Real-time Connection— Reimagining auto marketplace lead journeys with AI

This webinar brought together product leaders from Europe’s leading automotive marketplaces to examine how AI is reshaping the lead journey — from first search intent to dealer contact. Key themes include: how the lead itself is evolving from a one-way handoff into a dynamic three-way conversation; what enriched buying signals look like in practice; how marketplaces are navigating GDPR while building smarter connections; the competitive case for conversational search; and how AI voice agents are solving the missed-call problem that costs dealerships millions in lost revenue every year.

Panellists: Jonathan Turpin (Principal, AIM Group, moderator) · Karolina Edwards-Smajda (Chief Product Officer, AutoTrader UK) · Christin Schink (mobile.de) · Robin Suwe (EVP Mobility, Vend Marketplaces) · Henrik Lenerius (CPO, Kaisa)

The Lead Journey Today: Long, Nonlinear, and Mostly Online

Jonathan Turpin, AIM Group

The car buying journey today typically spans weeks or months. Buyers move between search platforms, comparison tools, social content and dealer websites — often circling back multiple times before making contact. The marketplace sits at the centre of that journey, but the moment a lead is submitted, most marketplaces hand it off and step back. The question this session is really asking is: should that be where the marketplace’s role ends?

Karolina Edwards-Smajda, AutoTrader UK

At AutoTrader, we see buyers who start their research with us months before they’re ready to purchase. They save vehicles, set price alerts, compare spec for spec. By the time they submit a lead, they know exactly what they want — but the dealer receiving that lead often knows nothing about them beyond a name and a phone number. There’s an enormous amount of context lost in that handoff, and that context is exactly what AI can help preserve and communicate.

Robin Suwe, Vend Marketplaces

In the Nordic markets — Finn.no in Norway, Blocket in Sweden — we benefit from very high login rates. Buyers are typically signed in, which gives us a complete view of their on-platform behaviour: what they’ve searched, what they’ve saved, how long they spent on a listing, whether they’ve looked at similar vehicles. That behavioural layer is incredibly valuable, both for understanding intent and for building richer connections with dealers.

Data as the Foundation: Identity, Behaviour, and Verified Vehicles

Robin Suwe, Vend Marketplaces

The Nordic markets have an advantage that’s easy to underestimate: BankID. Because so many people use BankID to log in to our platforms, we can verify who a buyer actually is. That matters enormously when you’re trying to pass a lead to a dealer. It’s not just “someone filled in a form” — it’s a verified individual with a real purchase history. That changes the quality signal completely.

When you add verified vehicle data on top — structured data from VIN lookups, condition reports, service history — you have a foundation for AI to work on that’s genuinely trustworthy. AI can only be as good as the data it’s built on. In markets where vehicle data is inconsistent or buyer identity is unverified, AI recommendations and lead enrichment will be much harder to deliver well.

Christin Schink, mobile.de

Germany is in some ways at the opposite end of the spectrum from the Nordics. Login rates are lower, and privacy expectations are higher. Buyers are cautious about sharing data, and rightly so. But what we’ve found is that when you can demonstrate clear value — a better search result, a more relevant recommendation — consent rates are actually quite good. People will share data if they understand why it helps them. The barrier isn’t data protection regulation per se; it’s demonstrating value clearly enough that buyers actively opt in.

mobile.de handles over 1.6 million listings and around 140 million visits per month. At that scale, even small improvements to lead quality have a very material impact. The data challenge for us is less about collection and more about consistent structuring — making sure vehicle data is clean, complete and comparable across listings so that AI can do something useful with it.

How the Lead Itself Is Evolving

Karolina Edwards-Smajda, AutoTrader UK

The traditional lead is a one-way handoff. A buyer fills in a form, the marketplace fires an email to the dealer, and the marketplace’s role is over. What we’re moving towards — and what AI makes genuinely possible — is a lead that is the beginning of a three-way conversation: buyer, dealer, and marketplace, all connected in real time.

The marketplace has context that neither the buyer nor the dealer has alone. We know how the buyer found the listing, what else they looked at, how long they spent on this specific vehicle, whether it’s been price-reduced since they first saved it. When that context travels with the lead, dealers can respond more relevantly. And when the conversation continues — when a dealer can send follow-up information back through the marketplace, and the buyer can ask further questions — you have something much more powerful than an email address and a phone number.

Robin Suwe, Vend Marketplaces

In C2C markets, this is even more important. A private seller has no sales infrastructure, no CRM, no follow-up capability. The marketplace is effectively their dealer — and AI can play the role of a helpful intermediary, keeping the conversation alive, answering buyer questions using the listing data, and nudging both parties towards a transaction. That’s a very different value proposition than just displaying a classified ad.

Buying Signals: What AutoTrader Now Sends Dealers

Karolina Edwards-Smajda, AutoTrader UK

We’ve moved well beyond sending a name and a phone number. A lead notification from AutoTrader today includes: how many times the buyer viewed this specific listing, whether they’ve saved it, whether they’ve looked at comparable vehicles from other dealers, their approximate location relative to the dealership, and where in their buying journey they appear to be — early research or active decision-making.

We also flag high-intent signals: someone who has set up a price alert and then submits a lead within an hour of a price reduction is a very different prospect from someone who clicked enquire by accident. Helping dealers prioritise those leads — and respond to them in the right order, with the right context — is where AI adds immediate commercial value.

Jonathan Turpin, AIM Group

Are dealers actually using this information when it arrives, or does it disappear into an inbox unread?

Karolina Edwards-Smajda, AutoTrader UK

It varies significantly. The dealers who use it well are seeing materially better conversion rates. The ones who don’t are treating it like any other lead notification. Part of our job is making the enrichment so clear and so actionable that it’s impossible to ignore — not a paragraph of context buried at the bottom of an email, but a clear signal at the top: this buyer is high intent, they’ve been looking for three weeks, they’ve seen five similar cars. Call them first.

GDPR and Consent: Navigating the Regulatory Environment

Christin Schink, mobile.de

GDPR compliance in lead generation isn’t just a legal checkbox — it shapes the entire lead flow. In Germany, buyers must explicitly consent to having their contact details shared with a dealer at the point of lead submission. The consent has to be informed, specific and unambiguous. You can’t pre-tick boxes or bury it in terms and conditions.

What we’ve found is that if you present consent clearly — explain exactly who will receive the data and why — opt-out rates are surprisingly low. Most buyers are submitting a lead because they want to be contacted. They’re not surprised that a dealer will call them. The friction comes when the consent language is confusing or the buyer doesn’t trust what will happen to their data next.

Robin Suwe, Vend Marketplaces

The Nordic approach has been to lean into privacy as a feature. Because we have strong identity verification and a clear data trail, buyers actually feel safer sharing information with us than they might on a less trusted platform. Trust is the unlocking mechanism for data. Platforms that have built genuine trust with their users over years are better placed to handle consent gracefully than those trying to extract it as an afterthought.

Karolina Edwards-Smajda, AutoTrader UK

Cross-border differences within the EU are a real operational challenge. The GDPR framework is harmonised in principle but interpreted very differently by national regulators. What’s acceptable in the Netherlands isn’t necessarily acceptable in France. For a marketplace operating across multiple markets, that requires building consent flows that can adapt to local requirements without fragmenting the product experience entirely.

Conversational Search: What’s Live and What’s Coming

Christin Schink, mobile.de

mobile.de has been running a conversational search assistant — we call it Mobi — for some time now. The idea was straightforward: car search has too many filters, and most buyers don’t think in filter terms. They think in need terms. “I need a family car with a large boot and I don’t want to spend more than €400 a month.” Mobi takes that kind of natural language input and maps it to our search index.

What we found in practice was that users who engage with Mobi spend longer on the platform and view more relevant listings. The quality of the search session improves. But adoption has been more gradual than we expected — changing search behaviour takes time, especially for users who have established habits. The interface needs to feel genuinely faster and easier than the filter-based alternative, not just different.

Robin Suwe, Vend Marketplaces

We’ve run experiments with conversational interfaces on our Nordic platforms, and the learning for us has been about expectation-setting. When a conversational UI doesn’t understand a query — or interprets it differently from what the buyer meant — the failure is very visible. With a filter, if you get zero results, you adjust. With a conversation, a misunderstanding feels more personal. The design challenge is handling those failures gracefully, so the experience stays helpful rather than frustrating.

Karolina Edwards-Smajda, AutoTrader UK

At AutoTrader we’ve been building the foundations. We launched needs-based search categories that allow buyers to search by life stage — “family car”, “first car”, “van conversion” — rather than make and model. That’s a step towards a more conversational experience without the full complexity of a natural language interface. We’re learning about how buyers respond to need-framed search before going further.

Kaisa: Using AI to Fix the Real-Time Connection Problem

Henrik Lenerius, Kaisa

The connection problem is simple to state but very expensive to ignore. When a buyer submits a lead or picks up the phone, they have reached the moment of highest intent in the entire journey — weeks or months of research have led to this point. What happens in the next few minutes determines whether that intent converts into a sale or evaporates.

The data we see across the market is stark. Around 40% of inbound calls to dealerships never reach a human — they hit an IVR, wait in a queue, or go unanswered entirely. More than 20% end in voicemail, and most voicemails are never returned. Meanwhile, research consistently shows that approximately 75% of used car buyers purchase from the first dealer they successfully speak to. Miss that call, and you very likely lose the sale — not to a competitor you know about, but to a competitor who happened to pick up the phone.

At Kaisa, we’ve built three layers of AI capability around this problem. The first is analysis: understanding where connection failures are happening, at what times, for which listings, and correlating that with lead outcome data. The second is reaction speed: using AI to route leads more intelligently and alert salespeople in real time. The third — and the most impactful — is prevention: AI voice agents that step in the moment a call is about to be missed and conduct a real, context-aware conversation with the buyer.

The key word is context-aware. An AI agent that answers with “how can I help you?” and asks generic qualification questions is a worse experience than a voicemail. But an agent that already knows the listing — the specific car the buyer called about, its price, its spec, the dealership’s location, current stock availability — can have a genuinely useful conversation. It can answer questions about the vehicle, explain what’s included in the price, confirm that a test drive can be arranged, and collect qualification information naturally in the flow of that conversation.

From our market tests in Spain, Germany and other markets, up to 64% of leads handled by AI voice agents were fully qualified — timeline, financing intentions and trade-in discussed, contact details confirmed. What began as a call about to be missed becomes a structured, high-quality lead delivered to the dealer with a full call summary. Dealers respond faster because they know exactly who they’re calling back and why.

Agents Talking to Agents: How Marketplaces Stay Relevant

Jonathan Turpin, AIM Group

As buyer-side AI agents become more capable, there’s a real question about whether they’ll bypass the marketplace entirely and go direct to dealer inventory. What gives a marketplace a defensible position in that world?

Karolina Edwards-Smajda, AutoTrader UK

Trust and data depth. A buyer’s AI agent will go where the most reliable, comprehensive data is. If our vehicle data, pricing benchmarks and dealer reputation signals are better than anything else available, the agent comes to us — not because it’s been blocked from going direct, but because we’re genuinely the best source. The marketplace’s job is to make sure that’s true, and to make our data accessible in whatever format AI agents can consume.

Robin Suwe, Vend Marketplaces

Identity verification is a moat that’s very hard to replicate. In the Nordic markets, BankID-verified buyer identity is something only platforms integrated with national banking infrastructure can provide. A dealer-side AI agent that wants to know whether it’s talking to a real, qualified buyer — and not a fraudulent lead or a competitor scraping their stock — needs that verification layer. Marketplaces that provide it become essential infrastructure in an agent-to-agent world, not bypassed middlemen.

Christin Schink, mobile.de

Marketplaces also aggregate the market in a way that a single dealer can never do. A buyer’s agent needs to compare across thousands of vehicles, across hundreds of dealers, in real time. That aggregation function — keeping the data current, structured and comparable — is exactly what marketplaces are built to do. AI doesn’t make that job redundant; it makes it more valuable.

Measuring Success: Moving Beyond VDP Views

Jonathan Turpin, AIM Group

Marketplaces have traditionally measured success by traffic volume — visits, listings viewed, leads submitted. But if lead quality is what actually matters to dealers, is the industry measuring the right things?

Karolina Edwards-Smajda, AutoTrader UK

This is something we talk about a lot internally. VDP views are easy to measure and easy to report. Whether the buyer actually bought a car is much harder to track — the data trail ends at the lead, and what happens after that is often opaque to us. We’re investing in attribution capabilities that can connect a lead submitted on AutoTrader to an eventual sale at the dealer. “Unconfirmed sales” — where our data strongly suggests a transaction occurred but we can’t directly verify it — is a metric we’re starting to develop.

Robin Suwe, Vend Marketplaces

In C2C markets, the attribution problem is even harder — we have no direct relationship with the transaction. But proxy metrics can still shift behaviour. If you reward salespeople and account managers for lead-to-contact rate and lead-to-sale rate rather than lead volume, you change the incentives throughout the chain. The industry has been slow to move away from volume metrics, but AI-generated lead quality scoring gives marketplaces something concrete to show dealers that volume and quality are not the same thing.

Christin Schink, mobile.de

We’ve started working with a subset of dealer partners who are willing to share outcome data with us. The patterns are revealing. High-intent leads — buyers who’ve spent significant time on a listing and saved it before enquiring — convert at dramatically higher rates than low-intent leads. If we can surface that signal consistently, dealers get better value per lead and are more willing to invest in the marketplace. It’s a better commercial model for everyone.

The Risks: Hallucinations, Fraudulent Leads, and Automating the Wrong Thing

Robin Suwe, Vend Marketplaces

The risk I think about most isn’t hallucination — it’s automating the wrong thing faster. AI makes processes more efficient, but if the underlying process is wrong, you just get to the wrong place more quickly. In the lead space, there’s a risk that AI optimises for lead volume at the expense of lead quality, or optimises for speed of response at the expense of a genuine conversation. You have to be very deliberate about what you’re optimising for before you hand it to an AI system.

Christin Schink, mobile.de

Fraudulent leads are a real and growing concern. AI makes it much easier to generate synthetic leads — fake buyer identities, realistic-looking enquiries, volume-generated to abuse pay-per-lead models or scrape dealer contact information. Verification at the point of lead generation becomes essential, not optional. This is another area where identity verification layers — like BankID in the Nordics — provide a structural advantage over markets where anyone can submit a lead anonymously.

Karolina Edwards-Smajda, AutoTrader UK

On hallucination specifically: any AI system that is generating content about specific vehicles needs to be grounded in verified listing data, not relying on general model knowledge. A conversational agent that invents a feature the car doesn’t have, or quotes a price that’s wrong, creates a support problem and an immediate trust collapse. The guardrails have to be very tight around automotive-specific claims — spec, pricing, availability, finance rates. These are not areas where creative confabulation is acceptable.

The Fully Connected Future: Finance, Insurance, and Beyond

Jonathan Turpin, AIM Group

Looking further out — is the goal a single, seamless flow where a buyer finds a car, arranges finance, gets insurance and completes the purchase, all within the marketplace experience?

Karolina Edwards-Smajda, AutoTrader UK

That’s the direction. Finance is the most logical next connection. A buyer who knows their budget shouldn’t have to re-enter their financial situation at the dealer — that context should travel with the lead. Pre-qualified finance decisions, where the marketplace’s platform has already indicated affordability based on the buyer’s stated parameters, would dramatically accelerate the dealer conversation. Insurance is similar.

Christin Schink, mobile.de

For new cars, the end-to-end digital transaction is getting closer. OEM agency models, where the OEM controls pricing and the dealer is essentially delivering a service, simplify the commercial flow. For used cars, human validation will remain important for longer — the condition of the vehicle introduces variability that AI cannot yet fully assess remotely. A physical inspection, a test drive, a human confirming what the data says — those steps will persist in the used car journey even as everything around them automates.

Robin Suwe, Vend Marketplaces

The Nordic markets are interesting on this point. We have platforms where the entire private used car transaction — discovery, price agreement, payment, ownership transfer — happens digitally. It works because the regulatory environment supports digital identity and digital documentation. In markets where title transfer requires a physical signature or a visit to a government office, you hit a ceiling. The technology is ready; the regulatory infrastructure often isn’t.

Henrik Lenerius, Kaisa

From where we sit, the near-term opportunity isn’t the fully automated transaction — it’s making every real-time human connection that does happen significantly better. The call that currently goes to voicemail should reach a capable, context-aware AI agent. The lead that lands with a dealer without context should arrive with a full picture of buyer intent. The follow-up that currently doesn’t happen should be triggered automatically. These aren’t futuristic capabilities — they’re available now, and the gap between what’s possible and what most marketplaces are deploying today is still very large.

This webinar was hosted by AIM Group and sponsored by Kaisa. Panellists represent their own views. The transcript has been edited for clarity and readability.