Read the definitions of workforce intelligence circulating this year and you will find the same ingredients every time. Recruiterflow describes it as internal workforce data combined with external labor-market evidence to guide decisions about roles, skills, capacity, location, cost, and talent supply. Analyst versions wrap the same ingredients in a decision loop (sense, understand, decide, learn) and fill it with questions about skills gaps, build-versus-buy, and which tasks AI should take over.
Those questions belong to a company with a people analytics team, a skills taxonomy, and an annual planning cycle (which describes a small slice of U.S. employers).
Even for them, the definition skips the outcome every one of those decisions produces, which is people moving. Someone gets hired, someone transfers, someone quits, someone gets let go, and every skills bet and automation plan shows up in that movement. The definition files it as a report at the end of the quarter, after the decisions it should have shaped were already made.
That makes most HR functions a coroner, the problem Paul O’Brien laid out in Your HR Platform Is a Coroner; It Should Be a Doctor. The separation report comes in after the person is gone, someone explains the cause in a slide, and the file closes. The behavior that came before the exit (a shift in attendance, a pay event, a drop in how often a manager and employee interact) was sitting in the systems for weeks, and nobody read it as a vital sign.
Strip out the new vocabulary and the definition describes talent intelligence, which the industry has been selling for a decade. It is built for the smallest slice of employers, and it cannot see the outcome that grades every decision it recommends.
The missing layer is movement intelligence, and workforce intelligence has no claim to being a new category until movement sits at its center.
Every Workforce Decision Ends Up as Movement
A skills strategy shows up in who you hire, who you promote, and who you lose to a competitor. An automation plan shows up in who you redeploy and who you let go. A schedule change at a franchise location shows up three weeks later in who stops picking up shifts.
Each workforce decision is a bet, and movement is how the bet gets settled.
Movement is also the most reliable data an HR stack holds. Skills profiles are self-reported and go stale the month after they are filled in. Hire dates, transfers, and termination dates are recorded by every payroll and HRIS system with near-certainty, because people get paid off them.
Skills tell you what a workforce can do and say nothing about whether that capability stays, a point O’Brien makes about entire cities in Your City Doesn’t Have a Talent Problem. It Has an Optionality Problem.
Leadership decides who gets hired and who gets let go; employees decide most of the exits. In June 2026, U.S. workers quit 3.2 million jobs while employers made 1.8 million layoffs and discharges, according to the Bureau of Labor Statistics JOLTS release.
Measured in aggregate, those employee decisions are the clearest read available on whether a workforce plan is working.
Enterprise workforce planning already runs on movement, whether planners call it that or not. The supply side of every workforce plan is a forecast of who will leave, who will move internally, and how many people must be hired to cover both. Adam Gibson, Global Practice Director, Workforce Strategy & Skills at The Adecco Group and author of Agile Workforce Planning, described his most common approach; take the last few years of attrition, run them through a quick simulation, adjusting for material internal and external factors to produce a forecast and a plausible range around it. He also made the point that over a three-to-ten-year horizon, greater forecasting precision rarely changes the strategic workforce decisions being made. That approach fits a strategic plan. The resource manager staffing next quarter’s shifts works on a 90-day horizon, where the question becomes which locations and shifts are losing people right now, and history alone can’t answer it.
Treat movement as the scoreboard for every other workforce decision you make, starting with this quarter’s exits.
Talent Intelligence Watches Movement After It Happens
Talent intelligence claims movement on paper; Recruiterflow’s same glossary lists candidates, companies, skills, markets, and movement as its scope. The definition practitioners work from leaves it out. Toby Culshaw, who led talent intelligence at Amazon and wrote the book Talent Intelligence, defines the field this way:
Talent Intelligence is the augmentation of internal and external people data with the application of technology, science, insights and intelligence relating to people, skills, jobs, functions, competitors and geographies to drive business decisions.
People, skills, jobs, functions, competitors, and geographies all made the list, and movement did not. Any definition of workforce intelligence built from the same ingredients inherits the same blind spot. Vendor practice follows the definition. Eightfold AI’s approach to talent intelligence promises to “turn deep understanding of people and work into real-time talent decisions,” lists internal mobility, redeployment, and a talent marketplace, mentions retention once in its copy, and never mentions attrition.
When talent intelligence does measure movement, it measures movement that already happened, usually at someone else’s company. Live Data Technologies tracks about 70 million people at 2 million companies and detects more than 30,000 job changes a day from the open web. That is useful market data, and every one of those changes is visible only after the person has updated a profile somewhere.
The line between the two categories is direction and timing.
- Talent intelligence reads movement from the outside, after it shows up in public.
- Movement intelligence reads it from the inside, from the behavior inside the employment window, before the resignation is typed.
One tells you where talent went; the other tells you where it is likely to go next, while there is still time to do something about it.
The analysts and vendors defining this category have built talent intelligence, and plenty of enterprises have bought it. They put skills inference and labor-market data on a slide labeled workforce intelligence and left movement outside it, as a planning exercise run once a year. That is a misdiagnosis. Every workforce plan they run depends on a movement forecast, which makes movement the vital sign that matters most, and they left it out of the one system they call intelligence while paying for the strongest tools everywhere else. The mid-market operators that definition leaves out had no part in writing it; they inherit tools that were never built for them.
Curious which of your own people are at risk?
Live insights on your workforce in two minutes.
The Signals Before an Exit Are Already in Your Systems
HR systems of record do their job well. They record hires, punches, pay, and separations after the fact, which is what they were built to do. They were never built to read those records forward, so a change in behavior gets logged as a static transaction.
People do signal before they leave. Timothy Gardner of Utah State University, Chad Van Iddekinge of Florida State, and Peter Hom of Arizona State, in a 2018 study published in the Journal of Management, collected 931 behaviors that managers and former employees had seen before people quit and narrowed them to 13 core pre-quitting behaviors. They then had 261 funeral home managers rate employees on those 13 behaviors and checked back 13 months later. Across the 193 employees they could track, the behaviors predicted who had quit voluntarily, beyond age, tenure, job performance, and the managers’ own guesses about who would leave.
Their study relied on managers noticing those changes. A district manager covering 40 locations cannot watch 900 people that closely, and a home-care agency with one HR person cannot either. Many of the same changes leave traces in system data anyway (a run of late punches, dropped shifts, a pay change, a missed check-in with a supervisor where those are logged).
I call the gap between when a signal enters a system and when someone reads it as a signal the Reporting Lag. For most companies the lag equals the time between the first change in behavior and the exit interview, which means the coroner gets the file while the physician never saw the patient.
Ask whoever owns your HRIS, payroll, and scheduling systems which of those signals they already log, and who reads them before an exit.
Workforce Intelligence Applies to Every Employer, and the Current Definition Was Written for the Largest Few
The SBA Office of Advocacy’s 2025 Small Business Profile counts 36.2 million small businesses, which make up 99.7 percent of U.S. employer firms and employ 45.9 percent of U.S. employees. A definition of workforce intelligence built around skills taxonomies and internal talent marketplaces describes the tooling of the other 0.3 percent and ignores the firms that employ nearly half the workforce.
The second group carries the heaviest movement.
In August 2026, the monthly quit rate in accommodation and food services was 3.5 percent, against 1.0 percent in finance and insurance, according to BLS JOLTS data. That is about one worker in 29 walking out in a single month. Franchise operators, home-care agencies, early childhood centers, and field-service companies live with that rate, usually with one HR person and no analyst.
Every one of those exits costs money. Heather Boushey and Sarah Jane Glynn of the Center for American Progress, in a 2012 review of 30 case studies across 11 research papers published between 1992 and 2007, found the median cost of replacing an employee was 21 percent of annual salary, and 16 percent for positions paying under $30,000. Run that against a 200-person home-care operator with 80 percent annual turnover and $30,000 wages, and 160 exits cost about $768,000 a year (illustrative math, not a client figure).
Their first workforce question is who is likely to leave and what a manager can do about it this week. That question gets waved off as an SMB retention problem, and that label misreads what the forecast is used for. The same read on likely movement tells an operator how many people to recruit next month, which location needs a manager conversation, whether last month’s schedule change is driving exits, and how much of next year’s budget will go to replacing people. Retention is one of the decisions it informs; staffing, scheduling, manager coaching, and budget are the others.
If your workforce intelligence plan has nothing to say to your highest-turnover population, rewrite the plan starting there.
Movement Intelligence Reads How Populations Move, Never How One Person Will
Can anyone predict who is going to quit?
John Sumser, the independent HR technology analyst behind HR Examiner, answered that with a Galton board when we recorded an upcoming episode of Leading Indicators.
A Galton board is a pegboard where a couple thousand beads drop through rows of pins and pile into a bell curve, and his has one gold bead mixed in. Run it as many times as you like and the curve comes out the same shape, but the gold bead lands somewhere different every time.
The curve is accurate and the gold bead is random, and both stay true no matter how many times you run it. Movement intelligence is the curve. It tells you a location, a shift, or a cohort of new hires is likely to lose a given share of its people over a given window, and it gets judged by calibration; when it puts a 30 percent likelihood on 100 new hires leaving before day 90, roughly 30 should leave. It can never tell you exactly which 30 with certainty. A new hire has no personal history to compare against, so early reads come from how comparable cohorts behaved, and the read sharpens as tenure builds.
The curve still reaches individual people, because managers act on one person at a time. A relative score places an employee against a population in similar roles, sites, and working patterns, and shows how strongly the curve leans for someone in that position and which patterns are driving it. It tells a manager where a conversation, a schedule fix, or a check-in is worth making first. It is never a determination about that person, and the person still decides where their own bead lands.
That is why a forecast should report how much signal stands behind it rather than a single accuracy score.
A location with years of scheduling, pay, and attendance history supports a strong read; a new franchise with three months of payroll data supports a weak one, and the forecast should say so.
Layoffs and terminations are leadership decisions, and behavior does not predict them. Movement intelligence measures them after the fact to judge whether the decisions behind them paid off (did the restructured team hold its people, did the replacement hires stay past their first year).
Reading the content of someone’s messages or screen is surveillance, the line we drew in Workforce Surveillance Is When Your Boss Reads Your Slack, Analytics Is Noticing You’re Not Happy. Reading aggregate patterns in the events a company already records for payroll and scheduling is analytics, and treating the two as the same thing helps nobody.
Keep outputs at the group level, route them to the manager who can act, keep a person in every decision, and document how the forecast was produced.
Movement Intelligence Is the Core of Behavioral Workforce Intelligence
Movement intelligence is the practice of forecasting workforce movement (voluntary exits, internal moves, and hiring need) from the behavioral patterns that come before it, measured in aggregate during the employment window, and of measuring involuntary movement to judge the decisions behind it.
It sits inside Behavioral Workforce Intelligence, the practice of reading the behavioral signals that build up between the hire event and the separation event. BWI reads the space between those two events; movement intelligence is what that reading predicts about the events themselves.
What belongs in movement intelligence:
- Likelihood of voluntary exit for groups (by location, cohort, tenure band, shift, and manager)
- Relative risk for individual employees, weighed against a population in similar roles, sites, and working patterns, used to direct manager support
- Early-tenure risk for new hires, read against comparable cohorts
- Hiring need driven by expected exits, ahead of the requisition
- Internal moves read as stay signals or leave signals
- After-the-fact measurement of layoffs, restructures, and other leadership decisions
What stays out of movement intelligence:
- Skills inference and skills taxonomies, which belong to talent intelligence
- External job-change tracking, which reads movement after it becomes public
- Monitoring the content of messages, screens, or keystrokes, which is surveillance
- Treating any individual score as a determination about a person, or using it to make pay, promotion, or termination decisions against them
Use that list the next time a vendor, an analyst, or your own team puts “workforce intelligence” on a slide and ask which side of it their version covers.
Workforce Intelligence Gets Judged by Who Stays and Who Goes
The definition of workforce intelligence most of the industry is selling leaves out the one outcome that grades every decision it recommends. Movement is recorded by every system, it costs real money every time it happens, and the behavior before it is already sitting in the data. Filing it as an end-of-quarter report is coroner’s work and selling that as intelligence is a misdiagnosis; reading it while there is still time to act is what a physician does.
This quarter, take your voluntary exits by tenure band and location, look at the 60 days before each one, and track first-year turnover cost as the number that tells you whether acting on those signals worked.
If every workforce decision you make eventually shows up as movement, why is movement the one thing your workforce intelligence still reads last?
