Five9's AI Revenue Grew 91%, Its Legacy Seat Business Grew 7.5%, and the Stock Fell 17.6%

Five9 beat on revenue and raised the top line, then lost more than a sixth of its market value in the next trading day. Third-quarter revenue reached $323.3 million, up 13% and ahead of the $319.0 million consensus, subscription revenue rose 16%, and non-GAAP earnings of $0.82 a share beat the $0.80 expected. AI revenue hit $45 million, up 91%, and now accounts for 17% of subscription revenue against 10% a year earlier. Adjusted EBITDA was $81.2 million, exactly 25.1% of revenue, the same margin as a year earlier. Underneath that flat EBITDA line, GAAP net income fell 19% to $14.5 million from $18.0 million on 13% revenue growth.

Guidance moved in two opposite directions on the same release. Full-year revenue went up $6 million at the midpoint and the 2026 AI revenue growth outlook was lifted 15 points, while the profit guide was cut 2.1% at the midpoint on higher revenue. The fourth-quarter revenue range of $328–$334 million implies 10.3% growth against the third quarter's 13.1%, so the acceleration reversed a quarter after it started. Five9 closed Nov. 5 at $34.14 and fell 17.6% the next day to $28.12, with about $369 million of market value erased in one session of trading after the print.

Five9's subscription is priced on licenses, defined in its 10-K as the maximum number of named agents allowed to reach the platform at once, with customers typically running more named agents than they hold licenses. AI solutions are charged separately on a consumption or capacity basis, and bundled plans charge one fixed monthly fee per agent seat that includes unlimited usage. That structure is what makes agent-driven deflation a revenue event and not merely a product cycle: every interaction an AI agent absorbs removes a concurrency slot from the license count.

Five9 signed a merger agreement on Nov. 5 with Ujwal, Inc., which operates as Level AI, for $325.0 million in cash subject to customary adjustments, with no termination fee and an outside date of Feb. 5, 2027. Management put standalone ARR at roughly $20 million as of Oct. 31, 2026, growing about 50% year over year, and framed the price as about 11 times next-twelve-month ARR, which implies roughly $29.5 million of ARR a year out. On today's ARR the price is about 16 times. Level AI brings AI agents, agent assist, quality management, coaching and Voice of the Customer on a single conversation-intelligence layer, handles more than a billion customer interactions a year, and ships its own Level AI Latitude model family positioned for lower cost and latency than frontier models. Five9 prices agent software at about 16 times revenue.

The gap is the whole argument. Five9's shareholders now pay about 1.7 times revenue for the entire company, while the company itself bought agent revenue at about nine times the multiple of its own revenue base. The cash is affordable: Five9 held $697.7 million of cash and short-term investments at Sept. 30 against $807.9 million of total debt, so the price consumes less than half its liquidity without new borrowing. The earnings cost is not trivial. Management expects the deal to dilute adjusted EBITDA margin by about one point in 2027, roughly $14 million of EBITDA against the $1.39 billion of consensus revenue it cited as a starting point, with Level AI standalone reaching EBITDA breakeven in about six months and margin accretion in 2028. Purchase accounting has not landed, since goodwill was unchanged at $366.3 million and net intangibles had fallen to $41.0 million at Sept. 30 because the merger had not closed, and no guidance includes Level AI at all.

The unbundling is happening one layer above Five9. Salesforce now sells Agentforce at $2 a conversation, at $500 per 100,000 Flex Credits, roughly $0.10 an action, and at a $5-per-user-month license that requires credits on top. The seat has become a pass-through for consumption and the vendor gets paid on work handled rather than on headcount licensed. Salesforce's Agentforce ARR of more than $1.5 billion is about eight times Five9's annualised AI revenue of roughly $180 million, and it grew 240% against Five9's 91%. Five9 is not the only one repricing; it is the one being repriced by, because it prices the agent by the conversation while Five9 still prices the seat.

Twilio is the control experiment, even though it also sells agent seats in Flex and Engage. Its revenue is billed on usage, and it raised full-year organic growth to 15% from 13–13.5% while generating $369.6 million of quarterly free cash flow at a 24% margin and 114% dollar-based net expansion. NICE is the warning: it carries 2.2 times Five9's AI scale, with AI ARR of $392 million against Five9's roughly $180 million annualised, yet total revenue grew only 8.0%, and it is cutting about 5% of its global workforce with $15–20 million of restructuring charges booked in the fourth quarter. More AI revenue has not yet bought faster total growth in this category, because the agent transition is running on monetization rather than on growth rates.

In the next three months the questions are narrow and checkable. The Level AI merger has to close inside the fourth quarter and before the Feb. 5, 2027 outside date, and with no termination fee a failed deal costs Five9 little beyond confidence in its 2027 plan. The February 2027 full-year report has to answer three things: whether adjusted gross margin moves back above 62% as management expects, whether trailing subscription net retention holds at or above 109%, and whether fourth-quarter AI revenue exceeds the third quarter's $45 million without the margin benefit disappearing with it. Capital allocation is the fourth lever: Five9 repurchased $90.0 million of stock under an accelerated agreement in May 2026 and nothing in the third quarter, and $200 million of the 2026 authorization was still unused with the shares 30% below their $40 52-week high. Three checkable numbers will settle whether the seat-deflation read is right or wrong.

Over one to three years the thesis is simple. If AI agents take over interactions, Five9's license count falls and its revenue has to come from capacity-based AI pricing and gross telephony minutes, both of which dilute gross margin because one is consumption-priced and the other is a pass-through. The whole model then depends on AI revenue per licensed seat rising fast enough to offset shrinking seats, and the data so far says it does: AI revenue of $45 million against a legacy subscription base growing 7.5% still produced 16% subscription growth, 109% retention and a 25.1% adjusted EBITDA margin. The risk is that the mix flips faster than the math absorbs, and the single metric that would turn this from a margin story into a revenue story is retention rolling from 109% toward 100%.

Fact: revenue grew 13%, AI revenue grew 91%, adjusted gross margin fell 110 basis points, GAAP net income fell 19%, full-year revenue guidance rose $6 million, full-year profit guidance fell 2.1%, and the equity lost 17.6% in one trading day. Inference: the market is disputing the price of demand rather than the volume of demand, because agent revenue arrives on lower-margin terms than the seats it displaces, and because the private price tag on agent software, $325 million for $20 million of ARR, is about sixteen times what the public market now pays for all of Five9's revenue. The counter-evidence is concrete too: retention improved to 109%, management expects adjusted gross margin to recover in the fourth quarter, and Five9's quarterly free cash flow of $53.3 million still grew year over year, so the deflation is a forecast rather than an observation.