Searching for the Next Metaphor
For years people have referred to me as SEO’Brien because of my domain while throughout, I got to watch the work we do get renamed more often than I replaced laptops. Decades ago, Marketing had to be distinguished from Digital Marketing because CMOs didn’t yet grasp the internet; search engine optimization became inbound marketing, got folded into content marketing, picked up a hoodie and called itself growth hacking, and this year the same advice is being sold as answer engine optimization to companies terrified that ChatGPT can’t find them. The work never changed once; marketing, making something worth finding, make it easy to understand, and earning the trust of people who will point to it. Every new label created complexity and confusion while the work remained foundationally the same.
Walking HR Tech in Las Vegas last week gave me the same feeling at industrial scale. The program was organized around tracks like The Intelligent Enterprise, the investor sessions introduced agentic workforce management as an emerging category, and nearly every booth used a word that didn’t exist two years ago. John Sumser, who has analyzed this industry for three decades at HRExaminer, summed the whole thing up in a phrase I haven’t been able to shake, “searching for the next metaphor.”
That phrase cuts two ways and I find most keep choosing the cheap one.
The expensive way is searching for a new frame that changes how we understand work and what we measure about it. This is genuine innovation; it’s difficult, transformative, and disruptive; this solves underlying problems. The cheap way is searching for a new word that makes last year’s solution sound like next year’s breakthrough; this is evident in merely apply “tech” to an existing solution, continuing to solve known problems with known solutions but making them techier (is techier a word or am I making that one up?)
The vocabulary changes so the sales pitch can change, and decision makers can appease their boss that they implemented the new solution; the problem it was bought to solve mostly stay put.
I am tired of it, and I think too are the HR leaders I talk to who keep buying the same filing cabinet in a new color. We used to manage people in a spreadsheet and now we have workforce data solutions (apparently there are hundreds of them)… that’s just a techier spreadsheet. Problems stay unsolved, the opportunities sitting in plain sight stay ignored, and “innovation” turns out to be a service business wearing a technology t-shirt, because a new word is the only thing most buyers have any way to evaluate.
HR Has Repainted the Same Filing Cabinet Every Few Years
Strip the UX design off most HR technology and you find three things; a database of people and the events that happened to them, a workflow that moves a form from one inbox to another, and a report about both. That’s a filing cabinet with a mail slot and a clipboard, and the industry has been repainting it:
Personnel became human resources, then human capital, then people, then talent, and now workforce with an AI prefix.
The applicant tracking system became a talent acquisition suite, then a candidate experience platform, then an AI recruiter, and this year an agentic sourcing layer.
HR reporting became employee data, then HR analytics with a dashboard.
The annual survey became engagement, then employee experience, then continuous listening, while somewhere in there is a culture assessment, all of which most companies read once a year.
Sumser prefers to call the whole bundle of math in HR (AI, machine learning, natural language processing, people analytics, data lakes, and other automation) by a deliberately boring name, Intelligent Tools, which strips half a dozen marketing categories back down to what they actually do. ADP’s Naomi Lariviere makes the same case from inside a company that runs payroll for more than a million clients. In her HRExaminer interview, she explains that ADP uses deterministic, rules-based AI for payroll and avoids probabilistic AI where accuracy has to be absolute. The system’s value comes from those rules, and the rules come from people who understand payroll, tax, and compliance well enough to write them down as absolutes; no data lake or machine learning model can substitute for that expertise.
Repainting the cabinet pulls a company’s attention away from where that value sits. HR’s real work is defining the rules, standards, and judgment that make any tool meaningful. Buying a new category for every new label produces a patchwork of variations instead of shared standards, and the company pays new-category prices for a drawer that opens exactly the way the old one did, because the expensive part, writing the rules, still falls to the same people whichever tool holds them.
Vocabulary Inflation Devalues Every Word the Industry Prints
Words in technology change for three reasons, and only one of them has anything to do with the customer:
Vendors rename to stay relevant. A vendor that isn’t in this year’s analyst category or conference track doesn’t exist to the buyers walking the floor, so the product gets renamed to stay on the map.
Vendors rename to close the sale. Budget follows nouns, and a request for “AI-powered workforce intelligence” survives a CEO request that “a better report” would not.
Occasionally a word explains something new. Language has to catch up with something genuinely different, the way the “desktop” once helped people who weren’t programmers operate a computer.
Call the result Vocabulary Inflation; a rising supply of category names (the words) against a flat line of change in what the work delivers (the outcomes), which devalues every word in circulation the same way printing money devalues currency.
When everything is “transformational,” the word stops buying anything. Sumser opened his governance series this August with the observation that “The word ‘transformation’ gets tossed around like a hand grenade,” and the rest of that piece argues that much of the trouble comes from AI being immature and implemented without regard for control and consequences.
On one of his HRExaminer conversations, Sumser and co-host Heather Bussing joined Matt Charney to critique the overuse of AI as branding and the gap between enterprise-focused HR tech and what smaller employers need. Three people who have watched this market for a very long time said out loud what buyers usually mutter in hallways.
The obvious objection is that some of the new words are honest; a vendor really did build something new and needed language to explain it. The honest rename and the sales rename look identical to the outside; the shared cause of that new metaphor is that neither one emerged with any measurement of the change in work because of the tool. Brandon Hall Group’s David Forry told MustardHub that organizations can report how often a tool gets used and not whether the work got better, and when adoption is the only number in the room, the label becomes the only thing a buyer can compare because they’re adopting a new tool, whether it’s spin to convince the use or a genuinely new solution.
Words inflate when outcomes go unmeasured; give the CFO and Head of HR a way to see whether the work changed and vocabulary stops doing the selling.
Too Much “Innovation” Is a Service Business Wearing a Tech T-Shirt
The new t-shirts got bad enough that regulators recently named it! In March 2024, the SEC brought its first enforcement actions for what then-Chair Gary Gensler called “AI washing,” settling with investment advisers Delphia and Global Predictions for $400,000 in combined penalties over claims about AI capabilities the agency found they didn’t have. Two small advisers and a modest fine; the incident matters far more to my point than the dollars because a federal regulator decided the gap between the label and the product had become a problem.
Then came Builder.ai, a London startup that raised hundreds of millions of dollars with Microsoft among its backers and collapsed into insolvency in 2025. Bloomberg’s reporting, as well summarized here, described engineers in India hand-coding client projects marketed as the work of an AI assistant, alongside allegations of inflated revenue. Former engineers have since pushed back, saying a small team built genuine AI tools and that alleged accounting fraud sank the company. It doesn’t really matter here and I’m not asserting anything, both versions make the same point; years and hundreds of millions of dollars in, nobody on the outside could say how much of the product was software and how much was people, and the valuation depended on the answer.
The root cause is plain venture economics. Investors pay software multiples for software margins and services multiples for services margins, so every founder with a people-heavy business has a reason to describe the people as a platform. A recruiting service with an algorithm on the intake form gets pitched as an AI recruiter though it might be doing nothing more than using Claude, while a consulting engagement with a dashboard asserts that it’s analytics.
@John Sumser’s January piece on faster horses is one of my favorites, “We are using new tech to make dumb things into faster dumb things.” A service dressed as software is the same horse with a racing number painted on its side; it might win a sprint, but it still eats, sleeps, and bills by the hour. That isn’t an innovative solution to a problem; and HR (and employment at large) is full of problems that need real solutions.
The Problems Under the New Words Haven’t Moved
If the renaming produced results, I would shut up and buy the new nouns like everyone else; the scoreboard says the renaming produced a dictionary.
As many as half of HR decisions ultimately turn out to be wrong so you might appreciate that my personal passion for startups and entrepreneurship finds this work Sumser has been doing very fascinating because my world lives in being wrong most of the time. What we can’t be is misled; innovation and startups depend on clarity of capability and impact. Gallup found that 52% of voluntarily exiting employees said their manager or organization could have done something to keep them, and 51% said nobody had talked with them about their satisfaction or future in the three months before they left. Those are the problems every repainted cabinet promised to fix; bad calls about people and departures nobody saw coming.
The startup entrepreneur in me cringes at realizing we’re failing people (employees) to such a degree. I cringe more seeing all of these new offerings in HR when ultimately THAT is the problem -> HALF of people who quit say their manager could have prevented it. “Here you go, here is another workforce data platform, this one has AI on top.”
As I poked around more since Sumser’s next metaphor observation, I found one of the reason problems persist I found his governance series. In The Gap Is Where the Job Gets Learned, he separates work-as-imagined (how planners believe the job happens), work-as-prescribed (what the rules and software say), and work-as-done (what people actually do when the day gets complicated), borrowing from safety scientist Steven Shorrock 🌻.
Every vendor sells into work-as-prescribed, because that’s where the procurement process lives. The problems live in work-as-done, where startups live as they uncover innovative solutions.
He shared the “human in the loop,” which may be the most successful rename of the AI era.
A policy can say a person makes the final decision while the actual workflow makes challenging the machine slow and socially awkward, “If the worker lacks time or permission to disagree, human review is theater.” The label reassures the board, the product ships with a checkbox labeled “oversight,” and the quality underneath doesn’t change at all.
A root cause under all of it is that HR systems record events (a hire, a pay change, a termination) and almost nothing about the behavior between them, a gap I recently traced through fifty years of HR research. Software that can’t see the work can’t prove it changed the work, and software that can’t prove it changed the work has to sell the word instead.
The Opportunities Everyone Keeps Walking Past on the Way to the Next Booth
While the industry rehearses new words, the problem-solving work sits in plain view, mostly ignored because it takes entrepreneurial thinking more than the problem-solution fit jargon that is preached to most startups in startup accelerators (if you’re not familiar with those buzzwords, just go with me here):
The data most employers need already exists. Payroll, scheduling, time and attendance, and onboarding records at nearly every employer already hold the signals of who is drifting and why, and as I argued going into HR Tech, almost nobody in economic development, venture capital, or education cares because it already exists.
Young companies create the jobs. Census Bureau research by John Haltiwanger, Ron Jarmin, and Javier Miranda found no systematic relationship between firm size and growth once firm age is controlled, which points at startups, and most HR solutions are still built and priced for companies that already exist.
Invisible talent sits in the gap. The people who aren’t obviously founders or leaders yet are often the ones running the workarounds that keep work-as-done functioning, and credential filters screen them out for not matching a job description written for the last metaphor. The people that can and do solve the real problems are overlooked because it’s easier and safer to defer to the new tech thing some big company rolled out under a new metaphor as the new solution.
The job of management itself has changed. Sumser’s essay Managers Don’t Make Decisions traces the role supervisor to coordinator to the judge of whose expertise to trust, concluding “Increasingly, it is the allocation of confidence.” Our systems still record approvals as though the manager were a foreman signing a timecard.
I found that John Sumser predicted that smart companies would automate away their chances to learn and grow while smaller operations that don’t know better reimagine how things fit together, and he added, “I doubt that it will happen where it is expected to happen.” That prediction is uncomfortable for every incumbent; that yes, existing companies and solutions are making smart decisions but that isn’t the same as developing real solutions; the next real solution in workforce technology is FAR more likely to come from a three person startup that never learned the old nouns and just decided to fix what’s wrong and ask for forgiveness later.
Incumbents can’t give up the old vocabulary because their pricing, sales training, and customer contracts are written in it; sunk vocabulary behaves exactly like sunk cost, defended long after it stopped paying.
A Real Metaphor Changes What You Measure; a New Word Only Changes What You Buy
Searching for the next metaphor is worth doing when it means the expensive version.
In the 1960s, philosopher Donald Schön worked with a team at Arthur D. Little trying to make synthetic paintbrush bristles perform like natural ones, the breakthrough came when someone said a paintbrush is a kind of pump. The team stopped staring at bristles and started studying how paint flows through the spaces between them, which pointed them at properties nobody had been testing. Schön later called this a generative metaphor, and he describes it as “a process by which new perspectives on the world come into existence.”
The desktop computer was one of those once. We’re still living with fifty-year-old metaphors for how we relate to computers, and that your monitor is no desktop; the files, folders, and trash can were props that let people who weren’t programmers operate a machine, and that changed who could use a computer and what they did with it, which sets the bar; a new metaphor earns the name when it changes who can do the work, what gets measured, or which decision moves
Something like “agentic” has to clear the same bar “desktop” did before it deserves attention.
Call the cheap version a Costume Metaphor; a new term that leaves the measurements (what gets counted) and the decision rights (who acts on the things) exactly where they were.
The paintbrush team changed what they tested and the desktop changed who could compute, while a repainted filing cabinet changes only the color.
The real next solutions will sound simplistic, likely unfamiliar with the “right” words, look low quality at first, and take a long time to digest, which is the opposite of a polished rebrand. Genuine reframes tend to arrive looking worse than the costumes, and a buying process tuned to polish will reject the actual solutions almost every time.
Judge every new word by the report it kills; a metaphor that retires no old number and changes no decision is marketing and should be priced like marketing.
The Same Test Applies to Us, and AI Makes It Harder to Pass
I’d be a hypocrite to write this from a company whose website says “behavioral workforce intelligence” and exempt us from it.
That phrase earns its keep only if it changes what an employer measures (behavior over time instead of events after the fact) and who acts on it (a manager before a resignation instead of HR after one). If it turns out to be a new label on the same quarterly dashboard, it is a costume, and buyers should treat it like one. We wrote down what the category has to prove when we argued that your HR platform is a coroner and should be a doctor, and that standard is the one we WILL be held to.
Generative AI makes the costume problem worse, evident in The Hallucination of Competence. Polished prose triggers what psychologists call the fluency heuristic, a mental shortcut that treats easy-to-read as likely-true, and Sumser explains the trap as “We mistake aesthetic perfection for conceptual reality.” Every vendor now has a language model writing its website, its RFP responses, and its analyst briefings, so the costumes have never fit better; the solution behind the copy is the only thing a model can’t polish.
I know this pattern from my past in search. Google spent years learning to stop trusting what pages said about themselves, from keyword-stuffed meta tags to white text on a white background, and started ranking on behavior (links, clicks, what people actually did). Workforce technology needs the same correction; stop trusting what the product says about itself and start grading it on what changed in the work.
Spotting a Costume Before You Sign the Contract
Since we’re full circle to search engines let me also point out that we’re coming full circle on innovation and startups, my wheelhouse now. There is a trick to we use in venture capital to discern the real innovation (meaningful solutions) from new words masking existing solutions. None of these steps require a budget and they work the same for a CHRO evaluating vendors as for an investor reading a deck:
Make the vendor describe the product without the category name. Ban “AI,” “agentic,” “intelligence,” and “platform” for ten minutes and ask what the software does on a Tuesday; if the description collapses into a database, a workflow, and a report, you have found the filing cabinet.
Ask which report the product kills. A real change retires an old tool, so name what is stopped before you sign.
Ask which decision moves, and whose. If the answer is “HR gets better visibility,” nothing moves; a real answer names a specific manager and the action they will take weeks sooner. In HR Tech, for the most part, the solutions will remove tech teams from the process because HR works in the bottom line (financial implications) of the workforce, there is nothing for a tech team to do (but implement) if a solution is actually meaningful.
Find out how much of the product is people. Development teams, analysts in the loop, and “customer success” doing the actual work are fine, as long as you price them as services instead pretending you have an innovative new resource.
Follow five real cases end to end. Sumser’s method for AI governance works for vendor evaluation too; trace five actual workflows with the people doing them and record every workaround, because the workarounds show where the current tool fails and where a new one should have to earn its name.
Investors should ask what the founder’s product retires. A founder who can describe the old metaphor their company replaces, and the number that dies with it, is building something; a founder who can only describe the new word is renting it.
The Next Metaphor Should Cost Somebody a Metric
Search engine optimization outlived four names because the work underneath it never changed, meanwhile HR has repainted the same filing cabinet for longer than that because the people buying it had no way to see whether the work changed for the better. The words keep moving while the bad hiring calls and the departures nobody saw coming stay the same, which is the clearest evidence that the renaming was aimed at the buyer instead of the problem.
We should stop rewarding vendors, founders, and our own marketing teams for the cheap version of metaphors and start asking every new word which old metric goes away.
I want to compare notes with anyone who has watched a renamed product change nothing, or watched an ugly, unnamed workaround change everything, because the next real metaphor in workforce technology will come from someone who noticed that difference before anyone gave it a category name.