AI, personal agents, and the future of online marketplaces: a Q&A with Steve Kooyers

The short answer is that AI is rapidly changing how people use the internet. We are still in the early innings, but search behaviour is already shifting. Some users have moved a meaningful part of their day to tools like ChatGPT or Gemini, while others are experimenting alongside traditional search.

What matters for marketplaces is that AI is becoming a default starting point for many online tasks, from researching a purchase to planning a move. As the models get better and cheaper, the question is not whether consumers will use them, but how far along the decision journey they will let AI go on their behalf.

For marketplaces that have enjoyed reliable flows of free traffic both directly and from search due to their strong brand recognition and superior user experiences, this is both a threat and an opportunity. The threat is loss of direct access to the customer if LLMs or personal agents disintermediate the marketplace from the consumer journey. The opportunity is to use the same technology to create better, more personalised vertical specific experiences than before.

Think of a personal AI agent as a trusted digital specialist who knows your preferences, budget and constraints, and can act for you online. Instead of you opening ten tabs to research flights, schools or neighbourhoods, the agent can do that legwork, filter options and even start transactions.

In a few years it is plausible that most people will rely on a small number of personal agents for a large share of what they do online and not actually interact themselves with the websites. That is the core disruption risk for marketplaces. Today marketplaces own the relationship with the buyer. In an agent driven world, that relationship could move to whoever provides the best personal assistant for the purchase in that specific vertical.

For instance, your personal agent might talk directly to an auto marketplace agent, a finance agent, and an insurance agent to present you with a short list of fully costed options, or to complete a purchase with minimal friction. If marketplaces are not part of that conversation and powering the best personal agent tools for their vertical to own the user journey, they risk becoming invisible infrastructure.

Marketplaces are much more resilient than the alarmist headlines suggest and have strong moats, but they will need to continue to innovate and evolve (as they have through past technological shifts). Their core strengths remain powerful:

• The most comprehensive selection of real-time inventory, often including both professional and private sellers
• Clean, structured data and search taxonomies building on years of historical listings and market specific data that improve comparison
• Vertical expertise and integrations utilizing proprietary data, such as valuation guides, history reports and financing tools which enhance listings and search
• Strong brands and high levels of consumer trust, which is especially important for high value, infrequent, emotional purchases like homes and cars
• Direct relationships with consumers, with the best marketplaces generating more than 70% of their traffic directly and more than 50% from Apps

Those things, and the network effects they create, are hard for generalist AI systems to replicate and remain the primary source of defensibility for leading online marketplaces. By concentrating a comprehensive view of real-time inventory, proprietary data, and specialised tools, marketplaces produce the best search experience for buyers while delivering a critical share of demand to sellers. This flywheel is difficult for AI agents or other generalist players to replicate as simply scraping and presenting listings and compiling publicly available data fails to create as compelling of a user experience.

However, two areas where we see potential for marketplaces to be more exposed are (i) how they monetise and (ii) the sustainability of their margins.

On monetization, many rely on prominence-based models, where vendors pay for better placement in search results. If an agentic solution is only surfacing highly targeted listings to a buyer, the traditional upsell for premium placement matters much less. Over time that could push marketplaces toward more transactional or performance-based models, especially where inventory is exclusive and outcomes are easier to track.

On margins and profitability, one of the many attractions of the marketplace business model is the high flow-through rates of incremental revenue which has enabled many businesses to sustain margins in the 60%+ range. If those businesses have historically underinvested in their technology platforms and capabilities, the margins could come under pressure as incremental investment is needed to enable them to compete in an AI world.

That said, these risks are nuanced by specific company and vertical with many marketplaces taking the right steps historically to build the foundation for future success. Appropriate diligence and sub-sector expertise is needed to understand whether it is a risk or opportunity to further extend market leadership vs. less well-prepared competition.

No, we see the potential impact and threat very differently by vertical and business model.

We believe that marketplaces which are focused on high value, low frequency, and highly emotional purchases where the consumer must balance trade-offs with unique inventory are much less likely to be disrupted. In these cases, the consumer wants to be guided and needs the vertical specific data and tools provided by a marketplace but often prefers to play an active role in the search. Traditional real estate and automotive marketplaces would fit into this bucket.

Meanwhile, marketplaces for more standard goods with higher frequency of purchase or lower perceived impact on users are more susceptible to being disrupted. New goods marketplaces is an example of this category, featuring undifferentiated goods where users are simply looking for the best price. Jobs marketplaces can also fit into this category with AI likely replacing many of the core tasks of recruiters today, limiting costs to users to multi-source or use personal agents to apply for roles.

So, if you step back, for consumer marketplaces we see a rough ordering of disruption risk from lower to higher: property and automotive, jobs, then new goods. Consumer to consumer marketplaces focused on unique used inventory with integrated transactional capabilities may be the most well protected, provided they keep innovating on product as well as trust and safety.

Outside of consumer facing marketplaces, we see transactional B2B marketplace models like automotive, or salvage auctions as well protected with robust moats as they provide mission critical proprietary data and tools (i.e. inspections, arbitration, transport, etc.) which enable the ecosystem and own the auction where the transaction occurs.

There are several moves I would classify as “no regrets” moves for leadership teams that we already see in action across our portfolio.

First, optimise for AI traffic, not only for traditional search. That means thinking about how your content is exposed to LLMs and agents, while still defending your own user experience.

Second, invest aggressively in your own AI powered experience. If consumers can get a conversational, guided journey from a general AI tool, they should get an even better one from a specialist marketplace. That might include building AI agents specific to your vertical, natural language search, smarter recommendations, richer listing descriptions and proactive alerts.

Third, double down on proprietary data and tools. Pricing and valuation tools, history reports, integrated financing calculators, and rich media all deepen your moat and enhance the differentiation of the AI agents and tools you are building.

Finally, use AI internally to drive productivity. Code assistants, automated moderation, smarter customer support, and marketing automation can lower your cost base and free up resources to invest in the product.

The best operators are doing all the things I just mentioned and are not waiting for a perfect roadmap. They are investing ahead of the curve and running controlled experiments across the business. On the buyer side, that includes internal AI agents that behave like expert shopping assistants, the introduction of LLM style search, heavy personalisation of homepages and search results, and improved matching algorithms.

On the seller side, we see tools that create or enhance listings automatically, from photo clean up to suggested pricing, and feature extraction. There is strong interest in workflow tools around chat, payments, and finance that move marketplaces closer to the transaction and make their value more tangible to vendors.

At company level, leaders are investing in data infrastructure, upskilling teams on AI, and appointing clear ownership for AI strategy. They are also becoming more deliberate about what data is freely available and what requires a relationship, so that when personal agents show up at the door, they can engage on the marketplace’s terms.

In short, the winners are treating AI as both an offensive and defensive tool. They are improving today’s economics while quietly preparing for a world where AI and personal agents are mainstream.