline staff is now a standard technique for making improvement stick, not a courtesy.

A modern operational excellence framework

The organisations getting the best results combine four elements, in a deliberate order.

• Lean discipline as the base. Value defined by the customer, flow mapped end to end, root causes pursued, improvement owned locally.
• Digital process intelligence. Use system data to find where time and effort really go, to size problems before choosing them, and to check that fixes hold.
• Selective automation. Simplify first, then standardise, then automate. Reserve AI for steps where judgement or unstructured input is the real constraint.
• Human-centred design. Build the new way of working with the people who will use it, and measure their experience alongside cost and speed.

The order matters. Automating a poor process produces a faster poor process, with the waste now hidden inside software where it is harder to see and more expensive to remove.

The goal of operational excellence has not changed: more value, less waste, better outcomes for customers and employees. The path to that goal has.

One more shift deserves attention. Improvement used to be episodic: a project, a kaizen event, a report. With continuous data, it can become a running system. Teams see their own flow measures daily, anomalies prompt investigation within days, and the benefit of a change can be verified instead of asserted. This is where decision intelligence meets operations: fewer dashboards, more decisions.

Where AI-assisted automation fits, and where it does not

Good candidates share some features. Volumes are high, inputs are messy, the rules are mostly clear but exceptions are common, and a person currently spends time reading, sorting or re-keying. Invoice handling, order entry from emails and first-line service triage all fit the pattern. Intelligent document processing is often the most practical entry point, and agentic workflow automation extends it to multi-step tasks.

Poor candidates are processes that are unstable, politically contested or due to be redesigned anyway. So are decisions where an error is costly and hard to detect. In those cases, use AI to prepare and recommend, and keep the decision with a person.

Two controls are worth building from the start. First, treat each automation as an asset with an owner, a performance measure and an exception route, so that someone notices when it degrades. Second, measure the whole flow, not the automated step. A step that runs in seconds is worth little if the case then waits three days in a queue for approval.

Where to start

Resist the large transformation roadmap. The highest-return first move for most organisations is a focused assessment of a few core processes, aimed at finding the two or three friction points where targeted action would produce fast, measurable results.

• Choose processes that matter. Pick two or three with clear customer or financial impact and an executive who owns the outcome.
• Look at the data before the workshop. Pull event data from the systems involved and see how cases really flow. Bring that evidence into the room.
• Walk the work. Sit with the people doing it. Ask what they chase, re-key and work around.
• Fix the simple things first. Remove unnecessary approvals, duplicate checks and handoffs. Many gains need no technology.
• Automate one step properly. Select a single high-friction step, automate it with clear ownership and exception handling, and measure the end-to-end effect.
• Make it a routine. Put flow measures in front of the team weekly, and give them time and authority to act.

The aim in the first quarter is evidence and capability, not coverage. A team that has improved one process with data, design and selective automation can repeat it on the next.

Common questions

Is Lean still relevant now that AI can automate processes?

Yes. AI changes what can be automated, but it does not tell you which work is worth doing, where the waste is, or what the customer values. Lean thinking answers those questions. Without it, automation tends to speed up activity that should have been removed.

What is process mining and do we need it?

Process mining reconstructs how work actually flows by analysing event logs from your systems. It is useful when processes are high volume, cross several systems and behave differently from their documentation. For simpler or low-volume processes, direct observation and basic system reports are often enough.

Should we improve a process before automating it?

Almost always. Simplify and standardise first, so that you automate a process worth keeping. The exception is where automation removes the need for a step entirely, in which case redesign and automation should be planned together.

Where Ganexa can help

Ganexa combines Lean discipline with process data, selective AI and change support, starting with people and the operating model before technology. Our business process improvement team can help you assess core processes, prioritise the friction points worth fixing and build the internal capability to keep improving, with organisational change and AI adoption support where new ways of working need to land. To explore where to begin, book a free consultation.