Labor Is 47% of a Dealership's Cost Structure. That Is Not the Same as 47% of the Problem.

On Monday, Aug. 24, Steve Greenfield, general partner at Automotive Ventures and host of CBT News' Future of Automotive, went on CBT Live and made a call that will get repeated in a lot of manager meetings this fall: within five years, the dealership BDC will look "archaic."

The reasoning is not hand-waving. Labor, citing NADA figures, is roughly 47% of a dealership's cost structure - the largest single line a dealer principal actually controls. Greenfield described a phased shift. First, AI makes existing people more productive: technicians billing more hours, salespeople closing more units per head. Then stores realize they can hold that output with fewer people and simply stop backfilling roles as staff turn over. He also noted that dealership net income before tax has sat between 1.5% and 2.5% for roughly fifty years - data he credited to Glenn Mercer and NADA - apart from a temporary pandemic spike, with last year averaging 3.3%. He expects AI's effect on labor and vendor costs to push that figure meaningfully higher.

Most of that is hard to argue with. The part worth arguing with is the order of operations.

The myth: The biggest AI payoff at a dealership is a smaller payroll. The data: In Cox Automotive's Q2 2026 Dealer Sentiment Index, dealers ranked expenses fourth among the factors holding their business back, at 33% - behind the economy at 55%, market conditions at 40% and the political climate at 36%. The same survey put the cost index at 74, its highest reading in more than a year. Source: Cox Automotive, Q2 2026 Dealer Sentiment Index, May 26, 2026.

A 47% labor share is not evidence of overstaffing

It is evidence of what kind of business a dealership is. A store sells a high-consideration, low-frequency product that almost nobody buys without talking to a person, then services that product against an appointment book that gets filled one conversation at a time. Payroll is the delivery mechanism for the revenue, not a tax sitting on top of it.

That distinction matters, because the two ways to act on a 47% number look identical on a slide and behave nothing alike in a store.

Reduce the denominator. Same output, fewer people, lower cost. The saving is immediate, precise, and lands on the next financial statement.

Raise the numerator. Same people, more conversations handled, more appointments set, more hours billed. The gain is just as real but slower to attribute, because nobody sends an invoice for the customer you never knew you lost.

Greenfield's own framing puts productivity first and headcount second, and that is the correct sequence. The practical risk is that step two is the one with a number attached, so it is the one that gets pulled forward.

What the same survey says dealers are actually short of

The Q2 2026 Cox Automotive Dealer Sentiment Index surveyed roughly 958 franchised and independent dealers between April 21 and May 4, 2026, with a margin of error of about 3 points. Index values run from 0 to 100, where 50 is the neutral line.

Read the two ends of that table together. Cost pressure at 74 confirms exactly what Greenfield is describing. Traffic at 36 and profitability at 36 say the binding constraint is demand capture, not headcount. A store that thins staffing while customer traffic sits 14 points under the neutral line is optimizing the side of the equation that is not currently binding.

Where dealers place the blame has been remarkably stable:

Expenses have barely moved in a year: 32%, then 34%, then 33%. What moved is everything on the demand side.

The volume picture says the same thing in different units. On Aug. 26, Cox Automotive forecast an August SAAR of 16.3 million, flat with July and roughly 200,000 below last year's pace, with total volume near 1.35 million units - down 8.5% year over year, largely on calendar effects. Last August carried both the run-up to the federal EV tax credit expiration and the Labor Day selling weekend, which falls in September this year. Cox senior economist Charlie Chesbrough attributed the resilience to a buyer profile skewed toward higher incomes, stronger credit and larger cash reserves, and told dealers to "stay focused during these volatile times, don't give up."

Demand is not collapsing. It is concentrating into a narrower pool of well-qualified shoppers, which raises the cost of every one a store fails to reach.

Where the archaic-BDC call is probably right

Greenfield's strongest argument is not about cost at all. It is about consistency: AI can learn a store's best practices and apply them the same way around the clock, and specialized roles such as F&I are more likely to be augmented than replaced, with less experienced staff getting real-time coaching that pulls their performance toward the store's top numbers.

That is the durable part of the thesis. A BDC's structural weakness has never been its price. It is that a BDC is a schedule. It has a Tuesday-at-10 version and a Saturday-at-4 version, a tenured version and a three-weeks-in version, and the customer gets whichever one the clock hands them. The variance is the defect, and removing variance is something software is genuinely good at.

So the BDC as a room of people manually working a call list probably does look archaic in five years. The BDC as a function - own the response, qualify it, book it, hand it off cleanly - almost certainly does not. What an AI BDC actually replaces is the schedule, not the job.

The sequencing risk that never shows up on a financial statement

Here is the failure mode worth naming, because it is quiet.

A store deploys AI, gets a genuine productivity lift, and then stops backfilling as people leave. Output holds. The expense line drops. Every reported number improves.

What does not appear anywhere on that statement is the call that rang out at 6:40 on a Thursday evening, the service customer who hung up after ninety seconds on hold, or the internet lead that got its first response Monday morning instead of Saturday afternoon. Missed demand has no ledger account. It is invisible by construction, and it stays invisible right up until market share moves.

That is why the order matters. Prove coverage first - answer rate, speed to first response, after-hours capture, and what share of answered calls actually become appointments - then let the staffing conclusion follow the evidence. Cut first and a measurable expense problem has been traded for an unmeasurable revenue problem, which is a poor trade even when the statement looks better.

It is also worth noting that cost leverage is being pursued on fronts that do not touch a customer-facing role at all. On Aug. 25, AutoTrust Dealer Alliance said in a company announcement that it had passed 300 member dealerships in under a year, a dealer-owned buying alliance built to give independent stores the vendor pricing leverage public groups get from consolidation. Vendor cost is the other half of Greenfield's thesis, and it is the half with no coverage risk attached.

What would change this read

Greenfield is describing a five-to-ten year arc, and the sentiment data here is a snapshot. Q2 CADSI fieldwork closed May 4, and conditions have moved since. If the traffic and profitability indices climb back through 50 over the next few quarters, the demand-capture argument weakens and the cost argument strengthens on its own terms.

The 47% labor share and the fifty-year margin band are also cited secondhand from a live broadcast segment, attributed to NADA and to Glenn Mercer's work. Anyone building a staffing plan on those numbers should read the underlying reports directly first.

And there is a version of this where both readings are right at once. If AI genuinely lifts throughput per person, a store can hold coverage and stop backfilling, because the same headcount handles more volume. That is the good outcome. It simply requires measuring coverage before concluding it is safe to reduce the people currently providing it.

Carbuki builds AI voice agents that answer, qualify and book for U.S. dealerships, so the coverage question can be settled with data rather than assumption. If that is the line you are working on this quarter, carbuki.com is a reasonable place to start.

Sources

• CBT News, AI could shrink the BDC and reshape F&I, Steve Greenfield says, August 24, 2026: cbtnews.com
• Cox Automotive, Dealer Sentiment Improves on Current Conditions in Q2, Outlook Weakens for Months Ahead (Q2 2026 CADSI), May 26, 2026: coxautoinc.com
• CBT News, Cox finds August new-vehicle sales remains resilient as economic pressures mount, August 26, 2026: cbtnews.com
• CBT News, AutoTrust unites over 300 franchise dealers to compete with America's largest public groups (company announcement), August 25, 2026: cbtnews.com
• NADA, NADA Data annual financial profile of America's franchised new-car dealerships: nada.org