The Voice-AI Roll-Up Is Starting

I have been keeping a small spreadsheet on my desk since December called voice-AI funding, restaurant-adjacent. It had four rows on Friday. After this morning’s news cycle pulled all three of them onto the same page of the Monday brief, it is time to write the column the spreadsheet has been pointing at for eleven weeks.

The thesis: the voice-AI category just split in two. The split is not about which models the vendors run, which TTS stack they license, or whose API is underneath. It is about TAM. One tier — call it the platform tier — sells voice agents to any vertical that answers a phone. The other tier — the vertical tier — sells voice agents to restaurants only. The platform tier just printed an $86 million round at $200M-plus in total capital raised. The vertical tier printed two rounds in eleven weeks. The platform tier has a clean acquirer list. The vertical tier does not. (Interpretation flag: the split framing is mine, not the vendors’. None of the four companies below would describe themselves as one tier or the other on the record.)

The three rounds, in order

Walk the prints in sequence. On December 15, PolyAI announced what is, by some distance, the largest restaurant-voice-adjacent round of the cycle. Per the Restaurant Business coverage of the broader category, the company “has raised $86 million in Series D funding, co-led by Georgian, Hedosophia, and Khosla Ventures,” and “PolyAI has now surpassed $200 million in total funding.” PolyAI is not a restaurant company. It sells to hotels, retailers, healthcare networks, banks, and — yes — hospitality groups. The restaurant book is one column in a wider deal.

Six weeks later, on January 21, Presto Automation closed a $10 million round led by Metropolitan Partners Group with participation from ElevenLabs CEO Mati Staniszewski personally. Per the Restaurant Technology News writeup, the money funds drive-thru voice deployments at the QSR end of the market. Presto is a restaurant company. The TAM disclosure is in the headline: drive-thru, U.S. quick-service.

Two weeks ago, on February 24, Slang AI announced a $36 million Series B — $28M equity, $8M debt — bringing total funding to $68M and the deployed footprint to 2,000-plus restaurant locations. Slang is also a restaurant company. The wedge is the host-stand phone.

Add the fourth row to the spreadsheet for completeness: Vox AI closed an $8.7M seed in August. Same vertical. Same wedge.

Three rounds, eleven weeks, four companies. The line through the four is not “voice AI is hot.” The line is that the platform tier is consolidating capital at a different velocity than the vertical tier, and the platform tier’s acquirers are already implied by the cap-table composition.

Platform versus vertical: the only chart that matters

Sort the four companies on a single axis — how much of their addressable market is restaurants — and the tiers fall out clean.

Platform tier. PolyAI, Retell, SoundHound. Restaurants are a vertical inside a larger commercial book. PolyAI’s hotel and bank work is the cross-sell that funds the restaurant work. SoundHound’s automotive book — which I wrote about in January from the CES floor — is the cross-sell that funds the Toast partnership. Retell sells horizontally to anyone with a phone tree. The platform-tier acquirers are obvious: enterprise CCaaS companies (Five9, NICE, Genesys), large platforms with horizontal voice ambitions (Microsoft, Salesforce, ServiceNow), and the model labs themselves if they decide voice is a moat layer.

Vertical tier. Slang, Presto, Vox. Restaurants are the only book. The product surface is the restaurant phone or the drive-thru speaker. The deployment language, the case studies, the sales motion, the integration partners — Toast, Resy, OpenTable, Olo — all run through hospitality channels. The vertical-tier acquirers are a much shorter list: Toast, Olo, Yum’s Byte platform, DoorDash, possibly Square, possibly a private-equity roll-up of restaurant-tech assets. That is six or seven names, most of whom already have a voice strategy.

This is the part of the analysis that the press releases will not write. The vertical tier raised the wedge. They have not raised the exit. A $36M Series B at the Slang stage builds a real company and gets the operator footprint to four thousand or five thousand locations. It does not, on its own, justify a billion-dollar acquisition by Toast — because Toast can build the same product, partner with PolyAI, or wait for the vertical-tier multiples to compress. The vertical companies have to either consolidate (Slang + Vox + a third) or platform-up (sell into hotels, retail call centres, anywhere a phone rings) to widen the exit list. The next eighteen months will tell us which.

What restaurants should care about

Three things, in order of operator impact.

One: the platform-tier roll-up is the one that affects you, not the vertical one. If PolyAI is acquired by a CCaaS company or a hyperscaler, your voice agent’s roadmap changes overnight. The pricing changes. The integration partners change. The data-flow contract may change. If Slang is acquired by Toast, the integration tightens and the pricing probably gets bundled into a Toast SKU — disruptive but legible. The platform-tier acquisition is the one to track on your weekly vendor-risk log.

Two: vertical-tier vendors will start selling into adjacent verticals this year. Watch the Slang and Presto websites for “now serving hotels” or “now serving healthcare” pages over the next two quarters. The economic pressure to widen the TAM is structural, not optional. When the case-study page changes, the company is repositioning for the next round and a wider acquirer list. That is the operator-visible signal that the vertical tier is reading the same chart I am.

Three: the platform-tier vendors are quietly winning the multi-property accounts. PolyAI’s hospitality book — hotels, casinos, and the larger restaurant groups — is where the average contract size lives. The vertical-tier vendors are stronger at independent and small-chain accounts because the integration and onboarding are tuned for that segment. If you are a multi-unit operator over fifty locations, the platform vendors will quote you better unit economics than the vertical ones inside twelve months. (Interpretation flag: this is my read of the deployment pattern, not a vendor disclosure.)

What I want to know before April

Three reporting threads I am pulling this month.

Whether any of the platform-tier vendors files a confidential S-1 or signals a strategic process. PolyAI at $200M-plus total raised and a Series D is the right shape for either path; the next move will tell us which.

Whether Slang or Presto announces a non-restaurant vertical. I would bet on hotels first, healthcare second, retail third. The case-study page is where the announcement leaks.

Whether Toast, Olo, or Byte moves on a vertical-tier acquisition before the vertical tier widens its TAM. The window for a Toast-buys-Slang trade at the current multiple closes the day Slang lists a hotel customer. Toast’s M&A team knows this. The AssemblyAI voice-AI-in-2026 deep dive makes the same point in horizontal vocabulary; the restaurant version of the trade is sitting right there.

For the broader voice-AI roadmap inside hospitality, the framework Eitan and I have been pointing operators at is the forthcoming Voice Agent Maturity Curve; for the M&A side of the same column, see the March hospitality M&A roundup. Both pieces sit upstream of this one.

The vertical players are good companies. The vertical wedge is a real wedge. The exit math is the part the funding announcements will not write. Watch the case-study pages. Watch the cap tables. Watch who buys whom by July.

— Maya covers restaurant tech for TableTransfers. Tips: [email protected].