GM Is Building an AI That Predicts What the Car Needs. Your Store Still Has to Take the Call.

An automaker just moved the service conversation into the dashboard

On July 31, General Motors told CNBC it plans to launch its own in-vehicle AI assistant later this year - one built to go further than the Google Gemini assistant GM rolled out across its 2022-and-newer vehicles earlier in 2026. Anna Santos, GM's director of product management for voice and AI/machine learning, described a system that understands the vehicle, the drive, and the customer's needs by combining conversational AI with GM's own vehicle knowledge and OnStar intelligence.

One line in the coverage matters more to a fixed operations director than the rest of it. GM says it is working with a large language model provider on automotive-specific capabilities that draw on vehicle telemetry and proprietary data to support predictive maintenance.

Read that as an operating statement rather than a product announcement. If it ships as described, a GM owner's car will start telling them what it needs - in plain language, in the cabin, before a warning light comes on and before anyone at your store picks up a phone.

The underlying technology is not new. Maintenance minders and OnStar diagnostic emails have existed for years. What changes is the interface, and interfaces are where habits form. A dash light is a prompt the driver has to interpret, remember, and act on later. A conversational assistant that knows the vehicle's service history can name the issue, explain the urgency, and answer the follow-up question - all at the moment attention is highest. The point of service intent moves into the vehicle, and it moves earlier.

So the question for a dealer is narrow and practical: when that prompt fires, where does it point the customer, and can your store actually take the booking?

Myth: Dealers lose service customers to independent shops on price.

Data: Cox Automotive's 2026 Fixed Operations and Ownership Study puts average consumer spend per visit at $261 at a dealership versus $275 at general repair. Dealers are slightly cheaper. Dealer share of service visits still fell from 33% to 29%. The leak is not the price tag - it is who is present at the moment the customer decides.

Source: Cox Automotive, 2026.

Why this lands on an existing soft spot

If dealer service retention were strong, an OEM assistant would be a nice convenience and nothing more. It is not strong, and the numbers are not ambiguous.

Two things stand out. The first is that dealers are losing share while charging slightly less per visit and while service and parts revenue per store climbed to roughly $9.23 million in 2025, up 33% over eight years. Revenue rose because vehicles got older and repairs got more expensive - not because dealers won the relationship.

The second is the gap between the 80% who intend to come back and the quarter who leave with an appointment on the books. Cox calls it a 50-point gap, and we have written about why intent does not convert into a booked first appointment. An in-vehicle assistant is, functionally, a very well-timed reminder aimed squarely at that gap. Whoever controls it controls a lot of downstream revenue.

The fleet is old, but the connected part of it is where dealers already win

Set the OEM assistant against the shape of the U.S. fleet and the strategic logic gets clearer.

S&P Global Mobility's annual analysis put the average age of U.S. light vehicles at a record 12.8 years, the eighth consecutive annual increase, across 289 million vehicles in operation with a steady 4.5% scrappage rate. Passenger cars average 14.5 years. Most of the country's service demand sits on vehicles that left any dealer relationship a long time ago, and no amount of in-dash software will reach them.

But the same analysis put the average age of battery-electric vehicles at 3.7 years - young, connected, and heavily software-dependent. And that is exactly where dealers still hold the customer: Cox found EV owners send 67% of their service visits to dealers, hybrid owners 50%, and internal-combustion owners just 28%.

That correlation is the whole point. Dealer retention is strongest precisely where the vehicle is newest, most connected, and hardest to service without factory tooling and data. An OEM in-vehicle assistant extends the reach of that connection - and it does so through a channel the manufacturer owns end to end.

Nothing about that is inherently bad for dealers. OEM-routed service leads generally land at franchised stores; that is how the model works. But "generally lands at a franchised store" is not the same as "lands at yours, and gets booked." When the prompt is generated by the manufacturer rather than by your BDC, your leverage shifts from generating the demand to capturing it.

What the evidence says about acting on vehicle-generated prompts

There is real, if limited, data on what happens when a dealer works telematics-triggered alerts rather than letting them sit.

Urban Science - a network and aftersales analytics firm that sells software in this category, so read it with the source in mind - reported that reaching out to customers at telematics-triggered intervals delivered, on average, a 3.5% increase in service revenue and close to a 4-percentage-point improvement in service retention. Those are modest numbers, which is part of why they are believable. They are also the kind of gain that compounds against a $12,000 lifetime-value figure.

The same firm's Harris Poll survey of 3,022 U.S. auto buyers found only 42% strongly agreed that they like to service where they bought the vehicle, and only 39% strongly agreed they are committed to servicing regularly at their dealership. The loyalty most dealers assume they have is thinner than the 80% intent figure suggests. A prompt with good timing can move someone who is genuinely undecided - which describes most of the customer base.

The part that does not change

Here is what an in-vehicle assistant does not do. It does not add bays, add technicians, or shorten your promise times. It creates a moment of intent. Converting that moment still requires the same unglamorous machinery it always has:

1. A phone that gets answered. Most service departments have never looked at inbound answer rate broken out by hour, which is where the 8-to-10 a.m. problem hides. If the assistant tells a driver to call the dealer and the dealer does not pick up, the prompt has just funded a call to the independent shop down the road. We have written about what unanswered service calls actually cost.
2. A scheduler the customer can reach on their own terms. Cox's high-performing dealers - the 58% who report fixed ops became both more efficient and more profitable - offer online scheduling, electronic estimate approval, and photos or videos of recommended work. That last one is not a soft benefit: customers who received photos or videos spent about $230 more per repair order, and 49% said visual evidence makes them likelier to approve recommended service.
3. Capacity to absorb the demand. Cox's high performers run bays at 90% utilization or higher. If you are already there, more captured calls produce longer waits rather than more revenue, and the sequencing matters - capture and capacity have to be solved in the right order.
4. A follow-up loop that runs without anyone remembering to run it. Declined work and open recalls are the cheapest revenue in the building and the most commonly dropped.

None of that is a new purchase. All of it is measurable this month.

The measured read

GM has announced an assistant, not shipped one. There is no name, no firm date beyond "later this year," and no public commitment that the thing will book a service appointment at all - predictive maintenance was described as a capability the technology could support, not a scheduling feature with a launch plan. Any dealer restructuring their fixed-ops strategy around a July press cycle is moving too fast.

The direction, though, is not ambiguous, and GM is not alone in it. Manufacturers have spent two years building direct-to-driver software channels, and 16% of consumers already used an AI tool somewhere in their last service journey without telling anyone at the dealership. The share of service decisions that get made before a human at your store is involved is going up, not down.

The useful response is not vendor shopping. It is finding out three things before the quarter closes: what your OEM's connected-vehicle and telematics program routes to you today and in what format, what your inbound service answer rate looks like broken out by hour, and what share of your new-vehicle deliveries leave with a first service appointment already on the schedule. Those three numbers will tell you whether a well-timed prompt from the dashboard would land in a store ready to catch it, or bounce.

The manufacturer is going to get better at telling your customer what their car needs. The part still worth owning is what happens in the ninety seconds after.

If service-drive phone capture is one of the gaps that shows up when you pull those numbers, that is the specific problem Carbuki works on with dealerships - and it is worth measuring either way.

Sources

• GM to launch proprietary in-vehicle AI system later this year - CNBC, July 31, 2026
• GM to develop AI assistant to deliver deeper vehicle integration - CBT News, July 31, 2026
• GM is building its own in-vehicle AI assistant to go beyond Google's Gemini - Quartz, July 31, 2026
• Dealerships Capture Record Fixed Ops Revenue - But Lose Market Share as Customers Drift to General Repair - Cox Automotive 2026 Fixed Operations and Ownership Study, April 9, 2026
• U.S. Vehicle Age Rises Again to 12.8 Years in 2025 - S&P Global Mobility, May 21, 2025
• Average U.S. vehicle age rises again to a record 12.8 years - Auto Remarketing, May 21, 2025
• How dealerships and OEMs can leverage the power of vehicle connectivity - Urban Science (telematics-triggered outreach results; Harris Poll survey of 3,022 U.S. auto buyers)