The CS Café

The CS Café

Before you forecast that renewal, count their headcount first

4 in 10 CSM roles talk about AI. Fewer than 1 in 50 own what it does to your seat count.

Hakan Ozturk | The CS Café's avatar
Hakan Ozturk | The CS Café
Oct 04, 2026
∙ Paid

The CSM Hired to Shrink the Customer’s Team

Your customer’s best quarter can be your worst renewal.

Hiver is hiring a Senior CSM in the Bay Area. The job is to help support leaders automate workflows with AI. From the customer’s side, that means fewer agents doing the same work.

Now read it from yours.

If that customer pays per agent seat, every automation win shows up at renewal as a smaller contract. Nobody churns. The invoice just shrinks.

I checked how common this is.

Across 1,681 Customer Success listings on TopCSJobs (January to September 2026), 40.7% of Q3 roles used AI or automation language.

Fewer than 1 in 50 gave the CSM any responsibility for what that automation does to the customer’s team.

The vendor sells the shrink. Nobody in the account owns it.

The Seat Math Nobody Puts in the Forecast

Take a support platform priced per agent.

The customer runs 120 agents on a three-year deal. Their AI agent starts resolving a third of incoming tickets.

They lay off nobody. They stop backfilling. Attrition does the rest. By renewal, the team is 80 agents. The customer renews happily, at 80 seats.

That is a 33% contraction with zero conflict.

No one threatened to churn. No one opened a negotiation. Procurement simply counts heads and sends the number.

Meanwhile, your forecast still says 120. The renewal sat in the commit column all year. The contraction only appears the week the new seat count arrives, and by then the quarter is already gone.

This is what makes AI seat compression different from classic churn.

Every signal you track looks healthy. Output per agent goes up. The champion hits their efficiency target. The account looks like a success story right up to the invoice.

It is also a different risk from the customer who replaces you with an in-house tool, which I covered in The Churn You Are Calling Budget Is Often a Build.

Here, the customer stays and renews with you.

They just need fewer seats.

What 1,681 CS Job Listings Show

The AI talk lives in the company pitch, and the job itself rarely touches it.

Company descriptions, benefits and legal text made up 44% of the listing text I analyzed. That is where most of the automation language sat.

The few roles that do assign it to the CSM are explicit:

  • An Intercom CSM role measures success by getting support teams to top automation rates with its AI agent.

  • A ServiceNow role posted earlier this year made the CSM accountable for getting licensed AI agents deployed and producing outcomes.

  • A ThoughtSpot role posted earlier this year had the CSM help Fortune 5000 customers deploy AI agents.

The second signal is pricing.

The share of CSM listings that make consumption part of the job nearly doubled this year. Explicit seat language barely registers.

Q1 2026 → 6.0% of 235 CS listings made consumption part of the role. Seat language appeared nowhere.

Q2 2026 → 8.2% of 692 listings. Seat language appeared in 1.0%.

Q3 2026 → 10.9% of 754 listings. Seat language appeared in 0.5%.

The shift holds when Salesforce is removed (6.0% to 11.6%) and within the same 53 companies hiring in both Q1 and Q3 (7.9% to 13.5%).

Some vendors now write both models into one job.

Anthropic CSM roles ask the CSM to manage underutilization across consumption-based API products and seat-based products in the same book.

Put the two signals together.

The vendors selling automation are moving their own CS teams toward consumption.

Most seat-based vendors still define the CSM role without mentioning the number that contracts.


How I measured this: the data comes from TopCSJobs, the CS job board I run, which pulls listings directly from employers' hiring systems. The sample covers 1,681 English-language CS listings from January to September 2026, with reposts and company boilerplate removed. Each figure was checked by hand on a random sample of matches. The customer-automation figure is an upper bound, and the consumption trend holds within the 53 companies hiring in both Q1 and Q3.


Three Places the Blind Spot Hides

  1. The forecast

    Renewal forecasts roll the current seat count forward. Contraction gets entered by hand once procurement sends a number.

    By then it is a fact, and your options are gone.

  2. The QBR

    Your customer presents their automation win in your business review. Ticket deflection is up. Handle time is down. You celebrate it with them.

    You are applauding the preview of your own downsell.

  3. The champion

    The person who knows the headcount plan is often the person whose team is shrinking. They have every reason to keep it quiet. Some are protecting their team. Others are protecting their own role.

So the question that matters is simple.

Which accounts in your book will renew with fewer seats, and how many months do you have to change the conversation?

Most CS teams find out at the invoice.


The accounts that will renew smaller are already in your book.

Below, paid members find them this week and settle the seat conversation four months before renewal.

What changes this week:

  • Monday morning → you know which accounts will shrink at renewal, ranked by exposure.

  • Your next QBR → their automation win opens a pricing conversation you lead.

  • Your next forecast call → you bring a seat number your VP can defend, before procurement sends theirs.

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2026 Hakan Ozturk · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture