Lemkin's Agent Doubled ACV in a Week. HubSpot's Credits Tripled a Bill. The Multiple Is Not the Signal.

Lemkin: 3% of a $30M ARR base paid double for an agent in week one. HubSpot's credits tripled a bill and got a churn call. The unit is the difference.

Falk Gottlob · · 8 min read

Foundation Falkster cover on pink: two cream invoices standing side by side behind a small navy block, the shorter one stamped with an orange checkmark, the taller one curling over at the top with a navy telephone handset leaning against it.

Jason Lemkin published his rule for agent pricing on November 12: a great B2B agent doubles ACV, and most AI features do not move it. This is a Field Notes entry on the two bills inside that post, because they point in the same direction and mean opposite things.

The short version

  • Jason Lemkin's November 12, 2026 SaaStr post reports that a portfolio company at about $30M ARR launched its first real agent, that it doubles ACV with no free version, and that 3% of the customer base bought it in the first week.
  • The same post prices HubSpot's credit model for a three-person team: about $1,300 a month became about $3,700, and a customer called to move the newsletter elsewhere. One bill doubled and one nearly tripled, so the multiple is not the signal. The unit is: a finished job paid from the labor budget, against metered actions that used to be unlimited.
  • His other reference points are Podium at $100M of agent ARR in under 24 months, Box at a 20 to 40% premium, and Salesforce at 60 to 80% with about 5% of knowledge workers on it.
  • I ran the good version: $800 a seat to $1.75 per resolved ticket took one account from about $16,000 to $31,500 a month. Then the base followed, and 14 of 35 mid-market accounts with edge use cases were churning.
  • The 3% who buy in week one are the cohort the agent already fits. Track his three numbers, and add usage on bought accounts and contested jobs per week.

The bill that doubled

The company is not named. It is vertical software, about $30M ARR, growing, with a loyal base and a sales-led motion. This week it launched its first real agent. The agent doubles ACV and there is no free tier. In seven days, 3% of the base bought it, with no discount and no trial.

Lemkin's explanation is the budget it is paid from. Software budgets are capped and CIOs are consolidating vendors to fund AI. An agent that competes for that budget fights its own seat line. An agent that does a job gets paid from the budget for the person, the agency, or the contractor who used to do it, and against those hours twice the subscription looks cheap.

He has made it his investing rule: no agentic B2B company unless it unlocks budget beyond software, with deal sizes at least 2x what pre-AI competitors get. And he gives the range from incumbents. Salesforce charges a 60 to 80% premium for its AI editions with about 5% of sales and service knowledge workers on them as of last quarter. Box's Enterprise Advanced carries a 20 to 40% premium and reached about 10% of revenue in under two years. Podium built $100M in AI agent ARR in under 24 months selling an agent that answers the phone, books the job, and follows up for home services businesses.

His read of that range: improving a workflow earns 20 to 40%. Doing the work earns the double.

The bill that tripled

Third section of the same post. HubSpot's new model, priced by his colleague Amelia for a three-person marketing team with 10,000 contacts, a weekly newsletter, and about 20,000 workflow actions a month. The old bill was about $1,300 a month. The new bill is about $3,700, most of it credits for workflow actions that used to be unlimited. A newsletter send costs the same credit as an outbound sequence.

One HubSpot customer they know called to say the newsletter was moving elsewhere, and got a discount.

Lemkin's summary is that the bill tripled and the customer got no new capability. Fair. But put the two bills next to each other and the interesting thing is how similar they look from finance. Both are a large multiple on last year's invoice. The bigger multiple is the one customers hated.

So the multiple tells you nothing. A CFO looking at "ACV up 2x" on a board slide cannot tell which of these two businesses they are looking at.

What the invoice counts

The difference is the unit. The first bill counts a job that got done. The second counts actions the customer had already been taking for free. I wrote about the second pattern in 41 Brands Price AI Five Ways. The Credit Is the New Seat.: the credit is a private currency the vendor mints, and the customer cannot see the exchange rate.

I have run the first pattern. In Per-Outcome Pricing: What Gets Clearer and What Gets Terrifying I describe moving a support product from $800 a seat to $1.75 per resolved ticket. The customer had been paying about $16,000 a month for what was about $450,000 of equivalent labor. The new price came to $31,500 a month at a 77% margin. Just under double, and the customer was happier, because the bill counted something they could count themselves.

That is Lemkin's story from the inside, and it is true as far as it goes. Week one looked exactly like his.

What week one did not show

Then the rest of the base had to follow. Field Report: What Broke When We Killed Our Per-Seat Tier is the long version. The short one:

14 of 35 mid-market accounts with edge use cases were churning. Their work was not the agent's common case, so the unit that made the first customer happy made them pay for outcomes the agent did not reliably produce.

14 of 35

mid-market accounts with edge use cases were churning when I moved a base to outcome pricing. The 3% who buy in week one are not that cohort.

12 accounts migrated on paper and were barely using the product. They counted as converted on every dashboard.

Disputes ran at 600 a week against 50 projected, plus more than 200 contested cases a week on what one unit meant.

None of that was visible in week one, and none of it was a pricing mistake. It was the difference between the accounts the agent already fit and everyone else.

Which is how I read the 3%. Lemkin says it himself: week-one buyers are the most motivated customers. I would put it more narrowly. They are the accounts whose job is the agent's common case. They bought before anyone sold them because they could see their own work in it. The question for the 2027 plan is what share of the other 97% has that same job, and what share has the edge version.

Five numbers, not three

Lemkin gives three to track after week one: attach rate at 90 days, renewal on the agent line, and core-product retention for agent customers against everyone else. All three are right, and all three are revenue numbers.

Two more, from the cost side.

Usage on the accounts that bought. Not attach. Jobs finished per account per week. Bought and barely using is the quietest churn there is, and it shows up as a surprise at the next renewal.

Contested jobs per week, divided by active accounts. A line paid from the labor budget gets audited the way labor does, by someone asking whether the work was done. I made the general case for this number in Percent Migrated Is a Vanity Metric.

And one split: cut attach at 90 days by whether the account's work is the common case or an edge case. If the second cohort attaches at the same rate as the first, be nervous, not pleased.

This week

Take the agent on your 2027 plan and write down the job it finishes in one sentence. Then go through your top 50 accounts and mark each one common case or edge case for that sentence. The common-case share is your honest ceiling for the double. The rest need a different unit, or a seat. The broader argument, that price moves to the outcome and margin moves with it, is in SaaS to AI Business Models.

Related answer: Can an AI agent really double ACV for a B2B software company?

Sources: Jason Lemkin, "A Great B2B Agent Doubles Your ACV. Most AI Features Don't Move It at All.," SaaStr, November 12, 2026, read in full, citing Ulrik Lehrskov-Schmidt on launching free. The $30M ARR, the doubling, the 3%, the $1,300 and $3,700, the 10,000 contacts and 20,000 workflow actions, the 60 to 80%, 5%, 20 to 40%, 10%, and $100M in under 24 months are his as he reports them; I did not open HubSpot's, Salesforce's, Box's, or Podium's own disclosures. The $800, $1.75, $16,000, $450,000, $31,500, 77%, 14 of 35, 12, 600, 50, and 200 are my own, from the posts linked above.

Share this post

Frequently asked

What did Jason Lemkin report about an agent doubling ACV?+

In his November 12, 2026 SaaStr post, Lemkin says a vertical software company in the SaaStr portfolio, at about $30M ARR with a sales-led motion, launched its first real agent that week. The agent doubles ACV, there is no free version, and 3% of the customer base bought it in the first week with no discount and no trial. He now invests by the rule that an agentic B2B company must unlock budget beyond the software budget, with deal sizes at least 2x pre-AI competitors.

What is the HubSpot example in the same post?+

Lemkin's colleague Amelia priced HubSpot's new credit model for a three-person marketing team with 10,000 contacts, a weekly newsletter, and about 20,000 workflow actions a month. The old bill was about $1,300 a month and the new one about $3,700, mostly credits for workflow actions that used to be unlimited. One HubSpot customer they know called to say the newsletter was moving elsewhere and got a discount.

If both bills rose by a multiple, what is the difference?+

The unit on the invoice. The agent's bill counts a job a person used to do, which the customer compares to the hours it replaces. The credit bill counts actions the customer already had for free. The multiple was larger on the bill customers resented, so the size of the increase is not what decides whether it holds.

What is your own doubling?+

I moved a support product from $800 a seat to $1.75 per resolved ticket. The account went from about $16,000 a month to $31,500, just under double, at a 77% margin, and the customer was happier. When the per-seat tier was later sunset across a base, 14 of 35 mid-market accounts with edge use cases were churning, 12 accounts had migrated on paper and were barely using the product, and disputes ran at 600 a week against 50 projected, plus more than 200 contested cases a week on the unit definition.

What would you track after week one?+

Lemkin's three, which are attach rate at 90 days, renewal on the agent line, and core-product retention for agent customers, plus two from the cost side: usage on the accounts that bought, and contested jobs per week divided by active accounts. Then split attach by whether the account's work is the agent's common case or an edge case.

Jason LemkinSaaStrHubSpotPodiumBoxSalesforceUlrik Lehrskov-SchmidtACV
Helpful?

Keep reading

The short answer

Pricing Migration: The 18-Month Quarterly PlaybookSix quarters, four mandatory CFO/CRO/Board/Lead-customer conversations, the gross-margin recovery curve, and the contract terms that defend outcome billing.Read the chapter

Part of

SaaS to AI Business Models

About the author

Falk Gottlob

Falk Gottlob

Product Executive · Founder, Falkster.AI

Thirty years shipping product, from Microsoft Research and Adobe to Salesforce, where he grew Quip into what became Slack Canvas. Four startup exits, including a $6.5B healthcare merger and a company Microsoft bought. Four-time Chief Product Officer. Now founder of Falkster.AI, building Heidi, the company brain for AI-native businesses. This notebook is written from inside the build, not above it.

Sign in with LinkedIn to leave a comment.

Sign in with LinkedIn