We at a16z have long been students of the marketplace business model. From 2010 to 2023 we wrote dozens of articles about marketplaces and funded companies like Airbnb, Instacart, and Whatnot…and then, we saw the category soften. As we moved through the mobile era, anything that could be tried, was tried – and post-2020, net new marketplaces became increasingly niche.

Two years ago, we predicted a marketplace resurgence, with a new generation of AI-native companies built from the ground up around LLMs. We were anticipating things like just-in-time product generation, agent-orchestrated price negotiation, and extremely personalized search. 

We were right that a shift would take place – but the AI transformation has unfolded in different ways than we expected. AI isn’t (so far) opening up brand new categories. Instead, it’s reviving categories where marketplaces previously failed, turning graveyards into greenfields. 

In retrospect, we probably shouldn’t be surprised. Past eras of the internet also illustrate that there are no bad ideas, just bad timing (lest we forget that 90’s-era Pets.com failed, while online pet goods today do tens of billions of dollars in sales on dedicated vertical marketplaces). In the age of AI, businesses that previously struggled to make it on the internet are getting another shot on goal. Some will win.

How? Marketplaces historically fail for one of two reasons. They may struggle to acquire converting customers, either because of a lack of demand or because making a successful match requires too much time, cost, or effort. In this case, the customer acquisition cost (CAC) is too high – talent marketplaces are one historical example. Or, they may struggle to generate enough value from these customers, either due to low repeat rates or low margins. In this case, lifetime value (LTV) will be too low – home services marketplaces are an example.

AI can transform this, making unattractive unit economics “work” – and removing bottlenecks that would previously be inhibitors to scale. We’ve seen founders tackle this in two primary ways: (1) letting AI be the middleman; and (2) changing the value proposition.

Letting AI be the Middleman

AI is taking familiar forms of jobs and services, and making them more viable and discoverable on the internet. This is because AI can vet supply and demand, make a match, and manage customer relationships in areas where transactions are infrequent, opaque, or heterogeneous – specifically, in places like skilled labor marketplaces, real estate, and repairs. 

This gives us two huge unlocks. First, you can significantly increase contribution margin per transaction, as the cost (in human time and salary) to enable a transaction goes way down. And you can significantly increase throughput (number of transactions), as humans are no longer the limiter to your marketplace growth.

Here are a few examples of how AI-native marketplaces are doing this.

Automated intake 

The very first obstacle many marketplaces face is intake. Complex matches can require significant human time to run interviews, screening calls, and credential verification, sometimes on both the supply and demand side!

AI voice agents can replace humans here. Jack and Jill and Dex, both marketplaces for high-skill talent, use AI to speak at length to candidates and employers, collect preferences and characteristics, and make matches. They can also proactively re-engage candidates as relevant roles come in, increasing the odds of a transaction. 

This should significantly increase the odds that a candidate finds a job and a company makes a hire through the marketplace, but with much less human involvement required. With an AI interviewer handling the conversations, it might cost a few dollars to make a match that would previously cost hundreds of dollars (if not more) in human labor. 

Lowering manual coordination 

Once you’ve onboarded supply and demand, you need to make a transaction happen. Previously, buyers needed to be hand-held through this process by a cast of supporting characters: account managers, sales reps, and more. 

AI can fill this back office role by handling all communications with both sides, and keeping both sides up-to-date on the best options. In house rental and sale, Spotlight Realty handles all inquiry responses, scheduling, screening, and document generation for listed properties. This massively cuts down on what a traditional broker (and their staff) would need to do to get to a rental or sale, allowing the marketplace to scale to more properties and transactions at once.

And the marketplace can even choose to “pass on” some of these labor cost savings to the end customer. Spotlight, for example, offers a seller commission of just 1.5 percent, against the 6 percent industry standard. This can transform the product’s value proposition (more on this below!) 

Driving repeat transactions 

Once you’ve gotten both sides to a first transaction, how do you get them to come back? This is particularly tricky for marketplaces where needs are infrequent and you need to catch customers at the exact right moment: when their sink is broken or their door needs to be repaired. AI can act as an always-on engagement manager. 

This can be as simple as calls or texts that are hyper-personalized based on recent service, come at the exact right time for each customer, and are conversational and not just rote reminders. One example: “You had your gutters cleaned with us last month – how are they doing now?” There’s no reason why you can’t now call every customer monthly with a follow-up that is specific to them. 

In the extreme case, marketplaces can become subscription businesses. Honey Homes charges a flat annual rate for a homecare plan executed by a human handyman, but coordinated by AI. This wouldn’t be possible without an LLM doing the heavy lifting behind the scenes, and locks users into a long-term relationship in a previously “one and done” category.

Changing the Value Proposition

On some marketplaces pre-AI, the value proposition wasn’t strong enough to get users to transact on the platform, or wasn’t strong enough to get them to be loyal to the platform. Buyers and sellers would operate across several marketplaces at once, or even disintermediate the marketplace to coordinate manually. 

AI can allow founders to change the fundamental structure of what they’re offering to either buyers or sellers. This can increase the chance that customers use the marketplace or increase the number of transactions that they complete on it.

Making a more attractive “product” for buyers 

Take a product that had critical flaws pre-AI, where buyers feel the deck is stacked against them. One example? Complex installations that are fraught with delays and overages. Professional services where the real “price” is hidden behind a long discovery process is another. With AI, you can make these services transparent, efficient, and offer them for a fixed fee.

In roofing, Remi uses AI to manage contractor matching, permitting, scheduling, and inspection, which allows them to offer one guaranteed price to consumers from the start. Lawhive similarly offers flat-fee consumer legal services with no hourly rates, thanks to AI that partners with human lawyers to streamline work.

This means that a marketplace’s offering is no longer competing against a sea of similar solutions. They can offer something unique to the end-user that cuts through the noise, which (ideally) reduces the cost to attract them, and increases the number of transactions that take place. 

Increasing supplier loyalty and throughput

Suppliers are another crucial part of marketplace health. If they’re loyal to the platform and work to build a business on it, this can drive major growth. Marketplaces can give AI superpowers to their sellers to encourage this behavior. 

In talent, Paraform is a marketplace where independent recruiters refer candidates to fill roles at companies offering bounties. The platform gives recruiters AI tools to scale themselves – a sourcer, CRM, notetaker, and scheduler – allowing them to work with more candidates and across more searches simultaneously. These tools also incentivize them to run these searches on Paraform versus anywhere else. 

In healthcare, Counsel Health has patients talk first with a medical AI to share symptoms and ask basic questions. Then, they connect with a human doctor to discuss treatment or get a prescription. This lowers visit costs for patients, and allows doctors to scale more appointments per day (and hopefully, with less frustration) by focusing only on the higher-value work that AI can’t do.

The New Playbook

AI can fix marketplaces where the problem was operational costs or matching complexity – plugging in an LLM into places humans previously operated can significantly improve unit economics. And AI can transform categories where buyer or seller experiences were particularly poor, opening up opportunities for a marketplace to thrive instead of sputtering along. 

AI won’t fix marketplaces where the fundamental constraint is scarcity, subjective quality, or trust requirements that are impossible to overcome. The highest chance of success is therefore likely to be in categories where the graveyard is filled with scale-ups, not early failures. If companies reached millions in revenue but couldn’t get past $50M ARR, that’t a problem AI can likely address. If they died at $1M ARR, the problem might be deeper and less solvable by AI – for now!

If you’re tackling one of these (or any other) legacy marketplace categories in an AI-native way, reach out to me at omoore@a16z.com. I’d love to hear from you and learn about what you’re building.