How to reorganize your customer experience team when AI arrives

How does a CX team actually get restructured?

Reorganizing a CX team for AI happens the way most real change does: quietly, over months, not in a single announcement. Fewer people stay on repetitive execution. More people move into training, supervising, and improving the AI layer underneath, until the team looks different without anyone marking the exact day it happened.

People don't disappear. They start needing different skills instead: more analytical, more strategic, with a soft skills piece geared toward experimenting, drawing conclusions, and making decisions with that information.

What’ we see is that around 40% of the team changes role within the first six months. A year in, close to 70% is already doing new work.

It actually lines up with what analysts are tracking industry wide too. In a February 2026 Gartner survey, nearly 80% of service organizations said they plan to move at least some agents into new roles.

• Day 0: Everyone is still doing their old job. Nothing on the org chart has moved yet.
• Month 6: About 40% of the team has shifted role, leaning more analytical and strategic.
• Month 12: About 70% is doing new work, operating AI agents and improving the data behind them.

The roles that used to be stuck in day to day execution often evolve into manager profiles, or into analyzing how to improve business data. In support teams, for example, it's not unusual for a group that used to have ten people answering tickets one by one to end up with just three handling the cases that truly need a human, while the rest train the AI, connect tools, and keep an eye on whether the automated conversations are working well.

Similar patterns show up in customer success, in onboarding, in quality. The ratio changes, but the shape of the change doesn't.

The work doesn't disappear. It changes shape, and that's usually exactly what throws people off.

Four speeds inside the same team

Almost no plan survives intact if it treats the team as one homogeneous block, as if everyone were going to board the same train at the same time. In practice, any team usually has at least four distinct profiles when it comes to AI.

1. Fast movers: the ones who take off on their own

They're naturally curious. As soon as a new tool shows up, they try it before anyone asks them to.

• They become the informal go to person: the rest of the team ends up asking them, not the manual.
• They need little hand holding: with access and context, they usually move on their own.

2. With support: the ones who need a nudge

They're usually good at their job, but change doesn't come naturally. With a clear guide and concrete examples, they get on board without much trouble.

• They respond better to examples than to theory: seeing a real case from their own account usually convinces more than any manual.
• They progress with structured training, not a loose link to a generic course.

3. Resistant: the ones who push back

They've usually spent years measuring their worth in closed tickets or calls made. Watching that number stop being the center of the conversation is scary.

• The fear usually has less to do with the technology and more to do with becoming obsolete.
• Without support, they can end up as the team's bottleneck.

4. New roles: the profiles that get born

They didn't really exist before. Now someone has to design the playbooks, review what the AI gets wrong, and translate account knowledge into rules the AI can follow.

• They combine customer knowledge with judgment about the AI: they're usually not purely technical profiles.
• They end up bridging the team and the system.

Almost no team is one hundred percent a single type. The norm is a mix of all four, and the pace of the group usually gets set by the slowest profile, not the fastest.

That's really where the leader's real challenge sits: adapting the transition to each profile, figuring out where each person fits best in the new structure, rather than running the same plan for all four at once.

How does each role change across post sales?

The pattern repeats area by area: fewer people doing the repetitive task, more people training, reviewing, and steering the AI layer that replaces it.

• Support: From ten people answering tickets to three handling escalations, while the rest train and monitor the AI.
• Customer success: From reactive check ins and manual scoring to part of the team owning key accounts and the rest tuning account health models.
• Onboarding: From one on one sessions with every customer to personalized paths by segment, with the team designing and measuring those flows.
• Quality (QA): From manually reviewing a handful of conversations to near total coverage, with the team designing the criteria the AI uses to score.
• Analytics: From weekly reports that arrive late to real time signals, with the team focused on turning patterns into decisions.
• Operations (WFM, workforce management): From manual spreadsheet planning to predictive staffing models, with the team redesigning how work gets distributed.

What does “adding value” mean once AI does the repetitive work?

When someone goes from answering tickets to training the AI that answers those tickets, their work doesn't become less important. It becomes harder to measure with the same old metrics.

That's exactly the shift Forrester points to as well: reps moving from directly handling inquiries to overseeing AI agents and resolving their exceptions (Forrester, 2026).

AI replaces tasks, not the customer experience, support, or CSM (customer success manager) role itself, and that forces a rethink of what it means to add value inside that role and that team.

The problem is that a lot of managers keep counting the same things: calls made, tickets closed, checks completed. But if half the team no longer makes calls, and instead tunes the model that decides which accounts to call this week, that number stops telling the real story.

Think of someone who used to run twenty manual account checks a week. Now they review why the model flagged three accounts as at risk instead of ten, and adjust the criteria for next week. It's different work, harder to summarize on a dashboard, but with more weight on the final outcome.

The team doesn't get smaller. It gets harder to replace.

The challenges that come with AI

No transition like this is purely technical. These are the challenges that come up, even when the AI system works well from day one.

• The fear of becoming obsolete: someone who's spent years measuring their worth in closed tickets sees that number stop showing up on the dashboard, and nobody's really explained yet what to look at instead.
• Still measuring with the old metrics: a manager keeps counting calls made per CSM (customer success manager), when half the team no longer makes calls. They train models.
• Deciding who trains the AI: the person who knows an account best isn't always the one most comfortable writing rules or reviewing edge cases, and that gap shows up fast if nobody fills it.
• The quality of the data feeding the system: an AI acting on outdated account information repeats mistakes faster than a human ever would.
• The uneven pace inside the same team: while some have been working with the AI for months, others are still waiting for someone to show them where to start, and that gap grows if nobody closes it on purpose.

Watch out for this

Confusing speed of adoption with talent is one of the most common mistakes in this transition.

The person who takes longer to get on board with AI isn't necessarily the one contributing least. Often, they're the one who knows the account best, and the one who has the most to unlearn before moving forward.

Pushing them at the same pace as the faster profiles usually gets the opposite of what you're after.

How Zentesimal supports this transition

Zentesimal treats whether AI is here to stay as a settled question. What it actually asks a team is whether they're ready to lead it.

We know this transition well. Metricool is one of the teams we've walked through it with, and their support and customer success evolution is documented here. It's the same 40% to 70% shift we see, project after project, across the CX teams we work with.

We sit down with each role (support, customer success, onboarding, quality, analytics, operations) to map out what changes day to day for that specific function, and build training around it instead of running one generic AI course for everyone.

Book a demo