Sales SaaS 大变局：ZoomInfo 与 HubSpot 暴跌一周对 AI 销售智能体意味着什么？
In a four-day stretch this past week, two of the most established names in sales software had the worst trading days in their public histories. HubSpot fell roughly 19 percent on May 8. Then on May 11 and 12, ZoomInfo dropped close to 30 percent, capping a 35 percent slide that started after the bell on Sunday night. Both companies beat earnings. Both got punished anyway.
If you've been following the "SaaSpocalypse" headlines for the past 18 months, none of this feels random. What's happening underneath the price action is a repricing of the entire sales software stack — and the question worth asking isn't whether ZoomInfo and HubSpot will recover (they probably will), but where the value is moving to, and why.
This is a read on the week, the underlying shift, and the practical opportunity it opens for AI sales agents — including, yes, the one we're building.
The week that broke the sales SaaS playbook
The two prints, side by side, tell a story that earnings releases on their own don't.
HubSpot, May 8: Q1 revenue up 23.4 percent year over year to $881 million. Adjusted EPS of $2.73, up 53 percent. Forward guidance of $3.00–$3.02 in adjusted EPS, above the $2.86 consensus. By every conventional yardstick, a beat-and-raise quarter. Shares closed down 19 percent at $197.35, intraday low $180.50. Bank of America issued a rare double-downgrade — Buy straight to Underperform — and the company hit a 52-week low. The reason: management acknowledged a slower start to Q2 as it "tweaks" its agentic AI offerings, having dropped pricing on its Customer Agent and Prospecting Agent in April and added 28-day free trials.
ZoomInfo, May 11–12: Q1 revenue up 1.5 percent to $310.2 million, adjusted EPS of $0.28 — both above consensus. Then the company cut full-year 2026 guidance to $1.185–$1.205 billion (down from $1.247–$1.267 billion), guided Q2 below the Street, announced a 600-person reduction in force (~20 percent of headcount), and reported that upmarket net revenue retention had fallen into the low 90s after running near 100. BTIG and others downgraded the stock. By Tuesday afternoon shares were down roughly 33 percent. One headline summarized it neatly: "ZoomInfo beat earnings, cut 600 jobs, and lost 29 percent of its stock price. Its database is being repriced by AI."
That last sentence is the real story. Neither of these is a failure to execute. ZoomInfo's database is still the largest of its kind. HubSpot's CRM is still the gold standard for SMB. Both are doing what good operators are supposed to do — ship AI features, cut costs, guide conservatively. The market is telling them, in unusually direct language, that it doesn't matter. The category they own is the one being repriced.
What 'AI-driven buying pause' actually means
The phrase showed up in three different analyst notes this week. It's worth unpacking, because it's both more and less dramatic than it sounds.
Buyers aren't refusing to spend money on sales software. They're spending more carefully, and the questions they're asking before signing a renewal have changed. A year ago, a sales ops lead renewing ZoomInfo would ask, "are we getting enough use out of our seats?" Today they're asking, "in twelve months, will I be building this dataset from scratch with AI agents that crawl LinkedIn, our inbox, and our calls, and will I still need this contract?" The honest answer, for an increasing slice of mid-market buyers, is no.
The numbers around vendor consolidation are striking. 68 percent of tech leaders plan to consolidate vendors in 2026, targeting roughly 20 percent fewer providers. Drake Star's Q1 MSP market report tracks software M&A running at record pace, driven by what it calls "premium-vs-commodity" gap widening. And IT budget growth is decelerating to 3.4 percent — money is reallocating from application software into AI infrastructure. A "Great Repricing" essay that's been circulating estimates more than $1 trillion of SaaS market cap has been wiped since the AI cycle began.
The same money that used to fund 15 separate sales tools is funding two or three AI-native consolidators. Lightfield, an AI-native CRM founded by the team behind Tome, raised $81 million at a $300 million valuation and gained 2,500 customers in its first three months — including over 100 Y Combinator companies — many migrating off HubSpot. Monaco launched in February with $35 million from Founders Fund and the Collison brothers, bundling AI-native CRM, a prospect database, and supervised AI agents into one product. Aurasell raised a $30 million seed from Menlo and Unusual with the explicit pitch of "replace 15+ GTM tools." Reevo raised $80 million from Khosla and Kleiner to rebuild the entire revenue stack — prospecting, outreach, pipeline, forecasting, customer success — on a single platform.
None of these companies are profitable. Some of them won't be in five years. But they're all telling the same story to the same buyer: "you can stop subscribing to your CRM and your data provider and your dialer and your enrichment tool and your sales engagement platform. Buy one thing. Pay for outcomes."
For sales leaders: Build a one-page audit of every tool in your stack with two columns — what we pay per month, and what specific outcome it produces (meetings booked, deals closed, pipeline created). Anything where the cost outweighs a clearly attributable outcome is a 2026 renewal you should be ready to walk away from. The leverage you have in those conversations is unusual right now; use it.
The value is moving up the stack
Here is the clearest way to frame what's actually happening: in the sales software stack of 2020, the value lived in access. Access to contacts (ZoomInfo). Access to a workflow system (Salesforce, HubSpot). Access to email automation, dialers, enrichment. You paid per seat for the right to use the thing.
In the stack of 2026, the value is moving to outcomes. Not the right to use a tool — but the result the tool was supposed to produce. A booked meeting. A clean CRM record. A drafted follow-up email that actually reflects what was said on the call. A renewal saved because the AI noticed three signals two weeks before the human did.
That's a fundamental change in unit economics, and it's why per-seat pricing is the casualty. If the AI agent does the work, why do I pay for seats? The buyer's question is no longer "how much per rep per month?" It's "how much per closed deal — or per qualified pipeline meeting — or per saved renewal?" Whoever can credibly answer that wins the next decade.
This also explains why HubSpot's price cut on its agentic products caused the stock to drop. Investors weren't worried that HubSpot couldn't ship AI. They were worried about exactly the opposite: that even when HubSpot ships great AI, it has to give it away cheaper than the legacy seat license it's cannibalizing — and the unit economics of the whole company change in the process. ZoomInfo's problem is structurally similar: even if it pivots to outcome-based pricing for AI features, the floor under its database-licensing revenue is dropping faster than the ceiling on AI revenue can rise.
Where AI sales agents fit in the new stack
If the value is moving up the stack toward outcomes, the most defensible place to operate is at the data source — the actual sales conversation. That's not a coincidence; it's where the richest information about a deal lives, and it's the workflow no contact database or seat-based CRM has been able to capture cleanly.
This is the bet that Laxis has been making for several years, and the bet that's looking more obvious by the week. The pattern that wins isn't "another point product." It's a small set of capabilities, tightly integrated:
Sit at the conversation, not the database. The call, the meeting, the screen-share — that's where intent and detail live. Laxis records, transcribes, and structures every call into action items, deal signals, and follow-up drafts. Static contact data, by contrast, ages the moment it's captured.
Integrate with what's left of the legacy stack. The CRM isn't going away in 2026; it's becoming the system of record that AI agents write into. Laxis pushes call summaries, deal updates, action items, and follow-up emails directly into HubSpot or Salesforce on the Premium plan ($15.99/month). The rep stops typing notes; the CRM stays clean as a byproduct. That's worth more than any contact database license.
Charge for outcomes, not seats. The reason the free plan exists with 300 minutes per month isn't generosity — it's that a 300-minute trial is enough for a rep to experience the outcome (a CRM that updates itself, follow-ups that send themselves) and decide that $15.99 a month is trivial compared to the time it returns.
Compress the stack, don't extend it. Laxis already bundles meeting intelligence, AI writer for follow-ups, CRM automation, and B2B lead generation (with 150M+ contacts) into one product. The same buyer who used to pay for ZoomInfo plus Salesforce Sales Cloud plus a follow-up tool plus a coaching tool can plausibly do most of that work in one place.
None of this is a victory lap. AI-native CRMs like Lightfield and Reevo, and conversation platforms like Gong and Chorus, are all running at the same opportunity. The point isn't that any one company wins — it's that the shape of the opportunity is now visible, and the companies still selling per-seat access to a static dataset are the ones the market is repricing.
For SaaS founders: Assume every renewal conversation in 2026 starts with some version of "can we get a 20 percent discount, or are we going to AI-native?" The defense isn't a price match — it's outcome-based pricing tied to something the buyer can verify (meetings, deals, retained accounts). If you can't quote your product that way, your renewal team is fighting with a knife.
What to do in the next 90 days
If you sit anywhere in the sales-software ecosystem — buying it, selling it, building it, or investing in it — the next quarter is unusually consequential. Three practical moves:
If you're a sales leader, renegotiate every contract that comes up before September. The leverage you have in 2026 won't last forever; ZoomInfo, HubSpot, Outreach, Salesloft, and every adjacent vendor knows what's happening to the comps and will trade discounts for term length. At the same time, run a small AI-agent pilot on the highest-friction workflow you have — for most teams, that's CRM hygiene and post-call follow-up. The pilot is cheaper than the discount conversation, and the data it produces is what gets you to the right answer on the renewal.
If you're a SaaS operator, stop shipping AI as a feature and start shipping it as the product. The pattern of bolting an "AI Copilot" onto a seat-based product is what BofA was downgrading HubSpot for — the unit economics get worse, not better. The companies pulling ahead are repricing themselves: per outcome, per workflow, per agent-run. That's a hard repackaging exercise, but the alternative is a stock chart that looks like the past two weeks.
If you're an investor, the cheapest exposure to the AI sales agent wave isn't always the AI-native startup at a triple-digit forward multiple. Sometimes it's a legacy vendor that successfully reprices itself — and trades at 4x revenue while it does. But the underwriting question is execution risk, not vision; ZoomInfo and HubSpot both have the vision, and both still got cut. Look for management teams that are already shipping outcome-priced products, not the ones still talking about them on earnings calls.
See what an AI sales agent does to your sales workflow.
Laxis turns every call into a CRM update, a follow-up email, and an action item — automatically. Free plan includes 300 minutes per month.
The bottom line
ZoomInfo and HubSpot are not going to disappear. They have customers, cash, talented operators, and the kind of distribution that takes a decade to build. The story of the past week isn't about whether they survive — it's about the shape of the sales stack that's emerging around them.
That stack is smaller. It's priced on outcomes, not seats. It sits at the conversation and the workflow, not the contact record. It does the work that humans used to file expense reports about doing — updating the CRM, sending the follow-up, surfacing the at-risk renewal — and asks to be paid for the result, not the right to log in.
The companies that figure that out fastest, whether they're new or old, win the next decade. The ones that don't get a Wall Street note every few months explaining, again, why a beat-and-raise quarter wasn't enough.
