Abhishek Yadav is the founder and CEO of Meza AI, an AI customer success platform for B2B SaaS companies.
Most customer success managers I have spoken with have had some version of the same thought in the past year, usually late at night and usually after watching a demo where an AI agent resolves a ticket, drafts a QBR deck and summarizes a call without pausing for breath.
The thought is simple enough. If the software can do the work, what exactly is left for me?
I understand the anxiety, and I want to respond to it with evidence rather than reassurance, because the data now available tells a more interesting story than either the optimists or the doomsayers have been telling.
AI is not emptying out customer success. It is rewiring what the function is, who sits inside it and how the work is organized. The practitioners who understand that rewiring early will be the ones running these teams in three years.
The Replacement Thesis Is Already Failing In Practice
Start with what is actually happening rather than what was predicted.
A Gartner survey of 321 customer service and support leaders found that only 20% had reduced staffing because of AI, while 55% reported stable headcount while handling higher customer volumes, and 42% were actively hiring specialized roles such as AI strategists, conversational AI designers and automation analysts. Gartner also expects that by 2027, half of the organizations planning major AI-driven workforce cuts will abandon those plans as agentless service proves elusive, and that no Fortune 500 company will have fully eliminated human customer service by 2028.
Read those three numbers together and the pattern is unmistakable. The dominant outcome is not fewer people, but the same people absorbing more volume alongside entirely new roles that did not exist on the org chart two years ago.
That is not replacement. That is a function being rebuilt while it runs.
Three Archetypes, And Where Our Roles Land
The most useful map I have found for what comes next is McKinsey’s work on the agentic organization, which identifies three roles emerging as humans begin working alongside agents.
There are M-shaped supervisors, described as broad generalists fluent in AI who orchestrate agents and the hybrid workforce across domains. There are T-shaped experts, the deep specialists who reimagine workflows, handle exceptions and safeguard quality. And there are AI-augmented front-line workers in functions like sales and service, who spend less time inside systems and more time with humans.
Anyone who has worked a customer life cycle will recognize all three immediately because they map almost perfectly onto roles we already have in embryonic form. The AI-augmented front-line worker is the customer success manager freed from the reporting treadmill, no longer assembling health summaries and renewal forecasts by hand because the system produces them, and therefore spending the recovered hours in the conversations that actually move accounts.
The T-shaped expert is the customer engineer or forward deployed engineer, the person who goes deep when a deployment breaks against a customer’s real data and security model, and who turns that exception into a workflow improvement rather than a one-off fix.
The M-shaped supervisor is the emerging customer success leader or CS operations owner who no longer manages only people, but manages a portfolio of agents alongside them, accountable for what those agents produce across the entire book of business.
What makes this framing valuable for practitioners is that it tells you where to invest your own development. If you are a CSM today, the question is which of those three you are building toward, because the generic middle of the role is exactly the part automation absorbs first.
Empathy Becomes The Scarce Input, Not The Soft One
There is a lazy version of this argument that says machines handle data and humans handle feelings, and I want to resist it because it undersells both. What actually changes is the quality of the input a human brings to a conversation. When an agent has already surfaced that a customer’s power user stopped logging in three weeks ago, that two integrations are failing silently, and that support sentiment has drifted negative, the CSM walks into that meeting with precision no amount of relationship instinct could have produced alone. The empathy is not replaced by that precision. It is finally pointed at the right problem.
Gartner’s own analysts make this point when they note that human agents remain irreplaceable for nuanced situations and for building lasting relationships. The nuance is the job. Telling a customer their implementation is failing, negotiating a renewal where value has not landed, and persuading an executive to reinvest after a bad quarter are all acts of judgment under social pressure, and judgment is the one thing in this stack that does not get commoditized by better models.
Designing The Organization Around Orchestration
The structural implication is that coverage models stop being a function of headcount. The old tiering, where a handful of accounts got high touch and the long tail got newsletters, existed because human attention was the constraint. When agents handle monitoring, outreach drafting and routine follow-up across the entire base, the constraint moves from attention to orchestration, meaning the real skill becomes designing what the agents do, defining when they escalate and auditing whether their outputs are any good.
That is why I think the next generation of customer success leaders will look less like player-coaches and more like system designers who happen to be deeply fluent in customers. The function is not shrinking. It is being handed leverage it never had, and the people who learn to direct that leverage rather than compete with it will find that the job has become more strategic, more commercial and, somewhat unexpectedly, more human than it was before.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?