Zoom's Single AI Spend Pool Makes the Commitment the Number Buyers Negotiate

Zoom and Gainsight took AI off the seat as model suppliers sought commitments

Zoom and Gainsight moved AI onto a commitment the buyer draws down this week, and two more vendors put a meter on it. The model makers underneath them showed why they'd rather sell a year of committed spend than a seat. I think your next AI-heavy renewal turns on three numbers: the size of the pool, how fast AI draws it down and whether the rate is locked.

At Zoomtopia on October 20, Zoom announced new Zoom CX capabilities and changed how its contact-center product gets bought. Customers commit to one annual spend that draws down across agents, seats, applications and AI. They can change the mix without a contract review, Zoom's head of CX marketing told CX Foundation. Zoom hasn't published rates. For a Zoom CX buyer, the renewal numbers become the size of the pool and how fast AI draws it down. The seat price is now one input among several.

The same day in London, Gainsight said Gainsight CS will stop selling per-seat licenses next year. The new price scales with the ARR and the number of customers a company manages through Gainsight, and no rate card is out yet. Its Atlas service pairs agents with renewal managers on long-tail accounts and already manages more than $75 million of customers' ARR. So a CSM seat count stopped describing what Gainsight sells.

Two more vendors moved the unit on October 19. monday.com launched Harmony, a voice agent that calls and qualifies sales leads, at $0.39 a minute of agent time plus a $250 monthly platform fee. Notion rebuilt its AI credits into two monthly allowances inside each Business and Enterprise seat, with no rollover. One covers 1,500 credits on models Notion picks, and the other 500 on models the user picks. Notion’s version is the seat-plus-usage shape we described earlier in October, with the meter living inside the seat.

Software makers expect more of this. In Stripe's survey of 200 SaaS platform leaders, published October 19, respondents expect seat-led pricing to fall to 11% of platforms in 2027 from 16% this year. About one in four plan to keep seats while billing agent work separately. Those are expectations from a sample in which half are Stripe users, as we noted in our hybrid-pricing piece. Salesforce is pushing the meter past its own products. The Information reported on October 22 that it plans to make companies register outside AI agents such as Claude as soon as next month. The step points toward charging each time they run inside its apps. That extends the credit meter Salesforce put on agents reading the CRM to agents Salesforce doesn't sell.

The suppliers under these products showed why they'd rather sell a commitment than a seat. On October 20 the Wall Street Journal reported on a draft of Anthropic's IPO prospectus showing $16.3 billion of first-half revenue and an $80.7 billion net loss. Most of the loss was accounting charges on convertible notes. SpaceX, which owns Cursor and Grok, filed results the same day showing $431 million of operating income on $7.56 billion of revenue in its AI segment. A model maker owes most of its compute whether or not customers use it. A year of committed spend is worth more to it than a higher rate card, which is why 2027 AI renewals will ask buyers to commit. The next day, Google Cloud said OpenAI's GPT-6 models are coming to its Model Garden and Marketplace, where purchases can draw down a customer's existing Google Cloud commitment.

Commitments also get priced one account at a time. On October 24, The Information reported that OpenAI is cutting prices to pull Anthropic's biggest API customers, according to The Decoder's summary. Replit got more than 30% off a contract worth over $100 million a year, per a person familiar with the deal. Rippling's CEO said OpenAI's share of its coding work rose to 52% from 16% in May. Cursor's October 12 repricing ran the same logic inside one product, keeping house models in a $30 seat and moving outside models onto a meter from the first token. The list price is public, and the discount for committing volume is negotiated deal by deal.

Town is the strongest case against all this. It raised a $108 million Series B at a $1 billion valuation on October 19. It says paying teams grew 9x since July on a plan sold at $59 a seat a month (company-stated, like every growth number here except Braze's). A Townie can only hand work to a colleague who has one, so every collaborator becomes a seat. The seat is still selling. Look inside it, though. Town's pricing page gives each Team seat 6,000 credits pooled across the team, with the overage cap set at $0 by default. The seat survives as the way a team commits, and the credits are what it commits to.

Buyers are rationing the other side of the same ledger. Since October 1, Accenture has funded $29 of each employee's monthly Claude use centrally and charged the rest to project codes, 404 Media reported. That made every project owner an AI buyer. Deloitte's CFO Signals survey of 200 finance chiefs at $1 billion-plus companies found 53% review AI costs only after the fact. And 74% need several business days or longer to pause usage that runs over budget. Last week's question was the ceiling on the meter. This week's is the floor a buyer signs for under it.

If you're renewing an AI-heavy product this quarter, here's what I'd get in writing before it turns into a pool. Put the draw-down rate for each product in the order form, along with rollover of unused commitment into the next term. Lock per-unit rates for the full term, so a discount on the commitment can't be clawed back through the meter. Decide who funds usage past the floor, centrally or by project code, before the vendor's console decides for you. If you're the vendor moving to pools, publish the draw-down rates before your sellers quote. Then settle whether reps earn on the committed amount or the consumed one, because the two diverge the first quarter a customer under-uses.

Our bet is that at least two more seat-built GTM platforms replace per-product seat commitments with one annual AI spend pool before the end of June 2027. If none does, Zoom's pool stays a contact-center exception and seats with credits on top remain the default.

Salesforce and investors paid for revenue that was already growing

Salesforce agreed on October 20 to buy Braze for about $5.66 billion in enterprise value, $47.50 a share in cash. Its marketing and commerce products grew 2.8% last fiscal year, the slowest of its five lines. Braze grew revenue 26.2% in its July quarter, so Salesforce is buying growth for its slowest business. Pigment raised $250 million at a $2.3 billion valuation on October 21 after passing $100 million in ARR. The Planner Agent meant to run the cycle is still marked coming soon, so Pigment sells the planning model first. Legora raised $550 million at $8.8 billion the same day and says its second $100 million of ARR took under six months, from the US and in-house legal teams. Daytona raised $60 million at $1 billion on revenue it says grew about 25x since January. It bills by the second for the sandboxes that run agent code inside n8n and Writer.

The exception was Different, which came out of stealth with a $48 million seed led by General Catalyst for AI that runs B2B demand generation. It named two customers and no prices. Its own design keeps a marketer approving every campaign.

Apart from Braze's filings, every growth figure here is company-stated and unaudited. Even so, the week's money went to motions that had already shown revenue. The one bet on a promise was priced on its founder's record at Nutanix and ThoughtSpot.

Anthropic and monday.com put agents on the owner's inbox and the lead's phone

Anthropic opened Claude Duet on October 21 as a waitlisted research preview for Pro and Max subscribers. It checks a small-business owner's connected apps every few hours, sends a morning brief and stops for approval before it pays, signs or sends anything. The launch post names QuickBooks, Stripe, Gusto, Docusign, HubSpot and Shopify among the tools it drives, which puts Anthropic's agent between owners and their software.

monday.com's Harmony calls leads, qualifies them and transfers warm ones to a rep. The voice technology came from OneAI, whose assets cost $14.1 million in cash. The only named result is a UK car-finance broker's 9% rise in closed deals with no holdout, so monday.com sells a voice SDR by the minute on thin proof. Podium said on October 22 that its AI agents passed $150 million in ARR, with more than 20,000 deployed. They're sold as an add-on to plans local businesses already pay for. Sierra added Stripe payments so its service agents can take a payment and send the receipt inside the call.

All four agents take over a first contact, whether it's the owner's morning look at the books, a new lead's first call or a customer's payment. Duet is a preview and Harmony's evidence comes from consumer lending, which is the case against reading too much into either. The vendor that runs the first contact holds the relationship the software underneath used to own. If you sell software to small businesses, start counting how much of your customers' activity now arrives through someone else's agent.

BCG, Gartner and Forrester found AI use in sales and buying ahead of proof

Boston Consulting Group's October 19 survey of 398 sales leaders found 90% report meaningful AI use. Only 13% name win rates as a top payoff, and 65% expect revenue headcount to grow next year. A day earlier, a Salesforce regional vice president said agents working leads nobody had answered produced more than $88 million in pipeline. No close rate was attached, which is the pattern in all four of its lead-agent disclosures. Gartner said the same day that only 13% of 2,002 CIOs report significant or transformational value from AI.

Buyers are checking. Forrester's 2027 predictions, published October 21, say two-thirds of providers will lose deals because buyers can't verify their claims, which turns every unsourced sales claim into a deal risk. In Bain's survey of 207 IT buyers, about 46% use AI most of the time to build their first vendor list, and 32% use it to negotiate price. Forrester's revenue enablement Wave found frontline managers posted the lowest adoption rating of any group, among reference customers using about 75% of what they bought. So the people who run coaching use the platforms least.

I wouldn't trust any one of these numbers on its own, since every one is a survey or a vendor's own count. But five samples pointing the same way are hard to wave off. A buyer's AI now checks a seller's claim before the seller's AI has a close rate to back it. Publish the sourced number, and test your agents against a holdout.

Meta and OpenAI scaled the ad products that creator and chat budgets run through

On Meta's October 21 earnings call, CFO Susan Li said partnership ads grew more than 97% to a $15 billion-plus run rate. Meta's total ad revenue grew 26% to $63.06 billion. Its Creator Marketing Hub now lists more than 10 million creators, which puts ad rights at the center of B2B creator deals. OpenAI said the same day that ChatGPT ads now run in 89 countries and that more than 100,000 advertisers have run campaigns. The ads still show only to Free and Go users.

On October 22, LinkedIn's connected TV ads became buyable in Google's Display & Video 360 with job-title targeting, in the US and through account reps. Scrunch data published the same day found that when an AI answer cites a sponsored YouTube video, the sponsor appears 40% of the time, against 32.3% when it doesn't.

Meta's and OpenAI's advertisers are mostly consumer brands, and ChatGPT ads don't reach business seats yet. The buying path is moving anyway, from a negotiated creator fee to an ad manager. A B2B creator contract should carry paid usage rights with an end date before the next campaign is booked.

Also this week

• Getty Images filed a prepackaged Chapter 11 on October 21, blaming warrant judgments, a failed Shutterstock merger and debt, while AI's damage shows up in its self-serve iStock buyers.
• Mercury's CEO said it passed $1 billion in annualized revenue on accounts that cost customers nothing, and 1Password passed $500 million in ARR, up from $400 million last November.
• LinkedIn's October 15 announcement put private profile notes behind Premium, with no disclosed route into a CRM.
• Adam Schoenfeld counted "GTM context" on 37% of GTM tech homepages in September, and the part that wins deals stays in-house.
• Anthropic and Hg agreed to back a European consultancy that embeds AI engineers in companies, the FT reported, and Uber put all of its roughly 30,000 employees on ChatGPT Work.
• Bending Spoons took $330 million from Berkshire Hathaway on October 14 for more deals, after buying SaaS companies at under 3x ARR.

The week in numbers

$0.39: what monday.com's Harmony charges per minute of agent time.

11%: share of SaaS platforms Stripe's respondents expect to lead with seats in 2027.

30%-plus: the discount Replit reportedly got from OpenAI on a $100 million-plus contract.

13%: sales leaders in BCG's survey naming win rates as a top AI payoff.

6%: Pegasystems' third-quarter ACV growth. On its earnings call it said the full year would land below 10%, against a 15% target.

The list price is public, and the discount for committing volume is negotiated deal by deal.

Before an AI-heavy renewal turns into a pool, put the draw-down rate for each product, rollover of unused commitment, and a full-term lock on per-unit rates in the order form.

By Business of GTM