The closing slide of this NADA Show 2026 session is a promise, set in large green highlight:
Dealers who act now will be 2X more efficient, and nearly 2X more profitable. (if you don’t, others will)
It is presented by Jessica Stafford and Derek Hansen, both Senior Vice Presidents at Cox Automotive, and that matters more here than in any other session we have covered. Cox is not a vendor with a product and a slide deck. Cox owns Autotrader, Kelley Blue Book, Manheim, vAuto and Dealertrack, which means it is simultaneously the largest data provider in American automotive retail and the supplier of the tools the session recommends. When a Cox study is quoted at this show, Cox is quoting itself.
That is not a criticism yet. It does mean this deck is the one where checking the numbers against the published research is most worth doing, because unlike most conference decks, the research actually exists and is public. So we checked all four headline figures. One holds up completely, two are real numbers pointed at the wrong question, and the promise above is not a finding at all.
The four numbers, checked
The “Future of Car Buying” slide carries three percentages side by side, and the close carries the fourth.
83 per cent say AI will impact how they buy a car: this one is sound
This traces cleanly to Cox Automotive’s own Car Buyer Journey Study, the sixteenth annual edition, published on 13 January 2026 on fieldwork conducted in the autumn of 2025 with 2,300 people who had bought a new or used vehicle in the previous twelve months. The study reports that 83 per cent believe AI will reshape car buying within ten years.
Named study, disclosed sample, recent fieldwork, and the slide’s wording matches what was measured. After a series in which we traced conference statistics back to retracted 2014 vendor press releases, it is worth stopping to say that this is how it should be done. The only caveat worth carrying is that the sample is recent car buyers rather than the general public, so it describes people already in the market.
64 per cent have experienced using AI: real number, wrong question
This is where the slide starts doing work the research does not support. Sitting between two car-buying statistics, “64% of consumers have experienced using AI” reads as though two thirds of your shoppers are arriving having used AI to shop.
They are not. Cox’s own Car Buyer Journey Study, in its first year of tracking the question, found that 19 per cent of all buyers and 25 per cent of new-vehicle buyers used an AI website such as ChatGPT or an AI-generated overview such as Google’s while shopping for their vehicle.
Sixty four per cent have used AI for something. Nineteen per cent used it to buy a car. Those are different facts, they differ by more than three times, and Cox published the smaller and more relevant one itself. Placing the general-experience figure next to car-buying figures makes shopper AI adoption look roughly triple what the presenter’s own company measured.
37 per cent of dealers think AI matters: hard to square with Cox’s own dealer research
The third figure exists to create a gap. Consumers are ready, dealers are not, therefore act now.
But Cox’s AI Readiness Study, based on 537 franchise dealership leaders across focus groups, interviews and a survey run between April and August 2025, reports that 81 per cent of dealers believe AI is here to stay and 63 per cent say investing in AI now is critical to long-term business success. Sixty three is not thirty seven. The two numbers are answers to differently worded questions, so this is not a flat contradiction, but the slide picks the framing that makes dealers look least engaged, and it does so without showing the alternative from the same company’s research published months earlier. The size of the gap on that slide is the entire argument for urgency, and the gap depends on which dealer number you choose.
The 2X promise: not a finding
This is the important one, and the problem is not the arithmetic. It is the tense.
Cox’s Digitization of Automotive Retail study, released in June 2025, reports that high-performing dealers using AI and offering key online retail steps show significantly higher close rates and profit margins. It does not publish a multiple, and we could not locate a published Cox figure stating that adopters are twice as efficient and nearly twice as profitable.
Even granting the number, look at what it would be describing. A comparison between dealers who have already adopted these tools and dealers who have not is a comparison between two groups that differ in many ways at once. Well-run dealerships with capital, process discipline and management bandwidth adopt new systems earlier, and they were more profitable before they adopted anything. The slide takes an observed difference between two populations and converts it into a forecast about what will happen to you if you buy in. That is a selection effect dressed as a causal promise, and the parenthetical “if you don’t, others will” is doing the rest of the persuading.
None of which means the tools do not work. It means the honest version of that slide reads “dealers who have adopted these tools are, on average, more efficient and more profitable”, and that sentence sells considerably fewer subscriptions.
What the session gets genuinely right
The framework underneath the numbers is better than the numbers. Digital leadership is defined in four words, Educate, Adopt, Act and Clean Data, and the last of those is the one most vendors leave out of the pitch. It then splits into three pillars.
Inventory intelligence is the strongest section, and its argument is structural rather than technological: acquisition, appraising and pricing should be one unified strategy across the inventory lifecycle rather than three activities owned by three people who do not share a number. Under it, acquisition splits new from used, with new being “stock based on data, do not stock too many or too few” and used being multi-channel sourcing. Appraising is reduced to three words worth putting on a wall: right money, right process, right recon. Pricing is market driven, dynamic and ROI focused.
That framing costs nothing to adopt and does not require anybody’s software. A dealer principal who simply gets the used car buyer, the appraiser and the pricing manager into the same weekly meeting with the same numbers has implemented most of it.
Merchandising precision is the most immediately useful. The demonstration shows AI generating feature descriptions from a vehicle’s specification, turning “surround-view camera” into a sentence about parking tight spots feeling effortless with a bird’s eye view. Whatever you think of the prose, the operational point is real: most dealer listings describe features as a list of nouns, and the work of turning nouns into reasons to care has historically not been done because nobody had time to do it 400 times. That is a genuine and unglamorous use of the technology.
AI-powered engagement is illustrated with a customer journey, and it is the slide that should make a South African reader reach for the compliance file. We will come to that.
The closing framework, “play the lifetime value long game”, loops initial purchase to ongoing service to future trade-in to additional vehicle purchases. It is not novel, but it is the correct answer to the affordability pressure every dealer is under, and it pairs with our piece on connecting customer data to service revenue.
The part that does not travel
This section is ours, and it is the largest gap in the session for a South African audience.
The three pillars quietly assume an infrastructure that does not exist here in the same form. Intelligent appraising and market-driven pricing in the American context mean Kelley Blue Book and Manheim auction data flowing into vAuto, all of which are Cox properties and none of which serve this market. A South African dealer’s equivalents are a different set entirely: TransUnion and Lightstone for vehicle valuation data, Cars.co.za, AutoTrader South Africa and WeBuyCars for market and listing signal, and the banks’ own systems on the finance side.
This matters practically. If you take the framework and go looking for the tools, you will find that the integrated single-vendor stack the session implicitly describes is not available to you, and that assembling the equivalent locally means stitching several providers together yourself. That is not a reason to ignore the framework. It is a reason to treat the session as an argument about how to organise your business rather than a shopping list, because the shopping list is for a different country.
The customer journey example needs the same translation. On the slide, Joe searched 2022 Jeeps on a third party site five times, favourited a blue 2022 Jeep Wrangler, indicated a 2017 Toyota Camry to trade, wants a payment of 500 dollars a month, and test drove the Wrangler. Converted at 16.21 rand to the dollar, the mid-market rate on 17 August 2026, that payment target is about R8,105 a month. In local terms the same journey is somebody who looked at Rangers or Fortuners on Cars.co.za, has a 2017 Corolla to trade and has a monthly instalment in mind, which is the shape of nearly every enquiry a South African dealer receives.
That payment target is also the only monetary figure in the entire deck, which is worth noting given how much of the session is about profitability.
POPIA, and why that journey slide is a compliance question
Also ours. Look again at what the connected experience slide is actually describing. A named individual’s searches on a third party website, the specific vehicle he favourited, the car he owns, the monthly payment he can afford, and the fact that he came in for a test drive, all joined into a single profile that follows him online, in store, and online again.
Under the Protection of Personal Information Act, most of that is personal information and some of it is close to financial information. Three obligations follow, and none of them are optional because an American slide did not mention them:
- The joining is the regulated act. Combining third party browsing behaviour with in-store activity and a named customer record is exactly the cross-context profiling POPIA is concerned with. You need a lawful basis for the combination itself, not merely for each source separately, and the customer needs to have been told it will happen in terms they would recognise.
- Third party data has to arrive lawfully. If behavioural data reaches you from a listings portal or an ad platform, your privacy notice and your contract with that provider both have to account for it. “The vendor gave it to us” is not a basis.
- The payment figure carries extra weight. A stated affordability target sits close to financial information, and using it to shape what you offer moves you toward the National Credit Act’s affordability assessment territory. Recording that a customer “wants R8,105 a month” and pricing to it is not the same as establishing what they can actually afford, and the Act cares about the difference.
None of this is legal advice, and the sequence matters: get the consent architecture right before you build the profile, because retrofitting a lawful basis onto data you have already joined is considerably harder than collecting it properly.
How to use this session
- Take the unified inventory lifecycle argument seriously. Acquisition, appraisal and pricing answering to one strategy is the highest value idea in the deck and it needs no purchase order.
- Start with clean data, which the session names and then moves past quickly. Every pillar above degrades to guesswork without it, and it is the least glamorous and most commonly skipped step.
- Use AI for merchandising copy now. It is low risk, immediately measurable, and the work genuinely was not being done before.
- When a vendor quotes you a multiple, ask whether it is a measured difference between existing groups or a prediction about you. Ask what the adopters looked like before they adopted. Our piece on measuring your own numbers is the antidote to taking anyone’s word for it.
- Assume roughly one in five of your shoppers has used AI to shop, not two in three, and revisit that assumption every six months because it is the fastest moving number here.
The frustrating thing about this session is that it did not need the overreach. Cox has better research than almost anyone in the industry, the 83 per cent figure is properly sourced and genuinely striking, and the inventory framework stands on its own merits. The 2X promise adds nothing except a reason to check everything else.
Source
- Master Digital Leadership: Defeat Disruption, Drive Value, presented by Jessica Stafford, SVP Consumer Solutions, and Derek Hansen, SVP Dealer, Lender and Inventory Management Solutions, both of Cox Automotive, at NADA Show 2026, Las Vegas, 3 to 6 February 2026. Reviewed from the 25 page handout and the full 29 minute recording.
- The recording carries no audio track, so the slides were read in full but the spoken commentary was not available to us. Everything attributed to the presenters is text shown on a slide.
- Primary sources checked: the 2025 Cox Automotive Car Buyer Journey Study, sixteenth annual edition, published 13 January 2026 on autumn 2025 fieldwork with 2,300 recent vehicle buyers, which is the source of the 83 per cent figure and of the 19 and 25 per cent AI shopping figures; the Cox Automotive AI Readiness Study, 537 franchise dealership leaders, April to August 2025, which reports the 81 and 63 per cent dealer figures; and the 2025 Digitization of Automotive Retail study, June 2025, which describes higher close rates and margins among high performers without publishing a multiple.
- The deck contains one monetary figure, a 500 dollar monthly payment in the customer journey example, converted at 16.21 rand to the dollar, the mid-market rate on 17 August 2026.
- Ours rather than the presenters’: the comparison of the 64 per cent figure against Cox’s own 19 per cent, the comparison of the 37 per cent figure against Cox’s own 63 per cent, the selection-effect argument against the 2X promise, the point that the Cox tool stack has no South African equivalent and what local providers stand in its place, and the entire POPIA and National Credit Act section. Nothing here is legal or compliance advice.