There Are Still Plenty of Open Niches in AI

01 /The Concept

Hotels worldwide spend $100B annually on staff. A significant share of that labor is consumed by phone calls – guests asking questions, prospective guests inquiring about availability and rates – rather than by direct, in-person service. The result: either hire more staff to cover both simultaneously, or accept that guests will wait on hold while front-desk staff are tied up on the phone.

The problem is compounded by language diversity. Callers speak different languages, which slows conversations, creates misunderstandings, and affects service quality.

Riviera is an AI phone platform built specifically for hotel guest services, capable of handling multilingual calls.

Answering guest questions is the core capability. At minimum, the AI can analyze an incoming call and route it to the right staff member. More valuably, it handles the full range of common inquiries on its own – available amenities, operating hours, local recommendations for restaurants, shops, and attractions. Hotel staff upload property information to the platform; local context is pulled from the web automatically.

Room service is a natural extension. The hotel uploads its restaurant menu, and the AI takes food and beverage orders – offering dietary guidance, making suggestions to upsell the check, and remembering guest preferences across calls within the same stay. When new guests check into a room, the system resets automatically, syncing with the property management system to stay current.

Handling reservation inquiries rounds out the platform. The AI pulls live availability and pricing from the property management system and answers booking questions accurately. During those conversations, it also proactively proposes room upgrades and add-on services to interested callers – increasing average booking value without any human involvement.

All conversations are logged with transcripts, so managers can review call history. The AI also self-assesses the success of each interaction, making it easy to identify the calls that didn't go well and upload additional context so the system handles those cases better next time.

Riviera is currently in Y Combinator and received its initial $500K from the program. The launch was announced on the YC platform two days ago.

02 /Why It Matters

The exact number of hotels worldwide is, oddly, contested – estimates range from 187,000 to 700,000 properties. Either figure represents a large addressable market. Riviera has the opportunity to do well – or very well – depending on where reality sits.

Riviera is a niche product. In a moment when AI is moving across every front simultaneously, it might seem like a small bet – shouldn't ambitious builders be working on horizontal platforms instead?

But horizontal AI platforms are a different game entirely. They require massive capital to build and even more to distribute. The competition is less about product quality and more about fundraising capacity. That's a different skill set – and a much tougher one for most founders.

Vertical AI – AI built purpose-built for specific industries – plays to a completely different dynamic. The competitive advantage in a vertical isn't how much funding you raised; it's how deeply you understand the specific workflows, edge cases, and vocabulary of the domain. And how precisely you can tune the system to the operational specifics of each individual client.

A horizontal AI assistant that handles any phone call reasonably well is not the same product as an AI that fluently manages hotel reservations, understands dietary restrictions in the context of room service, knows how to upsell a suite upgrade at the right moment, and syncs real-time with the property management system. The vertical player wins on specificity.

Y Combinator itself has argued that the vertical AI agent market will be ten times larger than the SaaS market – though they have an obvious incentive to encourage startups to explore every niche. But major venture firms are independently arriving at the same conclusion. Bessemer Venture Partners, for instance, has published explicitly that "the future of AI is in verticals."

03 /Opportunities

If the vertical AI market is going to grow as fast as YC and top-tier VCs predict, the opportunity is to build niche AI platforms under a guiding principle of "the devil is in the details" The distinguishing feature and competitive moat of a vertical AI platform is deep specialization down to the smallest operational nuances – and the ability to fine-tune for each individual client's specific processes.

The more detail and specificity that appears on the platform's website and pitch – real domain knowledge, edge cases handled, integrations built – the more credible it looks to a buyer from that industry.

A few questions point to strong vertical opportunities: Which industry do you know well – or could learn deeply? Where are staff spending meaningful time on structured, repetitive tasks? The answers define the problem. From there, the design question is how much domain-specific tailoring the AI needs to actually resolve problems rather than merely converse about them.

That last point matters more and more. Lorikeet ([related review](/review/nash-produkt-luchshe-jeto-nedostatochno-ubeditelno)) raised $14M for a customer support AI agent whose pitch is explicitly that it "does what no other AI agent can do" – it resolves problems instead of generating responses. The other agents talk; Lorikeet acts.

Problem-resolution capabilities are also vertical-specific by nature – the mechanisms for closing a hotel complaint differ entirely from the mechanisms for resolving a billing dispute at a software company. That specificity is yet another layer of competitive advantage for the vertical AI builder who gets there first.