AI Job Design: Redesign Roles Before You Rehire

AI Job Design: How to Redesign Roles Before You Have to Rehire

A growing number of companies have learned the same expensive lesson this year: cut a role for AI, then discover a few months later that the work still needs a person. Nearly a third of U.S. hiring managers, 32 percent, say they eliminated a role primarily because of AI and later rehired for the same or a similar position, according to Robert Half data reported by CNBC. Ford and IBM are two of the more visible names on that list, both bringing back experienced staff this year after finding their AI systems could not fully carry the roles they had eliminated, per the same reporting. Gartner expects the pattern to widen: the firm predicts that by 2027, half of the companies that cut customer service jobs for AI will reverse the decision, often rehiring the same work under a new title.

HR circles have started calling this the AI boomerang, and it is largely avoidable. The problem was never that AI could not help. It is that most rollouts skipped the design work: mapping which parts of a job AI can carry, which parts still need a person, and who owns the outcome when something goes wrong. That was the focus of a recent Achieve Leadership Network Playbook Lab session, AI Job Design, Done Right: Beyond the Cut-and-Rehire Cycle, hosted by Vanessa Cannizzaro, Vice President of Talent Management and Operations at Intelerad Medical Systems, a GE Healthcare company. What follows combines the frameworks from that conversation with the latest third-party research on where AI job design is actually landing.

Why the Cut-and-Rehire Cycle Keeps Repeating

Gartner's research points to a mismatch between ambition and readiness. In its polling of customer service leaders, only about a fifth reported that AI had actually driven their headcount reductions so far, yet the pressure to announce an AI-driven efficiency win keeps organizations committing to plans their systems cannot yet support, according to Gartner's February 2026 forecast. During the Achieve session, one participant offered an analogy that captured the pattern well: pulling a load-bearing wall out of a house to make room for a new window will bring the house down. Skipping the underlying architecture of a job, the systems, the training data, the escalation paths, to move fast on an AI rollout carries the same risk. It is a structural problem before it is ever a technology problem.

AI Automates Tasks. It Does Not Replace Skills.

McKinsey's most recent modeling of the human-AI workforce puts a number on this distinction. Currently demonstrated AI and robotics technology could, in theory, automate about 57 percent of paid work hours in the United States today, according to McKinsey Global Institute. The operative word is theory: McKinsey is explicit that the figure measures technical potential, not inevitable job loss, and its analysis found that more than 70 percent of the skills employers look for today show up in both automatable and non-automatable work. The skills mostly stay. What changes is which tasks they get applied to.

That distinction sat at the center of the Achieve session too. AI takes over tasks, not skills, Cannizzaro told the group, and organizations pouring resources into skill-building programs without first mapping which tasks AI can already absorb are solving the wrong layer of the problem. She walked participants through a live example: rather than hire a contract lawyer to review agreements full time, a role few people want to hold for long, the position was redesigned so AI absorbs the administrative volume while the person owns validation, client interaction, and the judgment calls that still need a human.

Where AI Actually Belongs in Your Org Chart

Microsoft's 2026 Work Trend Index backs this up at global scale. After analyzing productivity signals and surveying 20,000 AI users across ten countries, Microsoft found that organizational factors, culture, manager support, and talent practices, account for roughly twice the impact on AI outcomes that the tools themselves do, according to the Microsoft 2026 Work Trend Index. Microsoft's term for organizations that get this right is the Frontier Firm: one that deliberately designs how much of a workflow a person owns versus an AI agent, rather than layering AI on top of an unchanged org chart.

That lines up closely with the mapping exercise Cannizzaro walked the group through. Her framing questions: where does AI actually reside in the business, who has access to AI or AI agents, where is decision-making happening, and how are the parameters of AI use controlled? She compared building that map to building a job catalog: complex, but necessary before any rollout goes wide. One participant offered a related model worth borrowing: a hub-and-spoke structure, where a central hub sets governance, tone, and security standards, and individual functions build their own agents on top of that shared foundation instead of waiting for HR or IT to build every use case themselves.

The New Hiring Battleground: Verifying Who Is Actually on the Call

Job design has a hiring-side complication most organizations have not planned for. Gartner predicts that by 2028, one in four job candidate profiles worldwide will be fake, and in a 2025 survey the firm found that 6 percent of candidates admitted to some form of interview fraud, posing as someone else or having someone else pose as them, according to Gartner's July 2025 research. Session participants described living versions of this already: an interview filter so aggressive one talent leader could not tell whether the candidate on camera was a real person, and a separate case where the person who passed a rigorous interview was not the person who showed up for the job on day one, a mismatch caught only with support from the company's InfoSec team.

None of this is a reason to slow AI adoption. It is a reason to treat identity verification and interview integrity as part of job design rather than an afterthought bolted on after a bad hire. The same discipline applies inside the business: the World Economic Forum estimates that 39 percent of core job skills will change by 2030, according to its Future of Jobs Report 2025, which means the roles being redesigned around AI today will need a second look well before the decade is out.

Building Roles AI Cannot Own

This tracks with the accountability model at the center of the Achieve session. AI should not own outcomes, Cannizzaro argued, because it is not yet advanced enough to be trusted with decision-making, only execution. Practically, that means building review time and error-catch metrics into the roles that manage AI agents, the same way performance management already exists for people managers. One participant traced this discipline back to a pre-generative-AI publishing workflow that held anything scheduled for automatic release for eight hours so a human could take one more look before it went live. The tool changes. The checkpoint does not.

What This Looked Like in the Room

This piece draws on a recent Achieve Leadership Network Playbook Lab, a quarterly, cameras-on builder session where members co-create real, practical playbooks for the quarter ahead instead of starting from a blank page. This installment, AI Job Design, Done Right: Beyond the Cut-and-Rehire Cycle, was hosted mastermind style by Vanessa Cannizzaro, Vice President of Talent Management and Operations at Intelerad Medical Systems, a GE Healthcare company, and facilitated by Zech Dahms. Rather than presenting a finished framework, Cannizzaro opened the floor so members could bring their own AI job design questions and pressure test them together in real time, drawing on live examples from talent management, recruiting, compliance, and IT. Members left with a working set of questions for mapping AI into their own org design, along with a close look at how one HR team is building governance, policy language, and review habits that hold up past a single pilot.

Bring Your Own Playbook to the Next Session

AI job design is not a one-time project. It is a habit of asking the same questions, task by task and role by role, every time the tools change. The Achieve Leadership Network's Talent, HR & EX-Based Playbook Lab meets quarterly to build exactly that habit together, alongside a community of talent, HR, and people leaders working through the same questions in real time. Upcoming sessions this quarter include an AI for HR mastermind and a masterclass on talent density.

If you want to co-create the next playbook with us, learn more about joining the Achieve Leadership Network here.