Is AI Drafting in Your Engineering Firm a Software Change or a Layoff?
AI drafting hits your repeatable layer first. Whether that becomes a software change or a layoff depends on who signs that layer's paycheck.

It is a software change if the repeatable drafting layer of your firm sits on contractors and tools. It is a layoff if that same layer sits on your payroll. The tool does not decide which one it becomes. You decide, by how you build that layer before the tool shows up.
Which part of the work will an AI drafting tool take first?
The most repeatable part of the pod, and it will go there first. I mean the drafting that follows a standard your firm already set, where nothing in the task requires a decision from a licensed person.
- Base sheets and title blocks held to a firm standard
- Layer names, line weights and sheet indexing
- Plan and profile drafting from a design someone else settled
- Incorporating redlines from a review
- Quantity takeoffs and schedule updates
- The fourth revision of a detail that has not changed in a decade
The judgment layer is what is left. Which alignment. Which of three alternatives to recommend. What the reviewer will ask for at 60 percent. Whether the site will do what the drawing says. An AI drafting tool does not decide any of that.
So the tool does not remove the work. It removes hours from one layer of the work, and that layer has an owner in your firm right now.
Why does it matter who signs the paycheck for that layer?
A contractor is a rate. An employee is a fixed cost, a person and a promise you made when you hired them.
When the hours in the repeatable layer halve, contract spend falls on its own. Payroll does not. You still owe the salary, the benefits, the desk and the software seat, and the work that used to fill the week is now done in two days by something you rent.
That gap shows up in a month when cash is tight. The fastest lever to close it is a layoff, and it gets pulled by whoever is staring at the cash flow, on a Friday, about people who did nothing wrong.
That is not a character flaw in the owner. It is arithmetic done after the fact, and after the fact is the wrong time to do it.
What does the firm look like when you split the layers that way?
The licensed judgment layer stays on payroll, and it stays small and senior. Those are the people who choose the alternative, sign the drawings and sit across from the reviewer.
The repeatable layer flexes. Contract drafters when volume demands it, software when the software earns its place, and a mix of both while a tool proves itself against real redlines.
Growth then comes from adding tools and contract capacity until judgment is the constraint, and only then adding the licensed engineer. You are not staffing a drafting floor and hoping the work stays. You are staffing judgment and renting the rest.
The constraint in this firm has never been drafting capacity. It is licensed judgment, in nine states, across 525+ projects since 2017. That is the part I cannot rent.
Where do junior engineers learn if the drafting layer is not on payroll?
The same place they should be learning now, minus the two years of redlines that used to be the entry fee.
If your training model is earn your keep drafting for two years, the tool removes the first rung of that ladder. You would have to rebuild the ladder anyway. Better to build it before you are forced to.
Put juniors with a licensed engineer on the decisions. On the review comments. On the alternatives that got rejected and why they got rejected. That is the part of the trade that transfers.
The bench you need in five years is judgment, and judgment is the one thing the tool does not supply at any price.
What if the drafting layer is already on my payroll?
Then change what you hire next, not who you let go this quarter. Nothing has arrived yet. Nobody needs to lose a job for a tool that is not in the building.
Draw the two layers on one page. Put every role and every contract on one side or the other, by name. Then stop adding payroll to the repeatable side, starting with the next volume bump.
When work rises, try the contract path before the hiring path. Let attrition do the work slowly, and tell the engineers on that side where they are headed while there is still time for them to move with you.
The promise you made when you hired them is not a rounding error, and neither is the reputation you carry into the next hiring market.
What does the swap look like when the tool actually arrives?
You change a vendor agreement, a seat count and a rate. You run the tool beside the current layer for a month, keep what survives a real redline, and drop what does not.
Nobody's income changes in a week. That is what a software change looks like. It is boring, and boring is the whole point.
The other version, where the decision about people arrives at the same time as the tool, gets made in a bad month by a person under pressure. Do not build a firm that ends up there.
What I would do next week
One page, two columns, and an honest count.
- Write judgment on the left and repeatable on the right.
- Put every role and every contract on one side, by name.
- Add up how much of the repeatable side is payroll, because that number is your exposure.
- Take the next volume bump through a contract before you take it through a hire.
- Tell the engineers on the right side what their path to judgment looks like.
- Set a date to look at the page again, and put it on the calendar.
Decide the shape of the firm before a tool forces the decision. If the repeatable layer is a contract, the tool arrives as a line item. If it is payroll, the tool arrives as a layoff. Pick which one you want to be true, then build it that way on purpose.
From the book Engineer Your Freedom
Michael Groselle, P.E., is the founder of Modern Engineering Solutions, a multi-state water and wastewater engineering firm. He left salaried engineering in 2017 and built the firm from one resignation letter and zero clients to more than $2M in annual revenue, with a team of 37 and 525+ completed projects.
Sold on Amazon. Prefer the Kindle edition? Buy it on Amazon ($9.99).