Market and trends

4 min read · Chapter 6 of 10

Who holds the power

Power follows whoever owns the customer relationship, the model and the phone line.

Contact center incumbents (NICE, Genesys, Five9) own routing, the human agents' desktops and the handoff, and their AI revenue already exceeds most startups' total revenue. Their risk is that every contained call is a seat they no longer sell, as Five9's filing says.
Finished-agent vendors (Sierra, Decagon, Parloa, PolyAI, Fin) own the outcome definition. Investors value them highest, at roughly 45-105 times recurring revenue by my rough arithmetic, and CRMs are buying in: Salesforce is buying Fin. Sierra also wrote τ-Voice, the benchmark the industry cites, which I'd keep in mind when reading it.
Labs set the ceiling on task success and the price of the premium tier, and they're moving up the stack with full-duplex voice, direct phone connections and talent deals. They earn through every layer whoever wins the customer.
Speech model companies face commodity pricing on the basic tier and still earn a premium for expressive voices: on Retell, ElevenLabs voices cost 2.7-6.7 times the commodity set.
Communications platforms and carriers hold numbers, porting, attestation and the audio path. Telephony is only 5-12% of a mid-priced stack, but it's sticky, and Twilio, Telnyx and Bandwidth all sell their own agent layers.
Plaintiffs' lawyers and the carriers' analytics engines hold the most practical power over outbound calls. The lawyers price non-compliance and the analytics engines decide whether calls get answered, and both matter more as the FCC loosens its TCPA rules.
Developer platforms (Vapi, Retell, Bland) own the builder and are squeezed from below by model companies, from the side by telephony providers and from above by labs and contact center suites. Vapi's answer is enterprise work; Bland's is all-in pricing.

Tailwinds and headwinds

These are the forces I'd expect everyone in voice AI to watch over the next couple of years. Each one helps some players and hurts others. The two Gartner figures come from press coverage, because I couldn't read the surveys themselves.

Tailwinds:

What's changing	Who gains	Who loses
Service leaders are paying for AI out of budgets that barely grow, and a 2026 Gartner survey put the rise in their AI spending at 38%.	Finished agents (Sierra, Decagon, Parloa, PolyAI, Fin); contact center vendors that sell AI by usage (NICE, Genesys, Amazon Connect)	Outsourcers such as Teleperformance and Concentrix, whose revenue is shrinking; seat-priced contact center revenue, as Five9's filing says
A minute of voice AI keeps getting cheaper, and OpenAI's hybrid voice layer launched in September 2026 at $0.05 a minute.	Buyers; OpenAI and Google, whose voice layers set the floor; finished agents, whose per-outcome prices hold while their costs fall	Speech specialists on commodity tiers (Deepgram, AssemblyAI, Cartesia); offshore outsourcing, whose price advantage narrows
Contact center vendors and clouds sell voice AI inside contracts and cloud commitments buyers already have, and NICE and Genesys each report more than $350 million a year of AI recurring revenue.	NICE, Genesys, Amazon, Salesforce; finished agents listed on the buyer's cloud marketplace (Decagon, Sierra)	Developer platforms in enterprise deals (Vapi, Retell, Bland); vendors missing from the buyer's marketplace
Every AI call still rides a phone number, and Twilio's revenue grew 22% in the quarter to June 2026 with AI-native companies among its fastest-growing customers.	Twilio, Bandwidth, Telnyx; LiveKit; Hiya and other branded-calling services	Vonage, shrinking inside Ericsson; buyers, who pay the carrier surcharges passed through
Press reports put voice AI startup funding at about $7 billion in the first quarter of 2026 alone, so a few leaders can afford to price for market share.	ElevenLabs, Sierra, Decagon, Parloa; the companies buying their smaller rivals	Smaller platforms and speech companies without a new round, such as AssemblyAI; buyers later, when the subsidies end

Headwinds:

What's changing	Who gains	Who loses
Voice agents still fail too many tasks, and in a 2026 Gartner survey 87% of customers said a company using AI for service must offer a way to reach a person.	Vendors that keep a person in the loop (NICE, Genesys, Five9); vendors with published resolution data, such as PolyAI	Full-automation pitches; platforms whose customers leave after failed pilots (Vapi, Retell, Bland); per-outcome vendors that carry the failures (Sierra, Decagon, Fin)
Labs, clouds and telephony providers all sell a complete agent, which squeezes a platform fee that's still about 45% of a mid-priced stack.	OpenAI, Google, Amazon, Telnyx, Twilio; open-source LiveKit and Pipecat	Vapi, Retell, Bland; commodity speech (Deepgram, AssemblyAI, Cartesia)
Spain now bans customer service run only by AI at large companies and essential services, with compliance due by about the end of 2026.	Contact center vendors with human routing (NICE, Genesys, Five9, Amazon Connect); outsourcers	Full-automation business cases in the sectors it covers; AI receptionists sold to them; per-outcome vendors whose margin needs high containment
TCPA class actions hit a record in 2025 and rose another 20% in early 2026, and wiretap and voice-training suits now reach the vendors.	Inbound-focused vendors; consent and caller-ID services such as Hiya; vendors that keep no call audio by default	Outbound platforms and their customers in lead generation, mortgages and insurance; OpenAI and Twilio if courts treat them as callers; ElevenLabs over training data
Some of today's low prices are introductory, and Google's Gemini Flash speech models double in price in January 2027.	Outsourcers and offshore staff; vendors running their own infrastructure (Telnyx, Google); buyers who locked in multi-year rates	Month-to-month buyers; platforms that resell models at cost (Vapi); per-outcome vendors on fixed prices

The one thing I'd watch is the first independent test of chained and hybrid agents on the same tasks and phone audio, because task success sets containment, and containment sets almost everything else here.