The BeAIReady Brief | Week 22
May 25–31 | Uber Can't Prove Its AI Budget Was Worth It, Leadership Expectations Aren't Matching the Worker Reality, and HR's Dirty AI Hiring Secret
The April Personal Consumption Expenditure (PCE), a measure of US inflation, is up 3.8% year-over-year — the highest point since May 2023 and accelerating — effectively taking rate cuts off the table through 2027. That’s important, because we’re in an economy where the cost of running organizations is already stubbornly high, the payoff from AI is stubbornly hard to prove, and the people doing the actual work are less confident, more burned out, and more skeptical than at any point in recent memory. And now, the CEOs most responsible for AI’s hype cycle publicly admitted they were wrong about what AI would do to jobs.
Organizations are adopting AI at scale, at a pace that is often faster than they planned. Leaders are pulling the brakes and asking if any of it is working… for the organization… or for the people inside them.
This week’s coverage:
Uber Can’t Prove It, and the CEOs Who Started the Hype Are Admitting It
Uber burned through its entire 2026 AI budget in four months and still can’t draw a line from spend to value — and the CEOs who spent the last year warning about an AI jobs apocalypse just admitted they were wrong about that too.
Using AI Doesn’t Mean They Trust It
A majority of American workers now use AI on the job — and a large majority of them don’t trust what it produces, a combination that should worry every executive who equates deployment with progress.
Leadership Expectations Aren’t Matching the Worker Reality
A global study of 2,000 C-suite executives finds that leaders expect AI agents in workflows within a year, but fewer than a quarter are confident their organizations can actually get there.
The Human Cost Nobody Budgeted For
Two pieces this week pushed back on the productivity narrative with a more uncomfortable question: what does sustained AI use cost workers relationally, and what does that cost organizations over time?
HR’s Dirty AI Hiring Secret
The largest independent study of AI hiring algorithms ever conducted found clear racial disparities affecting tens of thousands of applicants — and the vendor’s own audits had been designed in a way that made those disparities invisible.
On the Bigger Picture
Microsoft’s May Copilot Studio update includes some genuinely consequential capabilities for enterprise automation — particularly for organizations still managing legacy systems without APIs.
Here’s what I was reading.
Uber Can’t Prove It, and the CEOs Who Started the Hype Are Admitting It
Uber’s story is useful precisely because it doesn’t involve a struggling company making excuses. This is one of the most AI-forward businesses in Silicon Valley — a company whose core product runs on AI — and its COO spent the week publicly questioning whether the money was worth it. According to reporting in Fortune, Uber burned through its entire 2026 AI coding budget in just four months after incentivizing adoption through an internal leaderboard that ranked teams by AI tool usage. By the time the COO started asking questions, 95% of engineers were using AI tools and 70% of code was AI-generated. The problem wasn’t adoption. It was that no one could draw a line from those numbers to anything that actually served users. “That link is not there yet,” COO Andrew Macdonald said. “It’s very hard to draw a line between one of those stats and ‘Okay now we’re actually producing 25% more useful consumer features.’” (Uber burned through its entire 2026 AI budget in four months. Now its COO is questioning whether it’s worth it)
What makes this particular moment important is that it coincides with a notable reversal from two of the people most responsible for shaping expectations about AI’s economic impact. Sam Altman and Dario Amodei — who between them spent the better part of last year warning that AI would eliminate entry-level white-collar jobs at scale — both walked back those predictions in public this week. Altman told an audience he was “pretty wrong” about AI’s impact on employment so far. Amodei reframed automation not as a destroyer of jobs but as a multiplier of output: if AI automates 90% of a job, the remaining 10% “expands to be 100% of what people do.” Goldman Sachs CEO David Solomon, who was skeptical of the apocalyptic framing all along, pointed to 145% employment growth since 1962 as evidence that technological disruption doesn’t follow the script its proponents predict. The reversals matter not just as data points about AI’s actual impact on labor, but as a signal about how much of the AI narrative has been shaped by founders who had powerful incentives to generate a sense of urgency. (Sam Altman and Dario Amodei are both walking back AI jobs apocalypse predictions)
The Uber and Altman stories land at the same pressure point. Organizations spent 2025 buying AI under the twin pressures of competitive fear and leadership hype — and now the ROI question is being asked in earnest, by finance teams, by boards, and increasingly by the executives who made the calls.
Using AI Doesn’t Mean They Trust It
A Quinnipiac University poll released this week shows that over half of Americans are now using AI to research topics, and nearly a third of employed adults are using it for their jobs. But 76% of those Americans say they can trust AI-generated information only “some of the time” or “hardly ever.” The math is uncomfortable: a substantial and growing portion of the American workforce is regularly deploying a tool it doesn’t trust to produce outputs that inform business decisions. That isn’t an adoption problem. As the article puts it, it’s a judgment and governance problem — one that sits at the intersection of workforce capability, organizational culture, and the credibility of the executives overseeing AI deployment. (Workers Are Using A.I. They Don’t Trust. That’s a Problem for the C-Suite)
The generational finding is actually most interesting to me. Gen Z — the cohort that was supposed to be AI’s natural champions inside the enterprise, the fluent digital natives who would pull adoption forward — is the most pessimistic of any age group about what AI means for their employment. 81% percent of Gen Z respondents believe AI will lead to a decrease in job opportunities, compared to 66% of baby boomers. This inverts a narrative that has driven a lot of AI change management strategy: the assumption that younger workers would lead enthusiasm from below. The Quinnipiac data suggests that familiarity with the technology has produced not enthusiasm but a clearer-eyed reading of where the risk actually falls — and entry-level workers know they’re in the exposure zone.
The backdrop for all of this showed up in Glassdoor data released this week – burnout rates are up 65% year-over-year, and employee confidence is at a record low. Only 43.8% of employees are reporting a positive six-month business outlook. Workers who mentioned burnout in their reviews were 76% less likely to give overall positive ratings, and 78% less likely to recommend their companies to friends. The technology sector posted the largest confidence drop of any industry, down 9.7 percentage points year-over-year. The AI productivity narrative is landing inside organizations where the people AI is supposedly helping are struggling in ways that don’t show up in token usage dashboards. (Burnout is increasing, while employee confidence is at a record low, research shows)
Leadership Expectations Aren’t Matching the Worker Reality
The Adecco Group surveyed 2,000 C-suite executives across 13 countries for its annual workforce research, and the headline number is one that won’t surprise anyone who has been reading these trends: 45% of business leaders expect AI agents to be integrated into workflows within the next 12 months, but only 30% of workers share that expectation. What’s new isn’t the gap — it’s how wide the failure of communication is underneath it. Only 36% of leaders say their talent strategy clearly demonstrates that AI will create opportunities for employees rather than eliminate them. Only 39% are involving employees directly in job redesign. Only 22% say they are highly confident their organizations are developing the future-ready capabilities needed to keep pace. (Global Study Finds Widening Gap Between AI Ambition And Workforce Readiness)
Adecco’s CEO put it directly: “AI may move at software speed, but organizational trust moves at human speed. Companies that ignore that gap will struggle to turn pilots into performance.” What the research also shows — and this is the part that tends to get overlooked in the readiness conversation — is that workers may actually be more ready than leaders assume. Seventy percent of workers said they feel ready to collaborate with AI agents. Only 39% of leaders believe their employees would be comfortable with agents. Leaders are underestimating their workforce while simultaneously failing to communicate, invest in change management, or redesign roles in ways that would make readiness real. The confidence gap isn’t just between the C-suite and the rest of the organization. It’s between what leaders say they want and what they’re actually doing to get there.
The Human Cost Nobody Budgeted For
Two pieces this week approached the human cost of AI adoption from different angles and arrived at the same concern. Workday’s Human Connection Workplace Index — a survey of 2,150 employees at large enterprises who actively use AI — found that while AI is reducing burnout and boosting productivity for most respondents, it is also creating what the report calls a “connection deficit.” Sixty-two percent of employees say their stress or burnout risk has decreased since using AI. But 20% of Gen Z workers took time off this year due to loneliness or isolation — and Gen Z is 12 times more likely than Gen X to report feeling completely disconnected from colleagues. The finding that stands out most is that 37% of respondents have turned to AI for companionship, citing its judgment-free, always-available nature as reasons they prefer it to colleagues. That is not a productivity metric. That is a sign of something structural shifting in how people at work relate to each other. (New Workday Global Research Finds AI is Easing Burnout but May Be Deepening a Connection Deficit at Work)
A Fast Company piece from the same week took a wider lens. The author — who has spent two decades working with leaders across 20 countries — argues that the AI conversation’s focus on jobs and roles is obscuring a more consequential shift: the dismantling of the relational infrastructure that work has always provided. Entry-level roles aren’t just where people learn technical skills; they’re where people learn to navigate difficult colleagues, earn trust without authority, read a room. Knowledge isn’t just what people know; it’s forged through mentorship and peer relationships that pressure-test ideas. “Implementing AI is not a technology problem. It’s a people problem. It always is.” The quote is from Charlene Li, cited in the piece, and it points at something organizations are systematically underprioritizing: the leaders who are struggling most with AI are those who built authority on knowing more than everyone else, and AI just democratized what they were hoarding. The leaders thriving are the ones who built authority on relationships and trust. (AI may be eating jobs, but it poses an even bigger threat)
HR’s Dirty AI Hiring Secret
The most important research I found last week came from researchers at Stanford, Chapman, and Northeastern, and it documents something that organizations deploying AI hiring tools should find genuinely alarming. The study — described as the largest independent analysis of AI-powered hiring algorithms ever conducted — examined more than 4 million job applications across 156 employers, all screened by the same vendor’s algorithms. The finding: more than 25% of all applications submitted by Black job seekers were directed to positions where the algorithm produced outcomes that trigger federal discrimination scrutiny. The vendor, Pymetrics, had conducted its own bias audits and found no problem — not because the analysis was dishonest, but because it was measuring at the wrong level. Pymetrics pooled all applicants and outcomes across all employers and positions. The researchers analyzed each of the 1,746 individual positions separately, which is how U.S. employment discrimination law is actually designed to be applied. When you do it right, 10.62% of positions show adverse impact on Black applicants. (Largest study of AI hiring algorithms to date finds ‘clear racial disparities’ — over 25% of Black applicants tainted by bias)
The second finding may be even more consequential for organizations using these tools. Because algorithms produce the same output for the same input every time, and because scores are stored and reused across employers, a candidate who gets screened by Pymetrics at one company isn’t really getting a fresh evaluation at the next company that uses the same platform. Researchers documented what they call an “algorithmic blackball” — where applicants rejected once are effectively rejected by all employers using the same vendor, without knowing it. The implication for any organization using a third-party AI screening platform is direct: you may be discriminating at scale, your vendor’s audit almost certainly used the wrong methodology, and the EU AI Act’s compliance requirements for hiring algorithms go into effect August 2. The gap between what organizations believe their AI tools are doing and what they are actually doing is not a hypothetical governance risk. It is a documented one.
On the Bigger Picture
Microsoft’s May Copilot Studio update is worth reading even for organizations that aren’t deep into the Microsoft ecosystem, because it marks a meaningful expansion in what enterprise automation can actually reach. Computer-using agents — which can now interact directly with websites and desktop applications through the UI, without requiring API access — are now generally available. This matters for the specific, persistent problem that has defeated a lot of automation efforts: legacy systems and vendor portals that were never designed to be automated. A company like Graebel, featured in the release, can now automate its relocation request processing end-to-end, even though its proprietary platform lacks API support, by building an agent that navigates the UI the way a human would. The May update also brings agent-to-agent communication to general availability, a new visual workflow designer for building multi-step agentic processes, and Work IQ REST API and MCP server support for connecting agents across enterprise systems. (What’s new in Copilot Studio: May 2026 updates and features)
The people closest to AI — the workers using it daily, the Gen Z employees whose entire careers will be shaped by it — are the least confident about what it means for them, and the most likely to be quietly paying a cost that doesn't show up in any dashboard. Altman admitted he was wrong. Uber's COO admitted the math doesn't close. A Stanford team proved the hiring algorithm was discriminating in ways the vendor's own audit couldn't detect. All of that happened in the same week that burnout hit a 65% year-over-year increase and worker confidence hit a record low. Organizations are not running out of data here. They are running out of reasons to keep treating the human side of AI adoption as someone else's problem to solve later.
That’s it for this week’s BeAIReady brief!
If you appreciate the depth of reporting and how I connect the dots, please like, share this post, and subscribe (or share the Brief with a friend!). Thanks!
~erick