AI Won’t Fix a Broken Process: Why Businesses Need Lean Before Automation

In this episode of "The Manufacturing Evolution: AI, Ops & The Future of Work," we delve into a critical truth for businesses: AI won't fix a broken process. Join us as we explore why a solid foundation in Lean principles is essential before embracing automation. We unpack how to effectively streamline manufacturing operations using the People Process Technology framework, emphasizing the human element in automation. Discover actionable strategies that can help you identify and rectify inefficiencies, enabling your organization to fully leverage AI and robotics in operations.

Welcome to The Manufacturing Evolution: AI, Ops and The Future of Work. I’m Ron. Today’s truth is simple: AI won’t fix a broken process. Automation amplifies whatever exists, good or bad. If flow stutters, data is messy, or roles are unclear, robots and dashboards spread pain faster. For New Jersey, Philadelphia, and Delaware Valley manufacturers, and SMB plants nationwide, I’ll keep this practical and jargon light. We’ll build a Lean foundation, then add right sized technology that truly moves the needle.

Quick alignment on the key terms. Lean is about creating more value with fewer resources by reducing waste and variation. People, Process, Technology is a useful way to think about improvement: equip your people, build clear and efficient processes, then use technology to support and scale what already works.

Technology should follow good standards, not define them. Automation amplifies whatever is already there—it can make strong processes faster and more consistent, but it can also accelerate poor handoffs, assumptions, and workarounds. That’s why the starting point should always be clarity, stability, and visibility.

How can you tell when a process is broken? Look for signs like late orders despite overtime, constant firefighting, inconsistent cycle times for the same SKU, high levels of scrap and rework, WIP piling up between steps, and inventory increasing while customers are still waiting.

The data often tells the same story: multiple spreadsheets showing different answers, important knowledge living only in people’s heads, and reports arriving too late to influence decisions. If your mornings are spent expediting and your afternoons are spent apologizing, something in the process needs attention. Before adding new software, take time to understand how the work actually flows, where it stalls, and what is creating the delays.

Go see the work for yourself. A short Gemba walk gives leaders and support teams the opportunity to observe where value is actually created, with operators guiding the conversation. Follow a single order from release through packout, paying attention to waiting time, handoffs, bottlenecks, and rework.

Then map the current process using something as simple as sticky notes. Lay out each step in sequence, use arrows for handoffs, triangles for queues, and clocks for delays. Keep the exercise focused—an hour is often enough to uncover valuable insights—and photograph the finished map for reference. The objective isn’t to create a perfect-looking diagram. It’s to give everyone a shared, realistic view of how the work actually happens.

Now look for waste in the process. Motion shows up when operators walk unnecessarily for tools, parts, or labels. Waiting happens when machines or people sit idle for approvals, materials, maintenance, or information. Defects include mismarked parts, incorrect revisions, failed tests, and rework.

Overprocessing can mean entering the same information twice, adding unnecessary inspections, or printing documents that could stay digital. Overproduction happens when teams make more than customers need, often based on forecasts that quickly change. Transport waste appears when pallets, parts, or materials are moved repeatedly because the layout doesn’t support good flow. Excess inventory and WIP tie up cash, consume space, and often hide deeper process problems.

The important point is that many of these issues do not require expensive automation to solve. Often, the biggest improvements come from simplifying the process, improving the layout, reducing handoffs, and making the work flow better before introducing robots or new technology.

Before introducing automation, establish a clear baseline. Measure the things that reveal how the operation is really performing: total lead time from order to delivery, OEE at the constraint, first-pass yield, scrap rate, changeover time, throughput, on-time delivery, and WIP across each area.

Then focus on stabilizing the process. Create simple standard work for your highest-volume products, use visual controls so teams can immediately see what is normal and what is abnormal, apply 5S to organize tools and materials, introduce error-proofing where mistakes commonly occur, and strengthen preventive maintenance.

As variation decreases, the process becomes more consistent and predictable. That is when technology and automation can add real value—supporting a stable process rather than trying to compensate for an unstable one.

Next, improve flow by identifying and addressing the constraint. Use throughput and queue length to find the true bottleneck, then protect it with clear scheduling, appropriate buffers, and fast support when problems arise.

Reduce changeover time using SMED principles: move as much preparation as possible outside of machine downtime, standardize tooling and setups, and use simple checklists to make changeovers consistent. Balance work across people and equipment so capacity is aligned with takt time and demand.

At the same time, prepare the data foundation. Clean up part numbers and routings, establish one trusted system of record, create straightforward rules for updating information, connect only the machines that provide meaningful data, and validate that the signals being collected are accurate. Better flow and reliable data create the foundation technology needs to deliver useful results.

Once the flow is stable, choose automation opportunities deliberately rather than automating everything at once. Evaluate potential projects based on business value versus implementation complexity, and start with the opportunities that offer high value without creating unnecessary complexity. Repetitive tasks such as pick-and-place operations, palletizing, label printing and application, or automated data capture are often good places to begin.

Then right-size the technology to the problem. That might mean inexpensive sensors to capture event timestamps, cobots for flexible repetitive work, a straightforward MES for tracking WIP and downtime, or lightweight ERP integration rather than a major system overhaul.

Most importantly, keep the approach human-centered. Design safe work cells, rethink roles around the new technology, and give employees the training they need to work effectively with it. The goal is not simply to add technology—it is to make technology fit the way people and processes actually need to work.

Legacy equipment does not automatically mean you need a full replacement. Many older machines can be retrofitted with clamp-on sensors, I/O modules, and gateways that translate machine signals into usable data. Retrofit when the process is already stable, the equipment is still reliable, and the machine has enough useful life left to justify the investment. Replacement makes more sense when reliability, capability, or safety can no longer meet customer demand at a reasonable total cost.

Funding can also make modernization more accessible. For small and mid-sized businesses in New Jersey, Pennsylvania, and Delaware, regional MEP centers and workforce development programs may provide access to grants, training funds, technical support, and partnerships that can help offset the cost of assessments, implementation, and employee upskilling.

A practical way to move forward is with a simple 30-60-90 day plan. In days 1–30, conduct Gemba walks, map one value stream, establish baseline metrics, and complete a focused 5S effort. In days 31–60, address one key bottleneck, run a SMED event, standardize the work, and clean up the critical data needed for better decisions. In days 61–90, select one automation opportunity, run a pilot with clear success criteria, document what you learned, and decide whether to scale, adjust, or stop.

The takeaway is simple: Lean first, then right-sized technology. Build clarity and stability into the process before adding automation or AI. That is how technology becomes a tool for lasting business improvement rather than another layer of complexity.