How to AI-proof your business: the case of service marketplaces

Is AI simply buzz, or a business must-have? Even as the technology continues to develop, AI is here, and it’s already transforming how businesses operate, scale and succeed – or fail. For service-based digital platforms, the sweeping changes that AI is driving are full of potential for optimisation and growth. At the same time, marketplaces are at risk of watching AI desintermediate one side of their platform.

From automation and AI agents to regulation and defensibility, AI is both reshaping existing playbooks and writing entirely new ones. “We’ve been doing a lot of work over the last couple of years at Launchworks and Co on how to embed AI into platforms and more generally organisations. And what we’ve seen is that AI redefines the rules of competition,” says Laure Claire Reillier, COO of Launchworks & Co. and co-author of Platform Strategy. “It’s both disrupting and accelerating businesses.”

AI: business enabler or existential threat?

There’s no doubt that AI is revolutionising how marketplaces, and businesses more widely, operate. “AI is truly touching every aspect of my daily work,” says Jeremy Gottschalk, founder & CEO of Marketplace Risk and author of Bulletproof Your Marketplace.

Before becoming co-founder and president of Edumapper, Sébastien de Lafond co-founded, scaled and exited MeilleursAgents, a major digital marketplace for real estate in France. At MeilleursAgents, Sébastien and his team used machine learning to enable consumers to navigate an opaque market. “Machine learning and deep learning could lead to better information, despite difficult access to data,” he says. “At the time, AI was not a threat – it was an enabler.”

‘Sometimes AI feels overwhelming, but once you start, you quickly see its value and impact.’ – Tito Unda

More recently, platforms like Job&Talent have applied AI throughout the user lifecycle. Tito Unda, ex-VP of Product at Job&Talent, explains: “We’ve used AI all across the journey, from the more basic machine-learning opportunities on the algorithm side to fine tuning and optimising the matching, but also to curate content on the marketplace.” That has included using AI agents to handle some of the most repetitive and replicable interactions. “Sometimes [enforcing policies] is easier with AI than with humans,” thinks Tito. In fact, Job&Talent’s AI agents are now capable of handling tasks ranging from recruitment calls to sharing performance feedback.

But not all the impacts are positive. AI isn’t just augmenting platforms – it’s also threatening to outperform or displace them. “If our platform is not able to provide a better answer than ChatGPT,” says Sébastien, “let’s give up.” Service-based platforms face a critical question: how can they not only keep up, but stay competitive? He admits that “we need to ask this question to ourselves because if we don’t, then the market will answer in our place.”

Building a defensible marketplace in an AI world

‘AI redefines the rules of competition. The question is, how do you build a defensible moat?’ – Laure Claire Reillier

For a start, businesses need to actively protect what makes their platform unique. “At the end of the day,” Sébastien says, “we’ll have to make our business defensible.”

One defense strategy is to build proprietary datasets that generalist models can’t access or replicate. As Jeremy warns: “If you’re leveraging these AI tools, all of the information and content that you input [ends up] in the public domain.” Organisations have to consider carefully what they’re willing to expose.

‘Keep your data confidential and manage AI carefully—it’s a regulatory wild west out there.’ – Jeremy H. Gottschalk

“It’s really a top-down approach: understanding the rights or loss of rights that go along with using some of these public tools,” says Jeremy. “And as an organisation, [deciding] what’s acceptable and what’s not – what information do you want out there and what don’t you want out there?”

Expertise is another pillar of defensibility. “Being the specialist and the absolute expert in your field is absolutely key,” Sébastien thinks, especially when trust and accuracy matter more than speed. Jeremy agrees: “Training these models for your specific use case is going to be where some win and others lose.”

Platforms can also tap into network effects, and Sébastien points to how Edumapper is building a data advantage. “There are other [data] sources that we will be able to aggregate through network effects, through the contribution of our customers, schools and courses and users of the platform,” he says. “They will contribute to building that [proprietary] database.” This emphasises the relationship between various defense strategies: domain expertise and strong network effects will likely be critical for building the proprietary datasets that can act as a business differentiator.

The new moat: trust, brand and AI alignment

Even as AI disrupts business models, trust will continue to be a cornerstone of long-term defensibility. For marketplaces, trust drives participation. Participation generates proprietary data, and data powers better experiences. This flywheel effect is what defines enduring platforms.

“We should never lose the trust of our users, because that’s our key currency” – Sébastien de Lafond

Sébastien emphasises that building this flywheel starts with credibility. “We should never lose the trust of our users, because that’s our key currency,” he says. “Building brand around that is important so that people associate your brand to that reliability and specialisation.” A strong brand can be a key signal to users that the platform’s AI is trustworthy and well-governed. “If you cannot trust that this AI is benevolent [and] unbiased, then the whole trust in the platform will disappear,” he thinks.

Establishing this kind of trust early is also pivotal. “The question is: how do we kick start the process so that we start building that trust and start to learn those network effects so that people contribute?” Sébastien asks. Once users see value and fairness in the system, they are more likely to share data, provide feedback and participate in ways that further strengthen the platform.

Tito agrees that trust is a prerequisite for defensibility. “First you have to generate trust with existing users,” he says. “That’s what really will build […] brand equity over time.” Once that trust is in place, the conditions are in place for compounding effects to begin.

When the marketplace becomes a product

Sometimes, however, defending your business means transforming it entirely. “There are many marketplaces that are in a different battle,” says Tito. For platforms that primarily offer access to information, AI could present an existential challenge. “What is it that your market provides for your clients? If it’s just knowledge, then it is going to be tricky,” warns Tito. “Then you have to really focus on building trust, [and] having access to information that no one [else] has.”

Chegg, the tutoring marketplace, was valued at $12 billion in 2022. But “when ChatGPT launched in November 2022, it significantly disrupted Chegg’s business model“ says Laure Claire. Chegg started to lose traffic as “students were asking their questions to ChatGPT instead”. “By November 2024, Chegg’s stock [price] had declined by 99% from its peak” she adds.

Tito predicts that some service-based platforms will find themselves at a crossroads. “You have two options: either give up, or transform your business from a marketplace to SaaS,” he thinks. According to Tito, the key is to shift from brokering services to acting as the product itself. “The demand [for the service] still exists,” he says, “and you’ve become an expert in that field. Instead of still being a marketplace where individuals get to offer that service, it’s more of a SaaS approach in which you become a product.” Although this represents a significant pivot for many marketplaces, he believes it could be necessary for some businesses in order to stay relevant.

Guardrails for growth: AI ethics and oversight for platforms

When platforms implement AI at scale, they also have to manage its risks, from bias to legal uncertainty. As Sébastien points out, ethical design is especially important for Edumapper and other EdTech platforms. “We built the Edumapper platform to foster equality of chances in a domain where access to the proper information is clearly biassing the trajectories of future students,” he says. The challenge is to avoid reproducing the very patterns the platform aims to fix. “As we train AI and our agents, we could reproduce some of the [biases] which we see in our societies, without actually realising it,” he adds. “We will have to ethically control the way that our agent works, so that we don’t reproduce the things that we’re fighting against.”

For recruitment and user lifecycle, as at Job&Talent, AI can be used to reduce unconscious human bias. “There is a hard-coded statement in the prompt that says, ‘do not consider the accent of the person’ [during the virtual recruitment process],” shares Tito. However, regulatory complexity does compound the difficulties when implementing AI. “It’s been challenging,” he says, “considering that we are live in 10 countries [that have] completely different regulations when it comes to AI.”

Regulatory differences across markets is something that Jeremy also sees as a serious obstacle. “The key regulatory challenges are complete inconsistency, right now,” he thinks. “It’s like a wild west situation where it’s nearly impossible to advise people.” Adding to the complexity, he thinks, is the legal ambiguity around content ownership. “What happens when there are hallucinations or you get some content generated that is inappropriate?” Although OpenAI recently won a defamation lawsuit in the US which would have held it legally responsible for user-generated content, this is a fight that is sure to continue in courtrooms around the world.

Ultimately, Tito emphasises the need for careful boundaries. “AI is like a super-energised intern. They really want to show off,” he says, “but you have to set proper guardrails.” And as Laure notes, the stakes are likely to keep increasing: “Today it’s an intern, and tomorrow it might be a senior contributor as AI improves.”

Scaling AI: start small, think big

For businesses ready to start using AI as part of their business operations, Tito recommends thinking proactively about specific use cases and behaviours. “It’s important that you define what you want that AI to do, but it’s as important to define what you don’t want that AI to do,” he thinks, adding that rolling out AI is behavioural as much as it is technical. “You’re also changing a behaviour on the user,” he says. “Users have been used to getting a phone call from a human or getting a notification, an email filling in a form, and all of a sudden you’re changing the game.”

That’s why Job&Talent started small before scaling AI in their processes. “We’ve been focused on making sure that we nail the experience first,” Tito continues. That early testing gave the team confidence that they were on the right track: “We had a very clear understanding that it was already working and it was just a matter of scaling it.”

In Jeremy’s opinion, many platforms are still stuck in the experimentation phase. “I’ve seen a lot of marketplaces just integrating tools like ChatGPT, and they’re not training it,” he says. This could reflect a larger issue of adoption without any strategic goals. “There’s a real tendency, particularly in the tech world, to just follow the buzzwords for funding and to see where they go,” he thinks. Instead, he urges platforms to think critically about the ‘how’, not just the ‘what’: “Making sure that it fits into your business [is really important].”

Practical advice for platforms navigating AI

The value of a clear strategy is critical when it comes to AI. According to Tito, success still comes down to fundamentals: “GMV [Gross Merchandise Value], liquidity, whatever your north star is – you have to understand how AI could contribute to that goal.” And, he adds: “Make sure that AI in your company is not just marketing and is really delivering an outcome that has an impact on your P & L.”

When he thinks about running a business in this space, including tackling the new challenges and possibilities of AI, Sébastien turns to metaphor: it’s a little like dancing with a bear. “You have to leverage the formidable strengths of that bear, and it’s a formidable opportunity, but you have to be aware also that if you’re not prudent, it could tear your head off,” he says.

For his part, Jeremy recommends giving employees time to explore and learn independently: “I’ve seen some really cool company-wide policies where they’re requiring every employee to play with a [GenAI] tool for [one or] two hours a week.”

At the end of the day, Tito reminds business leaders to stay clear-sighted. “It’s important not to lose focus,” he thinks. “Understand that AI is a method, a tool.” And Jeremy’s message to entrepreneurs and practitioners is simple: “Go for it.”

Shape the debate with Platform Leaders

This panel discussion with Sébastien de Lafond, Jeremy Gottschalk, Tito Unda and Laure Claire Reillier took place during the online Platform Leaders event hosted by Launchworks & Co on the 5th of June 2025. Hundreds of participants – including entrepreneurs, academics, policymakers, investors and industry practitioners – from around the world and across the Platform Leaders community were in attendance.

Platform Leaders is a global network of experts exploring the future of digital platforms. At our offline and online events, we focus on the defining issues – from emerging technologies to market regulation – that are shaping platform strategy today. Subscribe today to stay informed about upcoming events, and find session recordings, deep insights and much more on the Platform Leaders website.

To watch the full event, you play the video below.