Companies Are Deploying AI. Now They Must Redesign How Work Gets Done.

In this edition of “Office Optional with Larry English,” Larry explores why simply adding AI to existing workflows isn’t enough. As AI agents take on more of the work, organizations have an opportunity to rethink who does what, where human judgment matters most and how work gets done.

Picture a software engineer who no longer spends the day writing code one feature at a time. Instead, she directs multiple AI agents working on different features at once, reviews their output, makes the architectural calls and steps in when something goes off course. Her work hasn’t disappeared. She’s still a software engineer, but her role has shifted from doing to directing.

This is the kind of shift happening for many roles, as organizations move from using AI as a tool to having it do more of the work.

Far beyond a boost in productivity, AI is changing how work gets done, what skills matter most and how some jobs operate day to day. And when AI changes what people spend their time doing, it changes what organizations need people to be skilled at.

From Doing To Directing

Microsoft’s 2026 Work Trend Index found that nearly half of Microsoft 365 Copilot conversations involved analyzing information, solving problems, evaluating or thinking creatively. Additionally, 58% of AI users surveyed said they were producing work they couldn’t have produced a year earlier.

In other words, AI isn’t just helping people do their existing jobs. It’s expanding the range of work they can manage and making traditional boundaries less rigid between roles.

A product manager, for example, might use AI to analyze customer feedback, identify patterns, draft requirements, explore potential designs, build an initial prototype and test different approaches before bringing the work to other experts. The person doing that work doesn’t necessarily need to be the best designer, engineer or analyst. But they do need enough understanding of each discipline to direct the work, connect the pieces and recognize when something isn’t right.

Underlying expertise still matters, but people may increasingly be expected to expand their skills, gaining at least a working knowledge of other specialties to make connections.

Additionally, AI can make expertise more valuable because there’s more AI-generated output to evaluate, refine and direct. When an agent writes the code, someone needs to understand the problem well enough to recognize when it’s wrong. When an agent reviews and synthesizes hundreds of sources, someone must decide what information matters. That takes domain knowledge and the ability to learn quickly. Employees don’t need to become experts in everything an agent touches, but they do need enough understanding to guide it, recognize when it gets off track and redirect with better questions.

Redesign The Work, Not Just The Tools

The organizations that capture the most value from AI will be the ones willing to rethink how work gets done; they’re not simply adding AI to the way they already work.

At my company, Centric Consulting, we’re already starting to see what that looks like. When we first layered AI onto an existing software development process, we found it created new friction and inefficiency. Teams still had to produce traditional deliverables for human review, even when agents didn’t require them to perform work.

This waste caused us to reflect on the traditional software development process and redesign the software development lifecycle for an agentic world. The goal of the redesigned process wasn’t to take people out of the process. Rather, it was to put human judgment where it mattered most. Instead of having people perform every step and document each handoff, the redesigned process asks agents to do more executing while people set the direction and validate results. In a proof of concept, end-to-end delivery ran 4.5 times faster.

In another client engagement, a Centric team saw how AI not only changes the process but also requires a change in the work itself. Specifically:

• People move from doing the work to defining the work. Instead of performing every task themselves, they set the acceptance criteria, establish thresholds and decide where human intervention is needed.
• People move from checking every output to managing exceptions. Rather than reviewing everything an agent produces, people determine what can happen automatically, what needs escalation and where a human needs to step in.
• People move from catching individual mistakes to managing patterns and risk. The job becomes less about finding one error at a time and more about watching for drift, repeated failures or problems that could spread across many decisions.
• People move from documenting handoffs to owning outcomes. Human oversight becomes a defined part of the process, with clear decision rights, evidence and accountability for the result.

Technology makes this shift possible, but technology alone doesn’t create it. The real work is redesigning the process so agents can handle more execution while people focus on direction, judgment and the decisions that matter most.

That’s the bigger opportunity with AI. It isn’t simply making existing processes faster; it’s changing what those processes can accomplish.

AI doesn’t eliminate human judgment. It gives us an opportunity to move that judgment to the places where it matters most — and redesign the work around what humans and AI individually do best.

This article was originally published on Forbes.com.