Why start here
Everyone is talking about AI right now. Most of the talk is about tools. I want to talk about people, because that is where AI wins or loses.
This is the first post in a series about what I have learned bringing AI into real businesses. It starts where my AI story started: a small computer vision startup called Drishti, and the factory floors it served.
I came in through the side door
I did not start in tech as an engineer or a salesperson. I started as a contract office assistant. I opened the office, made the coffee, booked travel, and took notes in meetings.
That side door taught me the whole business. Over the next few years I helped with hiring, legal paperwork, and grant applications.
I also went on site to set up and fix cameras on assembly lines. I checked factory video by hand to make sure the data matched the real work. Then I moved into sales and spent my days talking with car makers and their suppliers.
Drishti used cameras and AI to help factories see how hands-on work was really getting done. Apple later bought it. But the lessons that stuck with me had less to do with the technology and more to do with the people around it.
Lesson 1You can’t fix what you can’t see
Machines on a factory floor make a lot of data. People doing work by hand make almost none. So big decisions about that work were often based on a small handful of samples.
Those samples had a catch. When an engineer stands on the floor with a stopwatch, everyone does the job by the book. The engineer leaves, and the real work comes back. The numbers looked clean, but they did not tell the whole story.
I saw how hard this is up close in 2019. I joined a week-long kaizen with DENSO’s team at their plant in Battle Creek, Michigan. A kaizen is a focused push to make a process better, one small step at a time.
We got stuck on a simple question: when does a job actually start? When the worker scans the part? When they reach for it? Does it change from one worker to the next?
I also saw how much of the week went into collecting and checking data before anyone could test an idea. With good data already in hand, that time could go into making the work better.
Lesson 2The best AI makes people better at their jobs
When I talked with plant leaders, one worry stood out. A big car maker told us they did not want to add more work for the people on the line. Some plants had union crews, and trust mattered a lot.
The AI we offered was not built to replace anyone. Think of it like spell-check. Spell-check does not write for you. It catches a slip before it turns into a problem. On the line, that could mean a quick prompt when a key step was skipped.
It could also help keep know-how in the building. One supplier was about to lose a wave of senior leaders to retirement. A video record of how the best people did the job was a way to keep that knowledge from walking out the door.
Lesson 3The tech is the easy part
Once the cameras were up, a plant went from having too little data to having more than it knew what to do with.
The hard question was never “Does the AI work?” It was “Now that we can see everything, what do we change?” Who looks at the data? Which decisions does it feed? How does a supervisor’s day look different on Monday?
I believe those answers decide whether a pilot becomes part of how a team works, or stays stuck as a pilot. That is not about who is smart or who is trying hard. It is about where a team starts.
Why this matters now
The tools have changed a lot since then. The lessons have not. Whether it is a camera on an assembly line or an AI assistant in a sales team, the same three things hold true:
- See the real work before you try to change it.
- Use AI to help people, not to work around them.
- Plan how jobs and decisions will change before you buy, not after.
That is the heart of what we do at Revenue Optimization & Impact. We help teams understand where they are, measure it, and design how AI fits into the way they already work.
In the next post, I’ll share the next stop on my path and what it taught me.
Ecologist. Translator. Builder.