The AI and Work Signals That Mattered This Quarter
AI Noise: 2026 saw overwhelming noise about AI at work, but meaningful discussions are much narrower.
Adoption Friction: Adoption is nearly universal, but extracting value from AI remains uneconomical for many companies.
Regulatory Changes: Colorado repealed its AI law, opting for disclosure measures over stringent regulations starting 2027.
Entry-Level Decline: AI contributes to decreased entry-level hiring, highlighting challenges in career development.
Governance Shift: AI governance increasingly involves senior leadership, improving organizational value capture efforts.
The first half of 2026 produced more noise about AI at work than any quarter on record, and most of it was the same press release with different logos. Adoption is near universal, returns are not, everybody has a survey.
Strip out the recycled statistics and a smaller set of signals remains, the ones that actually moved between April and June and that will still matter in September. Four of them stand out, and they share a spine.
Each one is a test of whether the people running organizations treat AI as a problem of headcount or a problem of stewardship, and this quarter the evidence on which way they are leaning got a lot less flattering.
Keep Reading—and Keep Leading Smarter
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Have an account? Log In
Adoption Numbers Stopped Being the Story
For two years the headline was the adoption rate, and the adoption rate is now boring. Roughly nine in ten organizations use AI somewhere, a figure that has climbed past the point of being interesting.
What changed this quarter is that the major research finally caught up to what anyone inside a company already felt, which is that getting the tool in the building and getting value out of it are separated by a wide and expensive distance.
Writer's 2026 enterprise survey, released at the start of the quarter, put hard numbers on the friction. A majority of C-suite executives said AI adoption is tearing their company apart, and three quarters admitted their AI strategy functions more as theater than as direction.
Only 29% of companies reported significant returns despite most spending past a million dollars a year. The more striking finding sat underneath those headlines. Ninety-two percent of the C-suite said they are cultivating a class of favored "AI elite" employees, and 60% said they plan to push out the people who do not adopt fast enough.
Meanwhile, a meaningful share of employees, and nearly half of Gen Z, admitted to actively undermining their company's AI rollout.
That is not a productivity problem. It is a trust problem dressed as one. BCG's research this quarter described a "silicon ceiling," with three quarters of managers using generative AI weekly against roughly half of frontline staff.
ManpowerGroup's Global Talent Barometer found regular AI use rising even as worker confidence in using the technology fell sharply, the kind of divergence that tells you adoption is being mandated faster than it is being supported.
The signal worth carrying into Q3 is that the companies treating low adoption as an employee compliance failure are reading their own data backward. People do not sabotage tools they trust the people deploying them.
The Toughest AI Employment Law?
The regulatory story this quarter ran the opposite direction from where it was pointed.
Colorado's Artificial Intelligence Act, the most comprehensive state law governing AI in consequential decisions including employment, was supposed to take effect June 30. Instead, on May 14, weeks before the deadline, Governor Jared Polis signed SB 189, a bill that repealed and replaced it.
The original law's spine, a duty of care requiring deployers to use reasonable care against algorithmic discrimination, risk management programs, impact assessments, reporting to the attorney general, all of it is gone. What replaces it is a lighter regime built around disclosure and notice, and it does not take effect until January 1, 2027.
What survives is real but narrower. Employers using automated tools to materially influence a hiring or other consequential decision will have to tell people the technology is in use, give them an adverse-action process with a path to human review, and keep records for three years.
Enforcement sits only with the attorney general, who has rulemaking to finish before the law goes live, and penalties run to $20,000 per violation with a cure period attached. No private lawsuits.
The federal pressure was real. The December executive order singled Colorado out, the DOJ's AI Litigation Task Force began operating in January, and xAI had already sued over the algorithmic discrimination provisions with the DOJ moving to intervene.
The legal consensus is still that an executive order cannot preempt state law on its own, and the Senate killed a moratorium attempt 99 to 1. The point that should hold a leader's attention is that none of that machinery is what defanged the Colorado law. Its own legislature did, weeks before the deadline, under a pro-innovation current that found bipartisan footing in a state that wrote the toughest rules in the country two years ago.
For the HR and operations leader, the practical read is that the compliance floor just dropped, and that is exactly the wrong reason to relax. The disclosure-and-notice obligations are coming, and the broader pattern across states is moving toward transparency requirements rather than away from them.
More to the point, the legal duty to test whether an AI hiring tool screens people out unfairly is not the same thing as the responsibility to do it. The law stopped requiring the audit. The applicant getting filtered by an unexamined model does not care which.
Erin Bortz, manager of recruiting at the cybersecurity firm Huntress, put the principle in terms that do not depend on a statute at all.
Regardless of laws and regulations, any decisions that affect a candidate directly should ultimately be owned by a human.
AI can surface the candidates who fit a role, but in her account the tool is only as good as its inputs and cannot be fully trusted with the call.
When AI capability jumped this year and expectations around its use unsettled her colleagues, Bortz said leadership at Huntress responded by standing up an open-participation AI council to set guidelines rather than issuing a mandate from the top, a choice about who gets a say that maps onto the same trust question the friction data raised.
The First Rung of the Career Ladder Kept Disappearing
The most human signal of the quarter was also the most documented. Entry-level hiring continued its decline, and the research linking that decline to AI moved from suggestive to difficult to dismiss.
The Stanford study using ADP payroll data found employment for workers aged 22 to 25 in the most AI-exposed roles, software development, customer service, fell roughly 6% from late 2022 to mid-2025, while older workers in the identical roles gained 6 to 9%.
Entry-level postings overall are down about 35% since early 2023 by Revelio Labs' count. New graduates made up just 7% of big-tech hires in 2024, half the pre-pandemic share.
The Stanford finding that deserves the most attention is the distinction between automation and augmentation. Where AI automated a task outright, junior hiring fell. Where it augmented human work, employment held or grew.
That is the whole argument in one data point. The damage to early careers is not an inevitable property of the technology. It tracks a choice about how the technology gets deployed, and right now the dominant choice is the one that hollows out the bottom of the organization.
This is where the stewardship question stops being abstract. Removing the first rung does not just hurt the class of 2026. It severs the mechanism by which institutional knowledge transfers and by which a company grows its own senior people.
Leaders treating entry-level automation as a clean cost saving are booking a near-term margin against a long-term capability they will have to repurchase later, probably at a premium, possibly from a labor pool they stopped developing.
Justina Raskauskiene, HR lead at the marketing software company Omnisend, sees replacing junior specialists with AI tools as tempting, but ultimately short-sighted.
Without a junior level, you’ll become increasingly brittle. Juniors bring a fresh outlook that becomes necessary the moment a novel problem shows up.
Her team still hires early-career people where there is demand for them. The change is what those people do once they arrive. Instead of the tasks AI now does faster, Raskauskiene said her juniors focus on verifying, questioning, and improving its output, which is the augmentation path the Stanford data associates with stable employment rather than the automation path that erases it.
Governance Moved Up the Org Chart
Against three signals that should worry anyone with a human-centric view of the workplace, one cut the other way. Deloitte's 2026 State of AI in the Enterprise report found that organizations where senior leadership actively shapes AI governance capture meaningfully more value than those handing the work to technical teams alone.
Workforce access to sanctioned tools jumped from under 40% to around 60% in a single year. The maturity gap is real, usage among those with access still lags, but the direction is that governance is climbing toward the C-suite rather than staying buried in IT.
That matters because every other signal this quarter is downstream of who owns the decision. The trust collapse, the compliance exposure, the gutted entry ramp, all of them are what happens when AI deployment is treated as a procurement exercise rather than a leadership responsibility.
The encouraging read is that the people best positioned to treat it as the latter are starting to take the seat. The honest read is that most of them have not yet, and the quarter's other three signals are the receipts.
What proved out between April and June is that the AI-at-work conversation has finally moved past whether to adopt and onto the harder question of who absorbs the cost of adoption.
So far it has fallen heaviest on the people with the least power to refuse it, the frontline worker handed a mandate without support, the applicant screened by a tool nobody audited, the graduate locked out of a first job.