Your CSMs are using AI. Their prep looks sharper. Their summaries are cleaner. Their account plans are more complete than they’ve ever been.
None of that tells you whether they’re doing the actual job.
This is the performance visibility problem AI just handed every CS leader, and most haven’t named it yet.
For years, the visible outputs of CS work were a reasonable proxy for capability. A polished QBR deck suggested preparation. A clean follow-up email suggested attention to detail. Consistent check-in cadences suggested relationship investment. You couldn’t fake the volume and quality of that work without actually doing it.
AI broke that proxy.
A CSM without real business fluency can now produce a flawless pre-call brief. A team that’s never owned outcomes can generate executive summaries that read like strategic insight. The work looks done. The capability isn’t there.
The outputs have never looked better. The judgment gaps have never been harder to see.
This isn’t a criticism of AI. It’s a leadership problem that AI surfaced, one that was already there, quietly undermining retention and expansion performance, long before the tools arrived.
What AI Actually Automated
To understand the visibility problem, it helps to be precise about what AI actually changed.
The tasks AI handles well in a CS workflow are largely the same tasks that used to consume the most time without requiring the most judgment: Synthesizing account data before a call, drafting follow-up documentation, building portfolio health summaries, preparing executive briefings from existing notes, flagging renewal risk signals across a large book of business.
These are real tasks. They took real time. Automating them is genuinely valuable.
The uncomfortable truth is that they were also the tasks that made a CSM look busy, look prepared, and look engaged, regardless of whether the deeper work was happening.
The deeper work is different. It doesn’t get automated.
Knowing a client’s business well enough to have a point of view on their competitive position. Connecting your solution’s value to a metric that gets discussed in the boardroom. Defining what success looks like for this specific client at this specific stage, and tracking it relentlessly. Showing up to an executive conversation with an observation the client didn’t already have.
That work requires judgment, curiosity, and commercial fluency. The ability to think three moves ahead on a chessboard you’ve actually studied.
AI compresses the time it takes to get ready for that work. It doesn’t do the work itself.
The CSMs who thrive with AI are those who already knew the job. They use the time AI frees up to go deeper: More preparation, sharper recommendations, more executive conversations, more proactive outreach. AI makes them untouchable.
Those who don’t know the job use AI to produce better-looking versions of the wrong motion. As a CS leader, you may not notice the difference until a renewal blows up.
The Hiring Gap AI Just Made Visible
Here’s the harder conversation.
For most of the history of Customer Success as a function, CS leaders have hired and developed their people around a specific set of capabilities: Relationship management, product knowledge, responsiveness, communication skills, and the ability to run a clean process.
Those things matter. They’re not wrong, but they’re not sufficient for the role the market now requires, and most CS hiring processes have never been designed to test for what is.
Business fluency. The ability to understand a client’s revenue model, competitive pressures, and board-level priorities, and connect your solution to those specifically. The willingness to show up with a recommendation rather than a status update. Commercial confidence in a renewal conversation. The instinct to anticipate a problem before the client feels it.
When did you last interview a CSM candidate on business fluency — not relationship skills, not product knowledge, but the actual business of their clients? When did you last ask a candidate to walk you through how their product moved a client’s specific KPIs, or to describe a time they changed what a client was asking for, not just delivered what was asked?
Most CS interviews don’t go there. Most CS teams have hired for the visible capabilities, those that used to differentiate, without explicitly testing for those that actually drive retention and expansion.
AI just made that gap impossible to ignore. When the visible outputs are automated, what’s left is the actual job. If your team wasn’t hired or developed for the actual job, no amount of tooling will close that gap.
The Deployment Order Problem
There’s a sequencing mistake most CS leaders are making right now, understandable because the pressure to adopt AI is real and the tools are genuinely good.
The mistake is deploying AI before fixing the fundamentals.
Rolling out AI to a CS team that doesn’t have business fluency gives you faster prep work for the wrong conversations. Automating follow-up documentation for a team that isn’t defining outcomes gives you cleaner records of the wrong activities. Scaling outreach cadences for CSMs who aren’t thinking proactively gives you more efficient versions of reactive account management.
An accelerant only works if there’s something worth accelerating.
The CS leaders getting the most out of AI right now didn’t start with the tools. They started with the fundamentals: Clear expectations around business fluency, outcome ownership, proactive engagement, and commercial accountability. They built or rebuilt their hiring and development models around those expectations. They changed what they measured and what they talked about in account reviews.
Then they handed their teams AI, and watched the gap between their best people and everyone else compress dramatically.
The sequence matters. Fix the job first. Then accelerate it.
What to Actually Look For Now
If AI has made traditional output metrics an unreliable signal of CSM capability, what should CS leaders be measuring and evaluating instead?
A few practical shifts:
Evaluate business fluency directly. In 1:1s and account reviews, stop asking about activity and start asking about understanding. Can your CSM tell you the client’s top three strategic priorities for the year? Can they explain how your product moves a specific metric that matters to that client’s leadership? If they can’t answer without pulling up a dashboard, you have a fluency gap, not a data gap.
Audit outcome definition at kickoff. Pull a sample of accounts that have been live for 90 days. For each one, ask: What specific outcome did we agree to drive for this client, in their language, tied to their metrics? If the answer is vague or product-centric, the three-link chain we discussed in the last post was never built. AI prep work on top of that foundation produces nothing useful.
Change what you ask in interviews. Add a business fluency component to every CSM interview. Ask candidates to walk you through how they connected their product to a client’s specific business outcome: Not the generic value prop, the specific client, the specific outcome, the specific metric. Ask how they handled a situation where a client’s stated need wasn’t their actual problem. These questions can’t be answered with AI-assisted prep. They require lived experience and genuine judgment.
Watch how your CSMs use the time AI frees up. This is the most revealing signal of all. When AI removes an hour of prep work from a CSM’s week, where does that hour go? CSMs who invest it in deeper client research, more executive conversations, or more proactive outreach are those doing the actual job. Those who absorb it into their calendar without a change in output are those whose capability gap AI just exposed.
Redefine what “prepared” looks like. In a world where AI can produce a polished pre-call brief in minutes, the bar for preparation has to move. A brief isn’t preparation; it’s the starting point. Preparation is having a point of view: On the client’s business, on where they’re headed, on what they should do next. That can’t be generated. It has to be developed.
The Bottom Line
AI is the most significant capability shift in Customer Success in a decade. It compresses research, surfaces signals, automates documentation, and frees CSMs to do more of what actually drives retention and expansion.
It only delivers on that promise for teams who already know what the actual job is.
For those that don’t, teams built around the visible work that AI just automated, it produces a more polished version of the wrong motion. The health scores look greener. The prep looks sharper. The outputs look professional. The accounts still churn.
The performance visibility problem is real. The hiring gap is real. The deployment order mistake is real.
All three have the same solution: Get clear on what the job actually requires first, business fluency, outcome ownership, proactive judgment, commercial confidence, and then build your hiring, your development, your measurement, and your AI deployment around that clarity.
Fix the job first. Then accelerate it.
That’s not an AI strategy. That’s a CS leadership strategy. AI is just what makes it urgent.
Andrea Mulligan is a B2B SaaS executive and advisor with 30 years of experience building Customer Success, Professional Services, and GTM organizations. She works with growth-stage companies on CS transformation, AI operationalization, and post-sale strategy. Start a conversation →