Most automation roadmaps begin with the current process and ask how to run it faster. That sounds sensible. The risk is that a new cell, robot or software layer locks in whatever assumptions the old process was built on, including the ones nobody has checked in years.

I call those assumptions orthodoxies: rules a plant follows because everyone has always followed them. As I wrote in Stagnation Assassin, “Most organizations can’t smash orthodoxies because they can’t see them since these beliefs operate below conscious awareness.” Automation makes that blind spot expensive, because once an assumption is built into equipment and code, it is much harder to question.

Here are three I see most often, and a short audit to run before the purchase order goes out.

Assumption 1: “We’re out of capacity, so we need more machines”

At a shopping cart manufacturer where I led a turnaround in 2018, sales had been stuck just shy of $50 million for three years running. Over the previous four years the plant had added $8 million in automation, with little to show for it. Leadership was preparing a multimillion-dollar expansion because reports said the facility was running at 72% utilization.

I spent the next week on the shop floor, timing the work by hand. Value-adding production accounted for 31% of the time. Setup and changeover took 18%, waiting for materials 14%, and inspection and rework 9%. The machines were busy, but most of that activity wasn’t producing anything.

We canceled the expansion. Within 18 months, revenue passed $60 million without it.

Audit question: Before you automate, has anyone timed what the current equipment actually spends its hours doing?

Assumption 2: “Big batches are efficient”

Each changeover there took about 47 minutes, and the line ran six to eight of them daily. To avoid them, everything ran in large batches. That logic looked efficient on a cost sheet, but it created about 156 hours of average queue time and $8.7 million in excess inventory. The lost changeover time cost about $180,000 a year. The batching used to dodge it cost an estimated $11.3 million a year.

Applying Shigeo Shingo’s SMED (single-minute exchange of die) method, changeovers came down to 12 minutes. That let batches shrink to 10 units from 50, cut queue time to 34 hours from 156 and took $5.6 million out of inventory.

If that plant had sized new automation around 50-unit batches, it would have hard-coded the problem.

Audit question: Is your planned batch size a customer requirement, or a workaround for slow changeovers?

Assumption 3: “More inspection means more quality”

When we fixed the plant’s main assembly bottleneck, the constraint moved to a final quality check. The reflex answer is more inspectors or, on many roadmaps, an automated inspection station. We asked a different question: Why did so much need catching?

Root-cause analysis showed that 78% of what final inspection caught had been created upstream. We fixed those sources, moved checks earlier and cut final inspection back to sample checks. Defects fell 52%, and inspection time fell 67%.

Automating inspection of defects you could prevent pays to sort scrap faster.

Audit question: For each inspection point on the roadmap, what share of the defects it finds originate upstream?

How to run the audit

The audit needs no special software. Take each assumption behind the project and ask four history questions: When did this start? What was it meant to fix? Is that problem still here? What happens if we drop it? Then push the answers with a Five Whys chain until you hit “because that’s how it’s done.”

Bring in people from outside your industry, too. In one strategy session at an appliance business, we invited three outsiders. One, a logistics manager, asked why we manufactured to forecasts rather than orders. Insiders rarely ask that kind of question, because the practice feels natural.

Finally, sequence the money. At the cart plant, the standard answer to the main bottleneck was a second assembly station costing $800,000. We first rearranged the station, staged material at the point of use and synchronized upstream batches to its pace, all without capital spending. Only then did we add $87,000 of automated fastening. Throughput at that station reached 142% of the original. Following Eliyahu Goldratt’s Theory of Constraints, the automation came last and was small.

Every capital request should face a real gate. As I put it in my forthcoming book, "10-Minute Transformation," “A build gate that has never rejected anything is not a gate.”

Automation multiplies whatever process you give it. Audit the assumptions first, and you will automate the right work.