Automation, AI, and the Risk of Doing Things Efficiently That Should Not Be Done at All 

You may remember seeing the results of an MIT study from July 2025 stating that 95% of generative AI pilots failed to produce meaningful results for the companies surveyed. That was the headline. The study unpacks several success factors, but the article highlights a common pattern we should all be aware of if we want use AI tools in ways that will truly benefit our business. 

The thing to watch for is whether AI is actually helping you add value for your customer by solving problems. Let me explain. 

Parade of the ants  

It was the late 1980s before the term “lean” was coined, and I was an engineering student at General Motors Institute. We were touring a manufacturing company on the west side of Michigan. Central to their process was a fleet of automated guided vehicles (AGVs) which scurried around the plant, following paths in the concrete reminiscent of the geoglyphs of the Nazca Desert in Peru. Their function was to pick up inventory from the various production departments, shuttle it to a large, automated stacker crane for storage, and then, when the materials were needed, shuttle it on to the consuming process.  

As budding engineers, we were in awe. How cool! We all wanted to quit our jobs and go work there. What we didn’t understand at the time was that rather than solving the problem of how to manage all that inventory, this company should have focused on the problem of avoiding the extra inventory in the first place.   

Automation can be good. It has a time and a place as a countermeasure to a specific problem. But sometimes, in the name of using the latest technology, we abuse automation. We forget that before considering the use of a new tool, we need to understand how it supports the design of the value stream. 

Tripping over your own feet 

A few years ago, I visited a company that was debugging a robot with a vision system. The robot’s task was to grab the right size and color of stopper from a jumbled bin and then place the stoppers, with the correct force, into holes on the product. After this step, the product would be dip-coated, and after drying, the stoppers removed. Placing the stoppers was done by hand, and the objective was to improve productivity and quality by having a robot place the stoppers. The vision system and robotic gripper were having a terrible time finding the correct stopper and reliably grasping it. The new approach ended up being much slower than the manual operation, and it was questionable whether the quality of the stopper seals improved.   

After observing the robot struggle for 20 minutes, I asked if we could go to the incoming stock area and see how the stoppers arrived from the vendor. I am not making this up: the stoppers were organized in separate boxes by size, and within the boxes, they were oriented in trays. So, in the name of using the latest technology (a vision system), the new process took fixtured and sorted stoppers and dumped them together so the robot could sort them back out while orienting them. This reminded me of the Peter Drucker quote, “There is nothing so useless as doing efficiently that which should not be done at all.”   

There is nothing so useless as doing efficiently that which should not be done at all. 

Think about the wasted resources leading up to this point: The designing and production of the trays; placing the stoppers into the trays; stacking the trays into the boxes; and finally, the effort of undoing all this work so a robot can redo it. 

So what? Just a few weeks ago, a colleague shared with me how they had used AI to automate a menial task that took up a lot of time each month. Although impressive on the surface, I couldn’t help but think, “Should this task even be done in the first place?” Not asking this question is a substantial risk for organizations rushing to use AI to “improve” the work. It is relatively easy to create an AI tool, and it is so much more difficult to first understand one’s processes and the value stream they support. 

Guiding principles 

Fast forward a few years down the road, and we may just have AI tools churning out waste (activities that consume resources, but do not create value), and because that waste is largely invisible, it will just go on and on. How do we avoid this trap? I coach leaders to leverage the simple advice of authors Jim Womack and Dan Jones in Lean Thinking:  

1. Have ears for value:  
2. Understand value from the customer’s perspective (it’s not necessarily what you make or currently provide).   

This aligns with Jim Morgan’s Lean Product and Process Development (LPPD) guiding principle “Understand before you execute.” 

1. Have eyes for waste and flow 
2. Identify and map your value stream 
3. Eliminate waste to flow value and pull value 
4. Continuously improve 

If you can do these things, then you can also “Design the Value Stream”, another LPPD guiding principle. If AI can help support your value stream in your ideal future state, great! But don’t skip the crucial step of asking how AI is helping you create value in the eyes of the customer. 

One note on automating any data-intensive process: think about how you will monitor whether the process is still performing as expected. Processes are in a constant state of degradation, and you must put energy into improving them. Without some way of monitoring this largely invisible process, it will just degrade without your knowledge until it has a catastrophic failure.   

One note on automating any data-intensive process: think about how you will monitor whether the process is still performing as expected. 

Remember, it is much easier to see problems in scenarios with physical automation. Grippers wear and cycle time increases, tracks get dirty, and vehicles wander and shut down. Your AI tool will never be this transparent; it WILL give you an answer. 

So happy AI-ing! Just apply these new AI tools to actual problems in your process that are deliberately part of your value stream. As my mentor Yamada-san used to tell me,” Lean is not a hobby, it is serious business.”