The AI Rehiring Whipsaw Is a Measurement Problem
About the Author
Rajeev Ronanki
Chief Executive Officer
Many large employers are making "whipsaw" rehirings, in reaction to AI-related job cuts. This article explores the impact of measurement and valuable signals on improving the decisionmaking and outcomes in workforce decisions and management.
It’s not hard to fall into the camp of thought that flags AI as the great job-stealer. Just look at a recent report from Challenger, Gray & Christmas, which counted 101,743 job cuts attributed to AI in the first half of 2026. That’s nearly double the same type of job cuts for the entire year of 2025, representing nearly 23% of all job cuts.
However, the script appears to have flipped. Individuals affected by AI-related job cuts are now being rehired, often into their old roles.
A late July Wall Street Journal story shares how large employers, across technology, transportation, defense, and industrial manufacturing, are telling investors they plan to add people again. One example is Booz Allen Hamilton, whose chief operating officer said, twelve months after the company shed thousands of roles, that the firm needs to accelerate hiring and is running behind. Add to this, a survey from Robert Half, which found more than three in ten US hiring managers who eliminated positions, after their organization implemented AI, had added those roles back.
The cost of a rehiring round trip
When it comes to measuring the costs involved with these cuts and rehires, the data tells the story. A recent survey by Careerminds polled 600 HR professionals, where over 52% brought back the roles they cut within six months. The financials shared by leaders included 30.9% who said rehiring costs were greater than the job-cut savings; 42.4% neither saved nor lost money from their decisions; and 33% who felt their organization’s workforce lost critical skills through the cuts and rehiring process.
One recent example of this whipsawing of the workforce, which came in a positive light, is that of Ford. After its inspection systems fell short of expectations, the U.S. automaker rehired more than 300 veteran engineers and quality inspectors. The returning engineers not only improved the systems but also mentored the younger workforce staff. As a direct result, Ford captured hundreds of millions of dollars in warranty and recall savings, plus a top finish in this year's JD Power Initial Quality Study.
Ford aside, it costs to bring lost judgement back to companies. It’s a high-dollar proposition through rehiring, as you carry costs related to severance, recruiting fees, and the salary premium.
Human-AI coordination: the line item not carried
Today’s agent-related tasks don’t often show up in most org charts, or even as budgeted line items. Such tasks can include checking ingested data, identification of exceptions, and parsing out the calls a human must own. Agents are often expensed through their license fee and token-use bills, which have little to no measurement of operational cost and delivered value.
Now, companies such as IBM are starting to rethink structure and measurement within the blended workforce. A great example is found in Big Blue’s AskHR agent, which reports an eye-opening 94 percent containment rate of common questions. The agent operates within a two-tier support model, where tier one takes routine inquiries, and tier two has human advisors to handle more complex needs.
The agent and model have helped reduce HR operational costs by a whopping 40% over four years. Their high rate of common question containment being more of a testament to measuring tasks vs broader operational metrics. This is a clarion call to the masses afraid of AI-fueled layoffs. To have them realize that AI in the workforce creates more value by absorbing parts of jobs than entire jobs themselves.
And what about measuring the coordination between humans and agents in the workforce?
Recently, the consulting firm Sage and research leader IDC came together to survey more than 2,000 senior finance leaders. When these leaders were asked about human-AI interaction in the workforce, 49% of these leaders reported spending 15 or more hours a week checking outputs; and 19% spent more than 30. The researchers also coined a new term “verification tax,” to identify these human oversight tasks: including reconstructing, validating, and defending AI outputs. Only 9 percent of leaders intend to pursue broad autonomy for measuring transactional work, while 71 percent would veto a 99-percent-accurate tool that cannot explain its answers.
While AI is increasingly integrated within company tasks and projects, many enterprises continue to run two workforce-related "payrolls": one as salaries for employees, and one as tokens for agents. Workforce salary is typically tracked down to the dollar and kept exact in measurement. While recent reports of skyrocketing, surprise AI token bills, sometimes realized by larger organizations several months later, have caught many leaders and budgets by surprise.
Focus on the junior pipeline
Due to AI’s extensive integration within organizations, many leaders decided to freeze their junior roles. Their belief: AI would lower costs and increase productivity without negatively impacting the organization.
And it shows, as recently as six weeks ago. Where economists from Indeed’s Hiring Lab, the site’s economic research arm, reported the rise of senior-level postings by almost 15% over the prior year. Specific to the popular segment of software development, senior roles accounted for 69.3% of job listings versus only 4.5% at the entry level.
Organizations appear to be attempting to “buy seniority” on the open market instead of producing it internally. The squeeze runs from the other direction too, as 30% of applications to entry-level postings came from workers with ten or more years of experience.
It’s clear that the junior level rung is getting more crowded these days.
Getting back to IBM, it chose to triple US entry-level hiring in 2026, as a part of redefining the “AI-first workplace”. They are not in the minority, as new data from advisory firm Teneo has 67% of global CEOs reporting that AI is helping to increase entry-level headcount and reshape the workforce.
What leaders could see, and what they could not
Three different systems have driven growth and innovation in enterprise software for the last 30 years. Systems of Record, which companies how many people they employ and at what cost. Then came Systems of Engagement that captured what the employees communicated with customers and each other; and finally, Systems of Insight that support the decision-making process through use of data.
All three enterprise software layers worked as designed. However, even today, they are often unable to uncover a company’s most valuable signals, crucial to making timely and effective decisions that impact the workforce and business results. Some of these signals to help answer:
• Where is the limited attention of our decision-makers going this week?
• Which human-AI coordination work is producing that load?
• What happens across the company when we remove one experienced person, or add fifty agents to a workflow that already has a review bottleneck?
Those questions, and many more, require a new layer of enterprise software known as Systems of Attention. A new category of technology that determines which signals deserve the attention of company leaders, as well as the delivery of such signals to the right individuals, in time to act. Plus, to build strategic institutional memory through the recording of every delivered signal, what decision it informed, what leader action was taken on it, and the resulting business outcomes. This compounding and shared intelligence helps organizations get smarter over time, and to pass patterns and lessons learned to present and future employees and agents in the present and future workforce.
The next era of measured workforce decisions
Today’s workforce plan is shown as a headcount spreadsheet with an automation assumption typed into a cell. Too often, when it’s put into motion, nobody regularly monitors and measures against the assumption.
With Systems of Attention, workforce decisions are informed by signals that carry evidence, proposed outcomes, and time-to-action. A workforce plan can now be loaded in the form of a set of surfaced and monitored signals versus a static document. This enables reskilling timelines, automation rates, and review load to carry a measured value across a timely signal—that can drift and be seen drifting in real time. When any signal value moves, the leaders accountable for the workforce plan and underlying objectives and goals know about and have time and information to properly act.
This above video shows a controlled simulation involving SignalOS™, an operating system for bringing greater attention to enterprise leaders. In it, a hypothetical workforce challenge delivers key signals to the chief risk officer (CRO), to rebalance labor and reskilling. The CRO makes a $10M crucial workforce plan decision that would normally take weeks, in only 47 minutes.
Actions before your next workforce decision
1. Price your whole workforce decision, then monitor it. Saving salaries requires measuring and monitoring aspects such as rehiring premiums, lost workforce skills, and shifting timelines.
2. Put both “payrolls” on a single ledger. If your company keeps employee wages and AI tokens separate, bring them together in the same book. Until both costs are shown side by side, every automation case compares a hard number against a guess.
3. Give the bill for AI token use to one owner. Whoever integrates the automation and agent use on tasks creates the spend, and whoever owns the budget absorbs it. Name the owner now, before confusion and skyrocketing costs set in.
4. Tie every automation to the task work it replaced. Usage is easy to measure, while value is more challenging. Before any automation scales, someone should be able to point at the work it took over and say what that work was worth.
5. Remember the missed signals. Think about some of the worst decisions or missed opportunities in your company’s recent history. Ask what the piece of information you had, or could have interacted with externally, was that brought to the attention of leaders too late.
Let’s revisit the 101,743 AI-related job cuts in the last six months. Nearly all those decisions were made by people informed from what their current systems showed them: headcount, cost, and last quarter. As many organizations have decided to quickly rehire those cut roles, their decisions continue to be linked to systems designed for other purposes, and analytics that create too much noise through a plethora of alerts.
We’re all part of an evolution that’s led to the arrival of the blended human-and-agent workforce. The decision to significantly cut and then quickly rehire is a clarion call for leaders to recognize the limitations of their workforce operating models and the means of measuring within them. Now, through Systems of Attention, enterprises can receive the most relevant signals to take better actions, some of which include pricing the review work, catching the loss and timeline of reskilling, as well as showing junior freeze costs over five weeks or five years.
Large company whipsaw rehiring is a symptom. The cure isn’t found in more data, dashboards, or alerts, but in recognizing that leaders have a scarcity of attention. Decisions on the workforce, and the betterment of the organization, must be better informed by time-sensitive and evidenced signals, delivered in time for leaders to act.