The Turn: To The Maxxxxx
A lesson about perverse AI incentives and short term thinking
I was still a teenager in the late 90s when all marketers apparently got together and determined that our lives were stupid and boring.
What did we need to make our lives better? Whatever we needed, it needed to be EXTREME.
Your sports? Extreme.
Your chips? Extreme.
You video games? Extreme.
Your deodorant? EXTREME.
Adding Xs or just dropping the E in front of extreme just added to how extreme something was. I even had one of the most extreme cars of the early 2000s: a Chevrolet Malibu Maxx.
I wish I was kidding.
The extreme marketing eventually (and thankfully) faded away. Until it didn’t.
Return of the maxx
There’s a whole Wikipedia page on the term “-maxxing.” You might’ve heard about looksmaxxing from the media somewhat recently but the suffix has been in use for longer.
If you don’t follow internet or gaming culture though, you probably have heard it in another context: tokenmaxxing.
With tokenmaxxing, companies set up leaderboards to encourage developers to max out their AI usage (typically measured by token usage). While your personal subscription to an AI tool may eventually run out of usage for a set period of time, these big organizations use AI via API, which will let you run it until you run out of money.
Some of the agents people were running were consuming 1000x more tokens than you would consumer from a regular chat session. People were using AI tokens for personal projects just to ensure they would stay top of the leaderboard.
Not surprisingly, this went poorly.
People who pay attention to pesky little details like money are sounding the alarm. CFOs are freaking out. The COO of Uber said that it was hard to justify the money spent on tokenmaxxing. An anonymous company blew through half a billion dollars of usage in a single month.
Maybe employees aren’t actually that expensive.
In spite of that, financial analyst say layoffs may continue as it is the only way some of these tech companies can fund their AI infrastructure, even if it’s not delivering value in the way they might hope.
This reckoning is coming at a bad time in particular for Anthropic, which has filed for IPO. Still, we shouldn’t feel too bad for them. After all, when is the last time the market acted rational? They’ll get their semi-truck full of cash.
What and how we measure matters
The real lesson here is that metrics matter when it comes to behavior. Companies told their developers to use AI. That they would be judged based on their use of AI. That high usage is what they wanted to see.
So employees followed suit.
Anyone who has studied human behavior knew what would happen. As The Conversation covered recently:
Metrics do have their place in an ordered and complex society. There are many instances in which we might happily defer to the scores produced by simple metrics, trading nuance for convenience. Aggregate ratings on product or restaurant review sites, for example, can simplify our decision-making, even if they aren’t tailored to our specific preferences.
The problem is what Nguyen calls “value capture” – when we uncritically allow external metrics to determine our own goals and behaviour. Resisting this process involves questioning what is being measured and reframing it.
You can’t solely manage a workforce based on the consumption of resources and hope that it translates to the outcome you want. When has that ever made sense as a long term strategy?
But in the AI land grab era, it seems to be par for organizations to do dumb things first and ask questions later.
What else is going on this week
The Best HR & People Analytics Articles of May 2026. David Green built May’s roundup around Gartner’s finding that 80% of companies implementing AI reduced headcount without improving financial returns. Take some time with this one.
Our Fears About AI Are Really Fears About Capitalism. The rogue AI scenario people fear is mostly just a description of what corporations already do, and the technology is amplifying the misalignment, not creating it. How well our needs are taken care of determines how scary the tool feels.
Remote Work, Not AI, Has Sidelined Recent College Graduates. New York Fed researchers say remote work explains nearly two-thirds of the unemployment spike among young college grads, which is a less satisfying villain than AI and also, conveniently, fixable with return-to-office mandates.
What’s Happening to Talent Acquisition Careers? (2026 Edition). TA employment declined for the first time in BLS data history in 2025, down 0.5%, and David Manaster lays out the honest version: the COVID correction may be nearly done, but whether AI has permanently moved the baseline is a question the trendline cannot answer yet.
Five Years of Recruiter Research, Revisited. Madeline Laurano and Aptitude Research went back with Greenhouse to see what changed in the recruiter experience over five years. Calling it a sequel undersells how different 2026 looks from 2021.
Master the Pivot: Resilient Leaders Need Learning Agility. The leadership content industry has thoroughly saturated the resilience topic, and most of it tells you to breathe deeply and embrace uncertainty. Donald Thompson actually addresses how to build the structures that survive it.
What Cornerstone OnDemand Is Actually Up To. Dani Johnson was at Cornerstone Connect in New York and came back with a clear-eyed look at what the company is building, Workforce AI included. Way more useful than a press release.
Fairness Is Not a Season. Structured interviews are twice as predictive as unstructured ones and remove most of the surface area where bias does its work, and if your commitment to fairness evaporates the moment it gets expensive, it was a campaign and not a value. Brian Fink writes this in June on purpose.
While You Were Working: May 2026’s HR Regulatory Whiplash. EEO-1 reporting proposed for elimination, the Biden overtime rule officially dead, Colorado narrowing its AI Act, Illinois proposing new AI employment rules, USCIS disrupting green card processing for millions. Robin Schooling rounds it all up, and the list is not a metaphor for whiplash, it is whiplash.
Salt in the Eyes. Mike Wood puts the EEOC’s push to kill EEO-1 reporting next to the Workday AI discrimination class action and names what’s happening: we’re sharpening the penalty and unscrewing the smoke detectors at the same time.
Long Live a Liberal Arts Education. Learning to read history, argue from evidence, and write clearly turns out to be useful regardless of what technology does to everything else, and Josh Bersin makes the case on his 70th birthday. The timing with the college grad employment panic is not coincidental.
Jevons’ Paradox, Engels’ Pause, and the Jobs Question. IKEA automated 57% of customer support, retrained 8,500 reps as Remote Interior Design Consultants, and generated $1.4 billion in new revenue instead of cutting headcount. Thomas Otter uses that as the anchor for an honest reckoning with AI and jobs.
The Case for Recognition That Still Works. Recognition, appreciation, and gratitude are still the most effective tools for engagement and retention, per the HBR research Laurie Ruettimann is sharing. A little embarrassing given how much organizations spend on software to solve the same problems.
The Hiring Manager of 2030. By 2030, the recruiter-as-coordinator is largely gone and the hiring manager loses the alibi that came with having someone else run the process, argues Kevin Wheeler. When the agent does what it’s told based on your inputs, the bad hire is yours.
The Trust Collapse: Why Your Workforce Has Stopped Believing Your AI Strategy. Jason Averbook names the crisis underneath the AI deployment numbers: trust. Forty-eight percent of Q1 tech layoffs got attributed to AI while the same organizations were publishing content about augmentation, and workers can read.
A Pattern Unlike Anything in 25 Years of HR and Work Tech. Decades in this market and George LaRocque says the current moment is unlike any of them. Article one of his new WorkTech intelligence series is live.
I Ran the Experiment. Here’s What Sameness Costs. Five recruiting CRM hero sections tested blind produced a 57-point engagement spread with identical layouts and only the copy changed. Bennett Sung ran it, and the finding is that the category converges on abstraction because “strategic transformation” survives internal review and concrete claims don’t.
What One Week and One Prototype Teach Us About HR Tech. Cliff Jurkiewicz spent a week building an HR tech prototype with AI-assisted coding and it exposed the same thing practitioners have been saying for years: most hiring software manages legal risk, not people.
Hiring: Sr. Director, Portfolio Marketing at UKG. John Schneider is hiring at UKG for all my marketing folks.
Hiring: Analyst Relations Director at Workday. AR in enterprise HR tech is an underrated discipline and this role touches most of the major research firms. Jennifer Neumann posted it.
An Independent Resource for Research and HR Tech. Tami Nutt left Jumpstart HR and is available for research and HR tech advisory work.
Open for People Leader Work. T. Tara Turk-Haynes is available as a fractional and advisory resource for people leaders.
A New Chapter with a Benefits Disruptor. Kristy McCann met the Previ team at Transform two months ago, liked what she heard, and joined as a brand ambassador.
Have a great rest of your week!