Interactive Research · Retail Automotive · September 2026

Every Job in the Dealership.

Everybody tells you AI changes the dealership. Nobody will name a job. So here are all twenty.

Todd Smith

CEO of QoreAI · Author of The Intelligent Dealership

September 2026 · 20 min read

The unit of change

AI does not take a job. It changes the tasks inside it.

5 departments

20 positions

66 average employees

Section one

The question nobody will answer

You know which advisor is carrying the drive and which one is coasting. You know the title clerk has been there nineteen years. You know what every one of those people costs you to the dollar.

Two years of conferences, vendor decks and LinkedIn posts, and not one of them has named a seat in your building. Not which one. Not when. Not what you are supposed to do with the person sitting in it.

“Walk into any rooftop at eight in the morning and you find the same meeting.”

The used car manager knows what aged past sixty. The service manager knows who is short on hours. The desk knows what died over the weekend. For forty minutes everybody compresses what they know into a summary and hands it to the room.

You have sat in that meeting a thousand times. It is not a management practice. It is a data pipeline made of people. Every one of your jobs got built around the fact that the store cannot remember anything on its own.

Nobody in your building has a job AI takes or does not take. Your advisor does fourteen things between seven in the morning and close. AI shows up at three of them this year and two more next year. Same advisor, same seat, different day. Nobody got replaced and nothing stayed the same.

That is the only honest way to look at this. Go seat by seat, task by task, and ask what happens to each one.

Section two

Three constraints that only apply here

Every article you have read about AI and jobs was written about somebody else’s business. Law firms. Banks. Insurance. None of them run a franchised rooftop.

Your store cannot remember anything. Almost nothing you know is written down where a machine can read it. It lives in the eight o'clock meeting, in a manager's head, and in four systems that key the same customer four different ways.
You do not own the tools you run on. Cancel your biggest vendor tomorrow and ask what you still have. If the answer is nothing, you rented it, and the vendor decides what it does next year.
The factory already wrote your org chart. You have five departments because the statement has five lines, and every pay plan pays against them. You cannot redraw that on a whiteboard. It is why the chart is the last thing to move.

Now go pull your statement

Look for the two departments that are not on it. There is no BDC line on a factory statement. There is no F&I department either, because that income reports inside new and used vehicle.

Two of the three most exposed groups in your building are the two you have never managed as a department. Nobody watches that payroll because there is no line to watch it on.

Section three

It does not arrive as a decision. It arrives as a Tuesday.

Nobody calls a meeting. Nobody writes a memo. Here is the order it happens in, and you have probably already done the first one.

01
Somebody sells you a tool for the BDC. It answers leads at ten at night. It works well enough that you stop checking. Six months later a rep quits and you do not backfill her, because the numbers held.
02
The phones go next. Calls get answered and routed without anybody picking up. The switchboard opens up and stays open.
03
The office gets quiet. Deals post faster. Schedules reconcile themselves. Your controller stops asking for help in December.
04
Your service manager asks for another tech. You approve it. The shop is full and hours are up. First time AI added a seat in your building and nobody called it that.
05
Two years in, you have fewer people than you started with. You never made a decision. You just stopped backfilling.

“That is the default path, and plenty of rooftops will come out of it fine. What you gave up was the say.”

The people you lost were whoever happened to quit, not whoever you would have picked to move. The seats that closed were whichever ones your vendors automated first, in the order they happened to sell you something.

Everything below is the same story with you making the calls.

Interactive research

What happens inside every dealership job?

Start with one average rooftop. Five departments. Twenty positions. Sixty six whole people. Then change the future and watch the same building reorganize around the work.

One average rooftop

New vehicle14

Used vehicle7

Service29

Parts6

Back office10

Every mark is one person. No fractions, no abstract labor units.

The average franchised rooftop, until you tell it otherwise

New vehicle14people

Used vehicle7people

Service29people

Parts6people

Back office10people

Total66people

Not 66? Put your own number in.

Everyone on your payroll, all five departments, full and part time.

Your employees

Defaults to the NADA average: 66 employees.

14new vehicle7used vehicle29service6parts10back office

The industry average, until you put your own number in above.

Where 66 comes from

NADA Data 2025: 1,123,100 employees across 16,990 franchised light vehicle dealerships, which averages 66.1 per rooftop. This model uses 66 whole employees, split by NADA's published position mix. Technicians are 25.2% of it, the largest group in the building. The money is built from the same source at standard operating ratios, which puts payroll at $97,036 per employee against the $98,644 NADA reports for average annual earnings. It ties to their number without being fitted to it.

Your store does not look exactly like this. Mine never did either.

Compare one position from each department

Chapter 02 · One seat you know

Your service advisor's day

Fourteen things between seven in the morning and close. AI does not take the job. It changes what happens to the work inside it.

0114 existing tasks

02Four outcomes

032 new responsibilities

Explore all twenty positions

Before the shift

One seat. 14 existing tasks.

01Greet on the drive
02Walk around the vehicle with the customer
03Calm down an upset customer
04Explain the work at delivery
05Recommend from the multi point inspection
06Quote a repair
07Write the warranty narrative
08Plan the shop's capacity for the day
09Send status updates
10Chase declined service
11Book the next appointment
12Order parts for the repair order
13Submit the warranty claim
14Build the estimate

Same position, same seat, different day. The unit of change is the task, not the job title.

Step 02 · Make five calls

First, choose the store that sounds like yours

The difference is not how smart the models get. It is whether you own the record and rewrite the pay plans. Pick the future that sounds like your store.

Chat on the website, an AI receptionist, a phone tool in service. Each one works a little. None read across the five departments, none write back into anything the dealer owns, and the pay plans never change. Most franchised rooftops are here right now and a lot of them will stay here.

+3%dealership gross

66people, from 66

0whole seats close

0whole seats open

Step 02 · Your five calls

Build the version you would actually run

None of this is about how smart the software gets. It comes down to five decisions, and every one of them is yours. Answer them the way you would actually answer them, not the way you wish you would.

Starting point: Bolt on

Decision 1 of 5

How far out are we looking?

This controls how quickly the tools arrive.

These are dates for what a franchised rooftop can put into production, not what the labs can do. The frontier is ahead of all three and has been the whole time. METR's January 2026 work puts the length of task an agent finishes doubling every four to six months, and the length it finishes reliably runs about a quarter of the headline number. Notice how little separates 2028 from 2030. Capability stops being your constraint well before the other four questions do.

Pay plans are holding adoption at 28%, not the 43% you set. Compensation still pays for the activity the AI does. Raise pay plan alignment or nothing above it moves.

Step 03 · See your dealership

Start with the people.

Every square is one employee. First see which seats stay, move, open, or close. Then inspect the work and money if you want the deeper view.

White seats remain. Coral seats close. Teal seats open.

Your modeled rooftop

Every square is one whole person.

StaysMovesOpensCloses

New vehicle

14 people

Used vehicle

7 people

Service

29 people

Parts

6 people

Back office

10 people

Your modeled statement

$8,549 in modeled annual net change

Here is the bridge from today's rooftop to your result. Gross gained comes first, then payroll changes, then the technology bill. The last line is what remains.

How the result reaches the bottom line

$1,912,889Starting net

+$119,334Gross gained

$-40,810Payroll change

$-63,000Technology

+$8,549Modeled net

Annual, one rooftopTodayYour modelChange

Total sales$76,515,541$77,104,387+$588,846

Total gross profit$10,329,598$10,448,932+$119,334

Personnel expense$6,404,351$6,445,161+$40,810

All other expense$2,012,359$2,019,333+$6,974

Added technology and data0$63,000+$63,000

Net pretax$1,912,889$1,921,438+$8,549

Net to sales2.5%2.5%-0.0 pts

Net to gross18.5%18.4%-0.1 pts

Absorption67%68%+0.4 pts

Net moves $8,549. You spent $63,000 and bought yourself almost nothing, which is what happens when the tools change and the pay plans do not.

No technicians added at these answers, so none of the gross above is shop throughput. All of it is pricing, product mix and follow up.

Send this to your GM

The link carries your store size and all five answers, so whoever opens it lands on exactly what you are looking at right now.

The average 66 person rooftop, run through the workforce model.

How it is set:
  How far out are we looking: Today, 2026
  Who owns your store's data: Our vendors do
  Do you rewrite the pay plans: No
  How much runs without a manager approving it: Approve everything
  When a seat loses its work, what do you do: Wait it out

What it says:
  66 people today, 66 after. 0 seats close and 0 open.
  Net to sales 2.5% to 2.5%. Absorption 67% to 68%.
  +$8,549 to the bottom line.
  Technicians +3% on pay, and we are 0.5 short of the work we could sell.
  Pay plans are holding adoption at 28% of the 43% available.

Run your own: https://toddsmith.ai/whitepapers/every-job-in-the-dealership#d=00001

Total sales is NADA Data 2025 across 16,990 rooftops. Total gross at 13.5% of sales and personnel at 62% of gross are standard operating ratios, not my invention. Run them and payroll comes out at $97,036 per employee, against the $98,644 NADA reports for average annual earnings. The structure checks against their number without being fitted to it. Your store is not the average, so put your own statement next to this one.

Finding 1

Where the gross actually comes from

Gross only moves two ways in your building. You sell more hours in the shop, or you make more per unit out front. Taking cost out of the back office does neither. That money shows up on the expense line, not here, and if somebody shows you a model that counts it in both places, walk away from it.

+1.2%

total dealership gross versus the same rooftop without AI

New vehicle+0.3

Used vehicle+0.1

Service+0.7

Parts+0.1

Back office+0.0

Points of dealership gross contributed by each of the five departments.

Service runs away with this, and most of it is not new hires. It is the technicians you already have turning more hours because they stopped waiting on a part, an approval or a dispatch decision. A technician carries about $212,000 of gross against $107,000 of cost, so their idle time is the most expensive thing in your building. Used vehicle is pricing, not headcount. Back office sits at zero on purpose. It has never produced a dollar of gross and it never will. What it produces is expense, and that shows up further down.

What your answers imply

Your building moves from 66 to 66 people, opens 0 whole seats, closes 0, and adds $8,549 to annual net.

The model does not predict who leaves. It shows where work moves. Your redeployment decision determines whether the people move with it.

Finding 2

The building does not shrink much. It reshapes.

people, against 66 today. A change of 0 seats.

New vehicle 21%11%Service 44%9%Back office 15%

Share of the building by department. Service goes from 44% to 44%. Back office goes from 15% to 15%.

Accounting clerks4

Warranty administration1

Reception and cashier2

Controller and office manager1

Title and registration1

BDC and internet3

Marketing1

Used vehicle manager1

Sales desk manager2

Dealer principal and general manager1

F&I3

New vehicle sales consultants6

Used vehicle sales consultants6

Service manager1

Dispatcher1

Parts manager1

Service advisors5

Parts counter and wholesale5

Technicians17

Porters, detail and lot4

Whole seats by position. The line is today's count.

0 whole seats close and 0 open. At your redeployment setting, 0 of the 0 people move into one of the seats that opened and 0 leave the payroll. Nothing in the model decides that split. The dealer does.

A seat losing its work is not the same as a person leaving your building. Move them into the department that is growing and you keep the person, the tenure, and everything they know about your customers. Do nothing and they walk out with all of it.

Finding 3

Saving half a person saves you nothing

This is where dealers get fooled. A vendor shows you a study saying their tool gives an advisor back four hours a week. Real number, real study. You have five advisors, so that is twenty hours, which is half a person. Then payroll runs and it costs exactly what it cost last month.

0whole seats that close

0whole seats that open

66people in the building

0.5 people worth of work disappears without closing a seat. It is spread in tenths across the positions you kept, so it never shows up on a headcount report and never shows up on the statement.

That is the quiet version of this whole thing, and it is the version most rooftops are going to live. Work leaves in tenths of a person, spread across twenty positions, and not one of them loses a seat. Your store gets measurably faster and costs exactly the same, because nobody ever put the pieces together into a decision.

Where these people go

Do not read this as a list of names to cut. Read it as who you move, and be honest about the ones you cannot. A closing seat and an opening seat are not the same thing. Some of these moves take a week. One of them takes four years.

Seats that close

None at these settings. Nothing is being cut.

Seats that open

None at these settings. There is nowhere internal for anybody to go.

Finding 4

Your techs are about to get expensive

They are already a quarter of your building and the hardest quarter to replace. Their work is the least automatable in the store and the easiest to sell more of. Demand goes up, supply does not move, and you already know what happens next because you have lived it every time the tech market got tight. The screen heavy seats go the other direction.

Technicians+3%

Service advisors+1%

Used vehicle manager+1%

Service manager+1%

Porters, detail and lot0%

Used vehicle sales consultants0%

New vehicle sales consultants0%

Dealer principal and general manager0%

Parts manager0%

F&I0%

Sales desk manager0%

Parts counter and wholesale-1%

Controller and office manager-1%

Title and registration-1%

Dispatcher-1%

Marketing-2%

Accounting clerks-2%

Reception and cashier-2%

BDC and internet-2%

Warranty administration-2%

Percent change in pay per person versus the same rooftop without AI.

Pay plans are the whole game. Nobody in your building uses a tool that lowers their check, and neither would you. If the plan still pays on the task the system now does, the tool sits there unopened and you paid for nothing. That is why the pay plan dial caps adoption instead of sitting next to it. You can buy every tool on the market and change nothing until you fix the plans.

Finding 5

You decide who keeps the money

When a person produces more, they get paid more. That is how every pay plan in your building is written. When a system produces more, nobody gets paid anything and the money just stays. More of your work runs on systems every year, so more of every extra dollar stops at the bottom line whether you planned it or not.

To people 61%To the dealer 39%

Of every dollar of dealership gross, the share paid out versus the share kept. Today it is about 62 cents to people. Here it is 61 cents.

Decide this one on purpose. Keep the whole thing, or run a smaller team, pay them better than anybody in your market, and stop losing the good ones. Both work. Backing into the first one without thinking about it is how your best advisor ends up at the store down the road for four hundred more a week.

Do one thing with this before you close it.

Put your own headcount in at the top, set the five answers the way your rooftop actually runs, and send it to whoever is going to have to live with the result. It takes about ninety seconds and the link carries everything.

Section four

Five questions about your store

Everything the model just showed you assumes you can act on your own record. If your record lives inside somebody else’s product, your vendor decides what these jobs become, not you. Answer honestly. You are the only one reading.

01
Can you pull sales, service and inventory into one view without exporting anything?
Not a report somebody builds you. A view you can ask a question of.
02
If you cancelled your biggest vendor tomorrow, what would you still have?
If it is a final export and a login that stops working, you never owned it.
03
Is one customer one record, or four?
Bought in 2019, services twice a year, just filled out a trade form. One customer?
04
Does your best tool take action, or hand somebody a list?
A list is work moved, not work removed. Lists have never changed a headcount.
05
Does any pay plan in your building still pay for work a system already does?
Go look. This is the one dealers are most surprised by.

Count your yes answers. Under three and everything above is theory for your rooftop, no matter how good the models get.

Section five

Fix the pay plans before you touch the org chart

The order matters more than the speed. Every step has a stop condition, so you know when you are done instead of guessing.

01
Own the record. Done when you can ask a question across all five departments without exporting anything.
02
Build the store model. Done when the system knows your capacity and inventory position without being told every morning.
03
Build the customer model. Done when one customer is one record instead of four that never met each other.
04
Open it up to your people. Done when a manager can ask for something and get an answer instead of a report.
05
Run one workflow end to end. Done when it runs start to finish and a person still approves it. One workflow, not eleven.
06
Rewrite the pay plans. Nothing above this line sticks until this is finished. This is the step everybody skips and the only one that costs nothing.
07
Move the org chart last. No stop condition. By the time you get here the chart is writing down what already happened.

“Own the record. Build the model. Fix the plans. Then move the chart.”

Section six

Where I could be wrong

No hardware. If service robotics or automated recon show up, the technician line changes and the whole picture goes with it.
No factory behavior, and you know as well as I do that the factory can override anything you decide.
No competitive response, so if every rooftop in your market does this, the gain compresses into price and nobody keeps it.
No body shop. No used vehicle cycle. No floorplan.

Regulation is in here, but at one point only. Credit is a consequential decision, so F&I is capped no matter where you set autonomy. What is not modelled is everything else the statutes might reach.

The bigger one is that every position is one average person. Your top advisor does three times what your bottom one does, and AI does not land on those two the same way.

This is a snapshot, not a decade. It shows what the building looks like at a given level of capability and a given set of decisions. It does not model the path between now and then. For the ten year arc, read Auto Retail 2035.

Methodology and sources

Headcount and the position mix are NADA Data 2025: 1,123,100 employees across 16,990 franchised light vehicle dealerships. That source average is 66.1 per rooftop. The model rounds it to 66 whole employees, split by NADA’s published position mix, with technicians at 25.2%.

Financials use the same source at standard operating ratios, 13.5% total gross to sales and personnel at 62% of gross, which produces $97,036 of payroll per employee against the $98,644 NADA reports for average annual earnings. Fixed operations gross comes from $164B of service and parts across those rooftops at a 44% margin, putting one technician at $212,273 of gross against $107,613 of cost.

How each position’s work splits and how exposed it is are my estimates from running stores and from what QoreAI sees in dealer data, not survey results. Structure adapted from the Anthropic Economics team’s scenario explorer.

Free Resource

Get the PDF: Every Job in the Dealership

The full paper, all twenty positions and the department tables, as a 23 page print-ready PDF. Enter your details and it unlocks immediately.

We respect your privacy. Unsubscribe anytime.