AI-First Customer Success Is an Oxymoron
I’ve recently been skimming my LinkedIn feed and noticing a recurring description in people’s titles: “AI-first Customer Success” Leader, Consultant, Expert, etc. It’s one of those phrases meant to sound relevant, but it makes me laugh how contradictory it is.
If you’re AI-first, you are not customer-first, and I’d like to think, even if only a few dozen people agree with me, that customer success should be customer-first.
Even in wordplay, the name customer success starts with the word “customer.” Put “AI-first” in front of it, and you have told everyone the customer is now second.
I have argued “Customer First, Technology Last” since 2019, and I have written it twice in the last two years: there can only be one “First,” and customer first always wins. I’ve even added a final qualifier: “Customer First, Technology Last, AI Maybe,” because in most cases, organizations implement AI to say they have AI, when regular technology is enough to do what they need. Whatever is first decides what gets funded, measured, and cut. “AI-first customer success” simply diverts focus away from the customer.
Read the definition the industry wrote
The industry has more than one definition of “AI-first customer success.” Take one. ChurnZero’s write-up of its own conference panel describes a growth strategy led by systems rather than hiring, with AI automating tasks, triggering outreach from signals, and finding revenue.
Here is how that strategy holds up.
1. The human has to be justified, and the customer isn’t the one who decides
In the AI-first model, people are added: “only when AI reaches its limits.” The trigger is the system’s ceiling, not the customer’s need. Customer success was built on staffing to the outcome the customer is trying to reach. This model staffs to the cost the vendor is trying to avoid.
2. The scoreboard has no customer on it
The new metrics surface revenue, identify risk, and accelerate expansion. Every one is a vendor outcome, and “risk” means losing the account, not the customer failing. If nothing on the dashboard tracks time to value, adoption, or the result the customer bought the product to get, you have built a sales function and kept the old name.
3. Most customers get an imitation, by the industry’s own standard
The same leaders who call human connection irreplaceable plan to serve digital and pooled segments with AI agents that mimic a dedicated CSM. If connection matters most, most customers are getting what the industry itself calls a substitute. The real thing is reserved for “high-value” accounts, measured by contract size, not by how much the customer has riding on the product.
Unfortunately, ChurnZero is not the only company to get this wrong.
Gainsight now calls itself a retention-as-a-service company. Its Atlas business runs renewals end to end with AI agents and human oversight. Retention is what the vendor wants, but I hate to state the obvious: customers don’t buy software to be retained.
The data doesn’t support it either. Bain reports that CSM headcount grew more than 50% in six years. Meanwhile, median net revenue retention at public software companies fell from about 115% in 2023 to about 107%. More CSMs didn’t drive higher retention.
You could read that as proof that humans aren’t the answer either, and I’d accept it. In most cases, companies aren’t hiring people to focus on the customer. They hire salespeople and point them at the wrong thing. Capacity was never the problem, and more capacity is all AI-first offers.
Nobody renews because a system kept in touch
Customers renew when they trust the people they work with. They stop trusting you when they sense an agenda.
A customer once agreed to a meeting with me on one condition: sales could not be in the room. I sat alone with about 15 of their executives. One asked why they should trust me. I told them I didn’t care whether they bought our product or used someone else’s service, as long as they found value in what they needed to achieve. Two weeks later, we won the deal. That wasn’t me as a sales leader. It was me as a CSM developing a strategic value proposition in person, with no agenda and no AI.
Now give an AI agent ownership of the renewal. It has an agenda by design, and it never stops following up. A CSM carrying a renewal quota has the same problem. The agent just does it at scale. Every automated check-in tells the customer the same thing: we are watching your contract date.
Your customers are turning against AI
Customers are losing trust in AI while vendors push more of it.
- 39% of Americans now say AI does more harm than good, up from 31% a year earlier (Bentley-Gallup, May 2026).
- Among adults 18 to 29, that share jumped from 36% to 47%, the largest rise of any age group.
- 27% trust businesses to use AI responsibly, down from 31%.
- 31% of Gen Z report anger toward AI, up from 22%, and excitement fell 14 points in a year (Walton-GSV-Gallup, 2026).
- 61% of customers prefer a human agent (Verint, survey of 5,000 U.S. consumers).
These are consumer surveys, not B2B ones. But consumers are the people who log in to your product, and young adults are turning fastest. Half of Gen Z uses AI every week, and a growing share is angry about it. These are your future buyers.
A critic will quote Verint back at me: 69% of customers who prefer humans would switch to self-service if it fully resolved their issue. Look at what they asked for. They asked for resolution, and resolution needs automation. Automation needs no AI.
Verint surveyed support customers, so consider the issues too. A password reset is a workflow. Calling in to cancel a service is friction the company built on purpose. Neither compares to learning a platform your employer requires you to use, which is the actual work of customer success.
Then ask why AI gets deployed in this function. It lowers support costs and operating expenses. The AI health score is the clearest example. No customer ever asked for one. It exists to tell the company whether the account will renew.
Hot take: AI Maybe in customer success
My rule is Customer First, Technology Last, AI Maybe.
AI in customer success comes in two parts. One is AI used by CSMs to do less work. The other is AI placed in front of the customer to do the CSM’s job. The first is mildly acceptable. The second is forbidden.
Start with the first part. Before late 2022, no CSM used generative AI, and they still got the job done.
The best case for AI in CS is Bain’s finding that CSMs spend about 65% of their time on low-value tasks AI could automate. I read that number differently. If a machine can do two-thirds of the job, two-thirds of the job was never customer success. It was data entry, reporting, and renewal paperwork.
Automate that work where it lives: operations, the CRM, finance. Better still, ask why it exists. I consulted a company that could have reached the same outcome faster and cheaper with a simple search tool than with AI.
The standard defense is that AI frees up time to spend with customers. I haven’t seen it happen. The hours go to internal meetings, or the company loads more accounts onto each CSM, which produces more reports. Nobody is spending more time with the customer.
The idea-generation defense exposes the same hiring problem. Companies hired salespeople and asked them to be strategic consultants. AI fills the thinking gap the hiring left. Hire people who already think in patterns and systems, and the tool adds nothing.
Take away the admin, and what’s left is the real job: a person who knows the customer’s strategy and tells them the truth. That is the job AI is being put in front of the customer to do, and it can’t do it.
Plenty of businesses already work this way. Restaurants, hotels, and retail stores build relationships and guest experiences that matter. They have technology, but the experience is better when a person talks to a person and makes their goal a little easier to reach.
The work that keeps customers involves no AI at all. I have run it and watched it close deals:
- Growth squads. An account executive, a sales engineer, and a CSM share a set of accounts. The AE owns the contract and the number. The CSM owns no number, so there is no agenda in the room.
- Onboarding before signature. Time to value starts in pre-sales, while the customer is still deciding.
- Strategic Value Sessions replace QBRs. A monthly working session on the customer’s strategy, executives invited, with no slide reviewing our own usage charts.
None of these three needs AI. They need a person the customer wants to talk to.
But if you really want to use AI, then here is the title to strive for: “Customer-First Customer Success that uses AI.” It’s less trendy, but your customers will notice the difference.