Here's what most managers get wrong about AI: they treat it like any other productivity tool. It summarizes your meetings, drafts your emails, and knocks out those tedious tasks sitting at the bottom of your to-do list.
Sure, that helps. But it completely misses the point.
The real transformation doesn't happen when AI assists you—it happens when AI starts acting. When systems resolve customer issues, trigger workflows, and make routine decisions without waiting for human approval, the entire nature of work shifts. And when work changes, jobs have to change with it.
From Assisting to Acting: What Changes
Think about airline lost luggage. Today, generative AI can tell a customer service rep what steps to take to recover a missing bag. But agentic AI? It actually finds the bag. It reroutes it. It arranges delivery. No human in the loop.
So what happens to that lost luggage specialist who used to spend their day tracking bags and updating systems?
They become a concierge. While the AI solves the logistics problem, the human focuses on the angry, stressed passenger who just landed in a foreign city without their stuff. They apologize. They offer vouchers. They recommend local shops where the passenger can pick up essentials. They turn a frustrating experience into something almost positive.
Same title, completely different job.
According to McKinsey, 78% of organizations now use AI in at least one business function. But here's the problem: most are layering it on top of existing roles instead of redesigning the work itself. That's a missed opportunity.
Three Ways Jobs Need to Change
1. When Tasks Disappear, Judgment Becomes the Job
Too many jobs are still structured like task lists: answer tickets, process requests, close cases. As AI absorbs the repetitive work, what's left are exceptions, trade-offs, and judgment calls that don't come with a playbook.
Take a service advisor at a car dealership. Until recently, their day was filled with scheduling appointments, sending follow-up emails, and making reminder calls. Agentic AI can handle all of that.
Now that the advisor can focus on decisions that require nuance. They notice that a particular customer is retired and has mobility issues. They see their appointment is scheduled for a morning when snow is forecast. So they call the customer proactively and reschedule for better weather. That kind of thoughtful, human touch? That's what sets this dealership apart and builds loyalty.
2. Measure What Humans Actually Contribute
When AI takes over volume, measuring people on speed and responsiveness forces them to compete with machines on machine strengths. That's backwards.
Evaluation should reflect what humans uniquely provide: quality of judgment, ability to prevent recurring problems, and the skill to improve the systems that learn over time.
That service advisor at the dealership? Don't measure them by appointments scheduled or cancellations rescheduled. Measure customer satisfaction scores. Track repeat business. Count meaningful interactions that led to upsells or prevented service issues down the road.
3. Make Human Accountability for AI Work Explicit
When AI is involved, someone must own the outcomes—even when the system takes the action. Someone needs to own escalation rules, workflow design, and quality reviews.
Without that clarity, AI doesn't reduce friction—it just shifts it to the moment something breaks.
In our dealership example, a human should still oversee the AI agents doing the scheduling work. If problems emerge, they need to catch them and solve them.
One of the biggest risks with AI isn't failure—it's neglect. Systems that "mostly work" fade into the background until they don't. Teams need protected time to review where AI performed well, where it struggled, and why. That oversight isn't an afterthought. It's part of the job.
This Is Already Happening
This shift isn't theoretical. Klarna has publicly shared that its AI assistant now handles a significant portion of customer service interactions—a clear example of how quickly AI can move from a support tool to a frontline worker.
Once AI is doing real work, old job descriptions stop making sense. Roles need to be redesigned. So do accountability structures, performance metrics, and oversight processes. Everything has to change together.
AI improves fastest when humans actively review and guide it—not when oversight is treated as an afterthought.
The next phase of work isn't about managing people plus tools. It's about designing systems where expectations are clear, ownership is explicit, humans focus on meaningful decisions, and AI quietly handles the rest.
If leaders don't redesign jobs intentionally, those jobs will be redesigned for them—by the technology itself, by urgent failures, and by the slow erosion of clarity inside their teams.