AI is reshaping the car-buying journey from the ground up, but without clean data and a cohesive backend, the industry risks stalling at the starting line. With autonomous agents on the horizon, success will depend not on shiny tools but on complex infrastructure and outcome-driven thinking.
Car dealerships have long been places of theatre. Polished cars under showroom lights, handshakes over coffee, the reassuring tap of a tyre by a hopeful buyer. Yet behind the gloss, inefficiencies have lingered for decades. As AI integrates into every part of the sales and servicing cycle, those inefficiencies are being exposed.
Johan Sundstrand, founder and CEO of the Swedish-British firm Phyron, has spent the past few years examining how AI can accelerate the evolution of the automotive retail sector. Still, he sees a fundamental mismatch at the heart of many dealerships’ digital transformation efforts. “They have embraced digital but not necessarily become digital businesses,” he says. “What we have is a duplication of physical and digital layers, not integration.”
The result, Sundstrand says, is an expensive sprawl of CMSs, DMSs, web tools, advertising systems, and compliance modules. “They have to be visible across marketplaces and social channels and are producing more formats and touchpoints than ever, but they have not made more money,” he continues. “If anything, they are spending more and struggling to measure return.”
From tools to outcomes
While many dealers have acquired new technology, few have revisited the business outcomes those tools were meant to achieve. The shift to AI has too often been driven by hype and vendor pressure, rather than a coherent view of how automation can simplify the complexity of selling cars. “You should not start with tools,” Sundstrand argues. “You should start with the outcome you want to achieve.”
AI is already making its mark on several fronts, particularly in content generation. Enhanced imagery, automated video, and even seller notes are being generated at scale. “Dealers all sell the same BMW 3 Series or Audi A4,” he says. “What matters now is who can present it in the most compelling, accurate and contextual way to a customer who is probably scrolling on a phone.”
This is not just cosmetic. AI can ingest complex technical specifications, highlight relevant features for specific buyer personas, and surface the most meaningful selling points. Instead of 50 lines of barely legible options, a prospective buyer sees a simple summary that makes the offer more approachable and transparent.
The friction between vision and infrastructure
The problem, as with many industries, is not a lack of vision but rather inadequate infrastructure. AI systems require structured, high-quality data. Yet many dealer systems remain patchwork at best. Legacy platforms often fail to communicate with each other. Ownership histories are buried in unsearchable documents. Inventory remains unpriced or mispriced because pricing tools do not utilise real-time data.
“The biggest frustration is not the dealers themselves,” Sundstrand says. “It is the systems they are forced to use. They are inadequate. And without clean, accessible data, 80 per cent of AI projects will fail. Twice the failure rate of non-AI projects.”
Until now, the sector has tolerated this dysfunction. Cars still sold, margins remained acceptable, and customers adapted. But that tolerance is waning. Buyers are now navigating hundreds of digital and physical touchpoints, across awareness, research, comparison, booking, test drives, and follow-up. Younger consumers, in particular, are voting with their thumbs. “We ran a survey of over 2,000 UK adults and found a third of 18-44 year olds would let AI handle the entire purchase,” Sundstrand adds. “Not just help with research or shortlisting. Everything. No human interaction. That is how complex and stressful this process has become.”
Reimagining the end-to-end sales engine
The real potential, Sundstrand believes, lies in orchestrating the complete sales engine. Autonomous agents will not replace dealers overnight, but they are steadily advancing toward taking over entire workflows. “Acquisition, pricing, advertising, negotiation, contract, follow-up, it is all on the table,” he says. “What matters is sequencing.”
A dealership looking to automate cannot simply plug in an agent and hope for the best. It must start by fixing the basics: ensuring data quality, integrating systems effectively, and having a clear understanding of the steps from intake to sale. Then, piece by piece, AI can assist.
Take pricing. Today, most dealers adjust prices on a weekly or monthly basis. Yet the cost of holding a vehicle in stock, up to €35 per day in some markets, demands greater agility. AI can scrape market listings, assess mileage, service history and model popularity, then recommend price adjustments far more frequently. “We are not talking about wild fluctuations,” Sundstrand clarifies. “We are talking about nuanced, data-backed adjustments that reduce holding costs and improve turnover.”
Similarly, in inventory planning, AI can predict which models are likely to move fastest, which should be de-emphasised, and what acquisition price would ensure resale margin. It is not just about having stock; it is about having the right stock at the right time, at the right price.
Trust and the invisible assistant
This level of automation inevitably raises questions of trust. Can buyers rely on AI to be honest? Do they even want to know that AI is involved? For many, especially those accustomed to the traditional showroom experience, there is still comfort in human advice. Sundstrand does not dismiss this. “AI should support complex decisions, not remove choice,” he explains. “It can triage, it can inform, it can guide. But in edge cases or high-value transactions, a human should always be available.”
At the same time, he is clear that consumers have already shifted their expectations. They research online, compare models, read forums, and watch videos before speaking to a salesperson. In this context, AI is not an intrusion; it is the assistant that was always missing. “No one in my house watches content that does not move,” Sundstrand says. “That is the new default. Across all generations, what people want is not less information, but less stress. If AI can summarise, personalise, and explain, it is adding value.”
This shift also has implications for trust in dealership brands. While the physical site may still be polished and carefully laid out, the online experience has become the new first impression. Generic listings, outdated descriptions, or inconsistent pricing are no longer forgivable. As Sundstrand puts it, “Reputation now travels through pixels.”
From showroom to software company
The long-term trajectory, Sundstrand believes, is one where dealerships become hybrid software operations. Not because they start coding platforms themselves, but because they embrace data-driven decision-making, automation, and customer engagement at scale. “To benefit from AI, you need to behave like a software company,” he says. “That means fixing your data, integrating your systems, and focusing on workflows. You cannot just throw a chatbot at the problem and hope it improves conversions.”
He is not suggesting this happens overnight. However, he believes that those who start with simple, outcome-based implementations, such as conversational AI, marketing automation, and dynamic pricing, can gain real momentum. “Then scale,” he advises. “Then link the tools. Then let the agent emerge.”
As for the question of whether AI will be able to sell a car in 2025, Sundstrand has already gone on record. “Technically, yes,” he concludes. “Human-in-the-loop, even more so. The barriers now are cultural, not technological.”
The showroom may not vanish entirely. However, increasingly, it is being upstaged by systems that understand your preferences before you even arrive. The deal is being struck not over a handshake, but in the subroutines of an algorithm, designed not just to sell, but to understand, anticipate, and deliver.
