Bench Notes: AI Doesn’t Replace Jobs. It Replaces Steps
A builder’s note on why most AI conversations start at the wrong level and the wrong place.
The Question
Matt walked in with a big question.
“What can I do with ChatGPT?”
It sounds simple when you say it out loud, but it usually carries something much larger behind it. In this case, he wasn’t asking about small improvements or saving a few minutes here and there.
He wanted to cut his staff in half.
Six employees down to three. The shop is busy, cars keep coming in, and the work itself isn’t slowing down. What’s slowing things down is everything around the work, like calls, explanations, updates, and paperwork that stretches simple tasks into long ones.
That’s where he thinks the machine fits.
Where the Question Breaks
The question isn’t really about the tool.
It’s about the framing. Matt is asking what jobs can be replaced, but the work inside those jobs hasn’t been separated yet. It’s all grouped together under titles that sound clean from the outside but aren’t clean at all once you look closer.
A service advisor isn’t one task.
It’s a bundle of smaller actions layered together throughout the day. That includes answering customer questions, writing estimates, explaining why something costs more than expected, sending updates when parts are late, and documenting what the technician found.
When everything is bundled, the question becomes too large for the tool to answer.
Big goal, unclear work, limited experience. That’s where the mismatch starts.
Breaking the Work Into Steps
The conversation shifts the moment the work is written down.
Not job titles, but steps. What actually happens during the day starts to become visible once you list it out in plain language. Writing an estimate, explaining a repair, responding to a confused customer, updating a file, and passing information between people all appear as separate pieces.
Once those pieces are visible, a pattern begins to show.
Some steps require judgment and experience, while others repeat themselves with small variations. The repeated ones start to stand out because they follow similar structures and language every time they appear.
That’s where the machine begins to make sense.
Not at the level of replacing a person, but at the level of reducing friction inside a step. AI fits into the repeated parts of the work, not the entire role.
The Capability Gap
There’s another layer that shows up right after that.
Matt hasn’t used ChatGPT in a structured way yet. There’s no experience with prompts, no systems, and no technical background. That’s not unusual, but it changes what is realistically possible in the short term.
The same tool offers very different outcomes depending on who is using it.
Someone at the beginning level can use it to draft messages, standardize explanations, and speed up communication. Someone with more experience can build repeatable workflows that reduce variation in the work. Someone with technical support can take those workflows and turn them into more advanced tools.
All of those paths exist, but they don’t start in the same place.
The Real Paths Forward
Once the work is broken down and the starting point is clear, the path becomes easier to see.
At the simplest level, the tool helps with writing and communication. Drafting explanations, responding to common questions, and creating consistent messages are all immediate improvements that reduce time and repetition.
At the next level, those improvements become structured.
The same formats are reused, the same logic is applied, and the work becomes more predictable. The operator stops rewriting from scratch and starts refining what already works.
Beyond that, the work becomes more complex and requires additional support.
Custom tools, deeper automation, and system-level changes all become possible, but they require time, technical skill, and often other people. Not every path is necessary to make progress, and not every path is accessible at the beginning.
What This Changes
When the conversation comes back to Matt, the goal starts to look different.
Not smaller, but clearer.
Instead of replacing employees, the focus shifts toward the work itself. Which steps take the most time, which ones repeat throughout the day, and which ones create the most friction become the real questions.
Start with one.
Then test it.
Then move to the next.
Progress shows up in small changes that compound over time. The work becomes smoother, the communication becomes faster, and the need for constant repetition begins to drop.
That’s where the real change happens.
Not at the level of jobs, but at the level of steps.
Until next time—peace.
Jefferson
Chief Mess Maker, AI Sausage Factory