Your AI Has a Whip Holder

82% of dealers use AI. Only 22% of users have seen it grow sales. Why bolt-on AI isn't paying, and what to do instead.

New engine. Same carriage. 1890s: cars built to look like carriages, down to the whip socket. 2026: a chatbot on a 2010 website.

In the 1890s, an engineer named Hiram Percy Maxim was building one of America's first electric cars in Hartford, Connecticut. Everyone on the project accepted one design rule: the finished machine had to look like a horse carriage.

Maxim thought that was absurd. So he suggested, sarcastically, that they add a whip socket. A holder for the buggy whip. On a vehicle with no horse.

The joke landed because it was barely a joke. Early cars borrowed almost everything from the carriage. The body. The wheels. The bench seat. Even the "dashboard," which was literally the board that stopped the horse from flinging mud in your face. We still use the word.

It took the industry years to accept the obvious. No horse means a different machine.

Automotive retail just bought a warehouse full of whip holders. We call it AI.

The receipts

Don't take my word for it. Take the word of a company that sells AI to dealers.

Cox Automotive's AI in Auto Retail Tracker for Q2 2026 found that 82% of dealers now use AI. 69% expect it to grow sales. Only 22% of the dealers using it have actually seen that growth.

The AI Return Gap: 82% of dealers use AI, 69% expect it to grow sales, 22% of AI users have seen that growth. Source: Cox Automotive AI in Auto Retail Tracker, Q2 2026.

Four out of five dealers bought AI. Roughly one in five users got paid. And about a third of dealers either aren't measuring AI's impact or can't say how they're measuring it.

When the seller's own research says the product isn't paying off, you don't have an adoption problem. Adoption is done. You have a whip holder problem.

The algebra of Whip Holder AI

My thesis in one line:

Old process + AI = old process with a bigger invoice.

Most of what gets sold as dealership AI is a feature bolted onto a workflow designed before the iPhone. A chatbot on a 2010 website. AI-written notes in a CRM holding five copies of the same customer. A summarize button on a screen nobody has redesigned in fifteen years. It's a mullet: business in the front, 2010 in the back.

I've sat with more than 200 dealership groups in the last year. Nearly every one tells me, "We're doing AI." So I ask one question.

What work stopped?

Silence. Every time.

I'm not throwing stones. I've spent 36 years in retail and technology. I've bought my share of whip holders. I've sold a few, too.

The AI Staircase

Map AI against every job in a franchised dealership and it lands on one of three steps.

The AI Staircase: Step 1 Bolt On, same process a little faster. Step 2 Rewired, the work is redesigned. Step 3 AI Native, built from the customer record up.

  • Step 1. Bolt On. Same process, a little faster. AI writes the note, and nothing else moves. This is where most of that 82% lives.
  • Step 2. Rewired. The work is redesigned. Steps disappear. AI prepares and checks, and people handle the exceptions.
  • Step 3. AI Native. Built from the customer record up. AI runs the routine. People decide, build relationships and approve.

Step one to step three isn't an upgrade. The dealer on step three isn't running your playbook faster. They're running a different cost structure. In a minute I'll give you the math to prove it on your own service drive.

AI × garbage = garbage at scale

AI is a multiplier. Multiply clean data and you get an edge. Multiply garbage and you get more garbage, faster, with better grammar.

Your DMS knows John Smith. Your CRM knows Jon Smith. Service has J. Smith with a different phone number. F&I has John Smith Jr., who is actually John Smith. Marketing has an email address and a prayer.

The Five Copies Problem: five fragmented customer records across DMS, CRM, service, F&I and marketing, versus one customer record every department reads.

Bolt AI onto that and you get five confident, wrong answers. The chatbot offers John a trade-in on the car he traded last spring. The service reminder goes to the number he cancelled. The equity alert fires on the wrong vehicle.

AI doesn't fix bad data. It scales it.

Gone beats faster

Take a 7:30 a.m. service appointment on a Tuesday.

Gone Beats Faster: in the Bolt On version AI touches one of seven service lookup steps. In the Rewired version all seven are prepared by AI before the customer arrives.

  • The whip holder version. The customer pulls into the drive. The advisor looks them up in the DMS, checks recalls, digs through prior ROs, hunts for declined work, checks warranty and walks back to ask the shop foreman about a comeback. Then AI writes a lovely summary of the visit. Seven steps. AI touched one.
  • The rewired version. Before the customer leaves the house, AI has assembled the visit: history, recalls, declined work, warranty, equity position and likely questions. The advisor walks out with everything in hand and spends the time on the customer.

The lookups didn't get faster. They're gone.

Now do the math on your own drive. I'm not going to hand you an industry average, because yours is the only number that matters.

Run it on your drive: Minutes of lookup per RO × ROs per day × 22 working days ÷ 60 = advisor hours per month spent on work a rewired dealership doesn't do.

Then ask the follow-up question. When service volume grows, does that number force you to hire another advisor? The rewired dealer down the street doesn't have to. That's what I mean by a different cost structure, and you cannot out-hustle a cost structure.

The same exercise works on your BDC, the used car manager's appraisal and the F&I menu. Run it and hit reply with your number. I want to see the spread.

Your customer already left the carriage

While dealers debate chatbots, the customer moved. In the same Cox tracker, 63% of shoppers say they'll use AI on their next vehicle purchase, but only 29% of dealers have started adjusting to AI-powered search. The share of dealers who know they need to adjust but haven't started rose from 26% to 32% in a single quarter.

The Customer Gap: 63% of shoppers will use AI on their next vehicle purchase; 29% of dealers have started adjusting to AI-powered search. Source: Cox Automotive AI in Auto Retail Tracker, Q2 2026.

The customer is arriving researched, priced and halfway decided, and walking into a dealership running a whip holder.

The gap is the opportunity. Or the obituary.

Owned vs. rented intelligence

Every strange decision in business makes sense once you find out who it's profitable for.

The platforms running most dealerships were architected before the iPhone. Rebuilding them from the customer record up would mean admitting the core product is the problem. No incumbent does that voluntarily. Not because they're evil. Because they're rational. So they bolt AI onto the carriage, put it in a press release and add a line to your monthly invoice.

Look at what that line item buys. Each bolt-on sees one department's slice. Each is fed by an export from a system you don't control. Each keeps what it learns, and takes that learning with it when you cancel.

That's rented intelligence. You're paying rent on what your own customers taught the machine.

The alternative is owned intelligence: one customer, one record, every touch, in a store you control. Sales, service, parts, F&I and marketing all reading from the same truth. Once that exists, AI can run the routine end to end, and your people stop feeding screens and start doing what machines can't: judgment, relationships and approval.

The companies that won the last century didn't build a better carriage. They started from the engine. In a dealership, the engine is the customer record, and the question of who holds title to it doesn't belong in IT. It belongs in the dealer principal's office.

The Whip Holder Test

Start with the Swap Test. If you cancelled every AI product in your store tomorrow, what would Monday look like?

For most dealers the honest answer is: exactly the same, a little slower. Nothing structural would have to be rebuilt, because nothing structural was ever built. That's how you know you bought whip holders.

Then, before the next demo, ask four questions. In writing.

The Whip Holder Test: four questions to ask before any AI demo.

  1. Can it read across departments, or only within one? One department is one slice of the truth.
  2. Does it write back what it learns into a store you control? If not, you're paying to improve someone else's asset. It's the most valuable term in any dealer software agreement, and almost nobody negotiates it.
  3. If we cancel in 18 months, what do we keep? An export is not understanding.
  4. Does it take an action, or hand a person a list? A list is a whip holder's signature. Price it like one.

These come from my whitepaper, The Dealership Nucleus. Screenshot them. Bring them to your next vendor meeting, and then to your own manager meeting.

Four predictions I'll be held to

I don't forecast past 18 months. Anyone who claims to see further in AI is selling something. So here are four calls for the window between now and spring 2028, each specific enough to be wrong.

  1. A top-25 dealer group drops a DMS or CRM vendor and names data access as the reason, on the record. Today that complaint lives in private. It goes public.
  2. Write-back becomes a standard ask in large-group software renewals. "Do we keep what your AI learns?" moves from whitepapers into procurement checklists.
  3. A top-150 group puts a data custodian on the org chart, reporting to the dealer and sitting level with the controller. One person whose job is the store's record of itself.
  4. "AI" stops being the headline on vendor booths at NADA 2028. When everyone has it, nobody can sell it. The pitch shifts to which steps disappear and who owns the data.

Bookmark this. I'll score myself in public.

The sort

Nobody remembers who built the best horseless carriage.

The next 18 months will sort dealerships into two groups: dealers with a lot of AI tools, and dealers with a different business. The first group will have much nicer dashboards. Literally. The second group will own their customer, run leaner and buy the first group's rooftops.

Rethink the work, not just the tool.

So be honest. What's in your dealership right now: a whip holder or an engine? Run the service drive math and hit reply with your number.

Frequently asked questions

What is Whip Holder AI?

A feature that makes an old dealership process look modern without changing it, like AI-written notes in a CRM that still holds five copies of the same customer.

How many dealers use AI, and how many see sales growth?

Cox Automotive's AI in Auto Retail Tracker for Q2 2026 found 82% of dealers use AI, 69% expect it to grow sales, and only 22% of dealers using it have seen that growth.

What is the AI Staircase?

Three steps: Bolt On (same process, a little faster), Rewired (the work is redesigned and steps disappear), and AI Native (built from the customer record up, with AI running the routine).

How do I measure the cost of bolt-on AI in my service drive?

Minutes of lookup per RO times ROs per day times 22 working days, divided by 60, equals advisor hours per month spent on work a rewired dealership does not do.

What should I ask before an AI vendor demo?

Can it read across departments? Does it write back what it learns into a store you control? If we cancel in 18 months, what do we keep? Does it take an action, or hand a person a list?