One Vertical, One Use Case, $17.5M – The Focused AI Playbook

01 /The Concept

Toma builds AI agents that automate customer service for automotive dealerships.

At its simplest: the startup created a voice AI agent that answers inbound calls. It handles routine questions on its own, and when a caller's request falls outside its competence, it transfers the call to the right human staff member.

The agent can schedule service appointments – asking the customer what they need, checking technician availability, and adding the visit to the calendar. It can also reschedule appointments when customers call to request a change. And it proactively calls customers to remind them of upcoming visits.

On the parts side, the agent handles inbound requests for spare parts. It can identify the exact part from the customer's description, confirm pricing and stock availability, and – if the customer wants to proceed – place the order automatically and route it to the warehouse or procurement team.

The platform also includes a workflow builder that lets dealership administrators create custom agent scenarios: recognizing specific request types and triggering predefined actions, such as notifying a specific employee or creating a new record in the dealership's CRM.

The results speak for themselves. One Toma customer had the AI agent handle 22,000 calls over 90 days, scheduling 9,000 appointments that generated $2 million in revenue – while cutting the inbound call load on human staff by 40%. Another dealership used the agent to book 180 additional appointments per month simply because the AI could take calls around the clock.

Toma was founded in early 2024, when its founders were going through Y Combinator. Today, more than 100 US dealerships use its AI agent. Shortly after graduating from YC, the startup raised $375,000 in seed funding – and it has just closed a $17 million round led by the respected venture firm a16z.

02 /Why It Matters

In a blog post explaining the investment thesis, a16z laid out why it backed Toma.

The automotive market in the US is enormous – there are almost as many cars as people. The country has 18,000 franchised dealerships selling $1.2 trillion worth of vehicles annually, plus 270 million repair and parts orders generating an additional $156 billion per year.

Phone calls remain the primary customer communication channel for dealerships – and managing that volume is a genuine cost center. Dealer staff are perpetually busy and can’t always pick up. Outsourcing to call centers adds cost without solving the problem: annual staff turnover at those centers exceeds 50%, roughly twice the national average, which means quality suffers.

Toma’s founders spent their earliest days calling every US dealership with their voice AI agent – and the result was simultaneously discouraging and validating: 45% of calls went unanswered.

Beyond the phone, Toma’s ambitions reach toward automating the bulk of dealership operations, so that human employees can focus exclusively on work that only people can do.

The same dynamic plays out elsewhere. More than 400,000 wholesale distributors operate in the US, generating $9 trillion in combined revenue – yet their sales reps spend less than 30% of their time actually selling, because the rest goes to answering basic questions: "What's the price?" and "Is it in stock?" More than 25% of inbound calls to distributors go unanswered.

That gap is exactly what Y Combinator graduate Kanava ([related review](/review/dva-kriterija-vybora-pravilnogo-rynka)) is targeting – a voice AI agent for wholesale companies that handles inbound customer calls and also takes voice commands from sales reps in the field who can't open a laptop between client visits to update the CRM or pull pricing for a quote.

Solo business owners face the same trap: they're either doing the work or picking up the phone – they can't do both.

Y Combinator graduate Cactus ([related review](/review/jeto-samye-prostye-dengi)) built an AI agent to handle those calls on behalf of independent operators. The agent answers questions, qualifies leads, and takes orders. Every call gets logged in a lightweight CRM the owner can review at their leisure, complete with the AI's recommendations on which leads are worth a follow-up call.

Dench ([related review](/review/tebe-nuzhny-tolko-horoshie-klienty)), another recent YC graduate, raised $1 million in February to build an AI receptionist for law firms. It handles client intake calls, asks the right qualifying questions, and – if the case isn't a fit for that firm – can transfer the caller to another firm's AI receptionist that handles that area of law.

And Riviera ([related review](/review/a-takih-svobodnyh-nish-poka-eshhjo-do-figa)), yet another YC graduate, built an AI hotel receptionist that handles calls in multiple languages – answering pricing and availability queries, taking room bookings, processing in-room dining and housekeeping requests, and giving local recommendations.

03 /Opportunities

It's no coincidence that every startup mentioned in this review is a Y Combinator graduate. YC has explicitly identified niche voice AI agents as one of the most promising areas right now – both because vertical AI services are hot, and because voice AI is hot. Their intersection is doubly interesting.

This analysis agrees with that view. Building niche voice AI agents for inbound customer calls is a timely and high-potential direction.

These agents genuinely free up time for both employees and solo operators who have plenty of other things to do. They also serve as a landing pad for deeper automation – which is precisely how Toma itself thinks about its trajectory.

One more pattern worth flagging: if a business's primary customer touchpoint is still the phone, it's almost certainly digitally underserved. That means new technology can deliver outsized efficiency gains – and companies in those niches will pay for it.

So the need for a voice AI agent isn't just a product feature – it's a qualifying signal that a niche is ripe for broader technological transformation.