A new hire in her second week used to spend the day on forty relatively easy calls. Order status, password resets, a return she could process while the customer waited for a reply. Within a few weeks, she knew the catalog, the three things that go wrong most often, and what a frustrated customer sounds like.
With the front end increasingly moving to AI-powered chat, those inbound calls are gone. Her first live conversation now is with a customer who has already been through two agent sessions, is on an edge-case contract, and whose frustration levels have already crossed a threshold.
I’ve watched people come out of that first job and end up running product lines twenty years later. The contact center was never designed as a school, which is part of why nobody is protecting it now. The curriculum was an accident of volume. Easy calls arrived in large numbers, hard ones arrived rarely, and a new person climbed that gradient without anyone planning it.
Very few systems around it recorded what was happening. Workforce management optimized staffing against volume. QA scorecards measured adherence on individual calls. The tiering model existed to keep expensive people away from cheap work. Learning was a side effect that no tool tracked and no budget line paid for, so it’s easy to automate away without anyone noticing it’s gone.
Agents take the volume first, because inbound volume is a game of large numbers. What’s left for people is work that needs judgment, context, and some tolerance for ambiguity. That’s a fair description of a senior role and a terrible description of an entry-level one. You can’t staff the top of a skill curve with people who never had a chance to climb it.
I think this will not be immediately visible, but it will have long term consequences. It arrives as a slow shortage of expertise, maybe even years from now, when the people who would have been your directors turn out never to have had the job that made them. But by then, connecting the dots from a successful AI chatbot deployment to management and judgement gap will be hard to make.
The hiring math changes as well. If the only human work left needs judgment on day one, you’re recruiting from a smaller and more expensive pool for every seat, and the internal pipeline that used to feed it is no longer producing. Most service organizations have built their cost models on the assumption that they can hire capable people at entry level and grow them. That assumption is shifting, and budgets haven’t caught up.
There’s a second-order version worth watching too. AI agents are learning from every exchange as well, and they’re learning faster than the average new hire did. A CX leader has to decide which of those two curves the company is actually investing in, because how both are treated will have long term implications.
The opening here is real, and it starts with something CX already owns. Every resolved case in the archive is a worked example. A new hire can sit with thousands of them, with an agent as a tutor explaining why this account got a credit and that one didn’t, which is a better education than shadowing whichever veteran happened to be free that week.
The wider point is that CX will hit this before the rest of the company does. Service automates earliest because the work seems repetitive and the volume is high. Finance, marketing, and legal are on the same path a few years behind. Whatever CX leaders work out about building expertise will become the template the rest of the business comes to ask for.
I don’t have an obvious answer for what replaces the queue, and it’s too early in the AI era to know. The opportunity is to recognize the long-term design refactoring that this process will launch. Recognizing the dynamics are being altered by the arrival of AI at the edge is the first step in surfacing visibility for the next era of customer experience.