· AI
Which Jobs AI Changes First, and Why Redeploying Beats a Layoff Memo
Before you cut a role because "AI can do it now," find out whether AI is replacing that person's tasks or multiplying their output, because the data says those are usually different jobs.
A business owner who just bought the team a Claude or ChatGPT seat is now facing a second decision nobody warned them about: what to do with the person whose job got faster. The instinct in a lot of boardrooms is to treat "faster" as "fewer needed" and start cutting. The research on what AI actually does to tasks says that instinct is right for one category of work and badly wrong for another, and most owners cannot yet tell which one they are looking at.
The steelman for cutting now
There is a real case for acting early rather than waiting for more data. Federal Reserve Bank of Atlanta and Richmond researchers surveying corporate executives in April 2026 found larger companies already planning AI-driven workforce reductions, while smaller firms expect employment to grow instead. If competitors are cutting cost out of routine functions now, waiting a year to "be sure" hands them a real margin advantage, and severance is a known, one-time cost while a bad hire is a recurring one. A survey of 1,250 business leaders run by AI Resume Builder in November 2025 found customer service seen as the most exposed function (54 percent of respondents), followed by administrative and clerical work (49 percent) and IT support (47 percent). If a function is already flagged as high-risk by people running businesses like yours, moving early is reading the room, not overreacting.
Where that reasoning oversimplifies
The Atlanta and Richmond Fed paper itself does not support a blanket cut. Its authors report "little evidence of near-term aggregate employment declines due to AI" once you look past individual company plans to actual hiring data, and the decline they do find is specific: routine clerical positions are shrinking while demand for skilled technical roles rises at the same companies. That is not the org chart collapsing. It is one category of task disappearing while a different category gets more valuable, inside the same businesses.
Anthropic's Economic Index, built from roughly a million real Claude.ai conversations, shows which category is which. At its original release, only about 36 percent of occupations used AI in even a quarter of their tasks, and just 4 percent used it across three-quarters of their tasks, so "AI already does most of this job" is still the exception even in exposed roles. When people do use it, augmentation beats automation 57 percent to 43 percent, meaning most real usage is AI working alongside a person, not replacing their output. And usage concentrates in computer and mathematical occupations (37 percent), with programmers and writers, mid-to-high wage skilled work, showing the deepest adoption and a measured 12x task-completion speedup on college-level work versus 9x for high-school-level work. The multiplier is biggest exactly where the Fed data says demand is rising, not falling.
Put the two data sets together and a pattern shows up that a single layoff memo misses: the jobs actually disappearing are routine, low-judgment tasks, data entry, tier-one ticket triage, first-draft scheduling, while the jobs getting supercharged are the skilled ones sitting right next to them, often in the same department or even the same person's role. A blanket cut to "the department AI can now handle" risks removing the skilled half along with the routine half, then rehiring for it in eighteen months at a worse market rate.
What to actually do with the headcount
Audit tasks, not titles. Break each job description into its actual tasks and sort them into two buckets: ones where AI produces a finished, checkable output alone (automatable), and ones where it makes a person faster at judgment calls, writing, analysis, or client work (augmentable). Let attrition, not layoffs, shrink the automatable bucket, since new hiring into pure data entry or first-line triage is the piece actually declining industry-wide. For the augmentable bucket, retrain people up the task ladder into the judgment-heavy work one level above what they do now, before looking outside. You already paid to recruit and train that person into your business; a new hire starts that clock over, and the same vendor survey found 67 percent of business leaders now rate AI fluency as a highly desirable hiring trait, meaning staff who already picked up the tools are worth more externally than they were a year ago, not less.
This is the exact split MojoAI works through with a client before recommending a restructure: which tasks are genuinely automatable this year, which just got a productivity multiplier, and who moves where. If you are staring at a department guessing which bucket it falls into, that is a conversation worth having before the org chart changes.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.