Our Take
The cost-reduction playbook ignores that AI works best when you redesign processes from scratch, not when you automate existing inefficient work.
Why it matters
With 87% of HR leaders planning layoffs in the next year (per LHH survey), companies risk expensive rehiring cycles while competitors who redesign workflows capture the real AI productivity gains.
Do this week
HR leaders: Map your core processes on a blank whiteboard with AI agents embedded first, before identifying which human roles remain necessary.
IBM's $4.5B savings came from workflow redesign, not headcount cuts
Josh Bersin told HR leaders at HR Tech Europe 2026 that the standard AI layoff playbook is strategically flawed. Companies typically inventory jobs, catalog tasks, identify what AI can automate, then calculate position cuts. This approach misses how actual work gets done.
IBM CHRO Nickle LaMoreaux provided the counter-example. Over three years, AI freed up $4.5 billion in free cash flow and saved 22 million people-hours (company-reported). Software developers shifted from spending 80%-90% of their time coding to higher-value work. IBM's HR function cut its operating budget by 40% over four years while driving employee engagement to all-time highs.
The difference: IBM follows "eliminate, simplify, then automate" rather than automating existing processes. LaMoreaux framed this as redeploying entry-level hires to capture small and medium businesses the company couldn't previously pursue, instead of just cutting those roles.
Most companies are creating expensive problems for later
New LHH research surveying 3,000 HR leaders across seven countries found 87% have conducted or plan layoffs in the next 12 months, up from 77% in 2023 (survey-reported). AI transformation and skills displacement were primary drivers.
The costs show up later. Among HR leaders who track rehiring costs, 73% say it costs more to rehire than redeploy (per LHH). While 77% of HR leaders claim their organizations offer redeployment programs, only 19% of employees recognize those programs. That 58-percentage-point gap suggests most "redeployment" exists on paper only.
Meanwhile, Meta announced roughly 10% workforce reduction and Microsoft offered buyouts to about 7% of employees. Investors rewarded both moves, creating the incentive structure driving these decisions.
Start with the whiteboard, not the org chart
Bersin's core recommendation: design the AI agent first, then determine what human roles operate around it. The sequence matters because "the actual work that somebody does in a job is not really written down in the job description," he said.
The flaw in task analysis is that it preserves inefficient processes. If you're maintaining products manually today, automating that maintenance misses the opportunity to focus developers on new products entirely. LaMoreaux's question captures this: "We don't want to just say AI is taking away the drudgery. OK, to do what?"
For organizations already planning cuts, the data suggests focusing on genuine redeployment infrastructure rather than paper programs that employees don't recognize. The 73% who find rehiring more expensive than redeployment have the math to justify investing in the latter.