A few of us got together last week over drinks to talk about where CX is actually headed.
No panels, no slides, just conversation. Here are the main themes that came up, and what we think you should practically do with the food for thought.
The AI bubble is narrower than it looks
Someone made a point that stuck: we talk about AI like it’s everywhere, but most people haven’t touched it.
Ask the teachers at your kid’s daycare. Ask your neighbors. Ask the person behind you in line. The AI conversation is happening in a small room, and the rest of the world is doing just fine without it. That matters for CX.
If you’re designing support around the assumption that your customer is comfortable with chatbots, patient with imperfect answers, and willing to self-serve before talking to a person, you may be designing for a customer who doesn’t even exist yet.
Most customers still want a real person when something goes wrong.
What to do with this
Pull your deflection data and look at who’s actually self-serving versus who’s bouncing to a human after one bad bot interaction. Segment it by customer tenure, age, and issue type.
You’ll likely find that your AI-first flows are working for a specific slice of your base, and quietly frustrating the rest. Build escalation paths for that frustrated slice before you expand automation further.
We’re quietly starving our leadership pipeline
The second thread hit even harder.
A lot of brands have stopped hiring entry-level. No copywriters, no recent grads, barely any Gen Z or Gen Alpha. AI is doing the starter work now, so the starter roles are disappearing.
One CEO put it bluntly: we are sucking the potential from our future leaders. The short-term math works, but the long-term math absolutely doesn’t. The people who become VPs in ten years are the people doing the junior work today. If nobody is junior now, who runs anything later?
CX feels this first because most CX leaders came up through the frontline. Cut that pipeline and you don’t just lose headcount, you lose your bench.
What to do with this
Audit your org chart for the next two levels of leadership.
If the five-year column is thin, that’s your problem to solve this year, not later.
Protect at least some entry-level roles, even when AI could technically do the work. And build structured paths from frontline to team lead to manager.
The ROI isn’t in this quarter’s P&L, it’s in whether you have anyone to promote in 2029.
Everyone wants to automate 80%
The last theme showed up in almost every conversation: “We can automate 80% of this.”
The number isn’t the problem, that’s fine, the problem is the other 20%. That 20% is the hard stuff. The angry customer, the billing edge case, the three-way issue that touches fulfillment, returns, and a bug in the product.
That’s where judgment, empathy, and context matter. It’s also where most automation plans fall apart, because nobody staffed for the hard part, they staffed for the easy part they no longer need.
Automating the easy 80% doesn’t simplify your CX operation, it concentrates it. What’s left is harder, denser, and higher stakes. If your team isn’t built for that, you didn’t save money, you just moved the problem.
What to do with this
Before you automate anything, map what your 20% actually looks like.
Pull a sample of your most complex tickets from the last quarter and ask:
- who on the team can handle these today
- how long they take
- what “good” resolution actually looks like
That’s your baseline. Then staff, train, and compensate for that work specifically.
The team handling the hard 20% should be your most experienced people, not your cheapest. And their metrics should reflect resolution quality and customer effort, not just handle time.
If you don’t have the tooling or QA structure to pull clean ticket samples yet, start lighter. Ask your most senior team members which tickets they dread, which ones eat their afternoons, and which ones they wish they could hand to someone more experienced.
That conversation will get you most of the way to the same answer, and it’s a good forcing function for building the QA structure you’ll eventually need anyway.
The through-line
The AI conversation is smaller than it feels. But inside that small conversation, people are making decisions that will shape hiring, staffing, and customer experience for the next decade.
The companies that win won’t be the ones who automated the fastest. They’ll be the ones who kept humans in the loop on purpose, and built their operations around what humans and AI actually do well together.
