SHRM26 Debrief · AI, Data & Tech · #2510

2026 State of AI in the Workplace

with Kenny Pyle

▶ Watch on the SHRM portal ~56 min, distilled

AI adoptionAI use policyshadow AI use

"34% of workers have violated AI policy and said that they uploaded confidential company information into an unauthorized AI tool. Oh, yeah. Think about that for a second. That is terrifying."

What it was about

AI adoption at work is real but far messier and less uniform than headlines suggest. Usage, policy compliance, aptitude, and impact vary sharply by job level, and organizational culture matters as much as the technology itself in determining whether AI integration succeeds.

By the numbers

2 out of 3 vs 6%

workers with high trust in leadership reporting higher job engagement vs those with low trust

34%

policy violators who uploaded confidential company info into an unauthorized AI tool

individual contributors gain ~1 hour; managers and directors gain ~3 hours

net time saved after subtracting correction time, by job level

Key notes

Develop and actively communicate a clear, accessible AI use policy, since only 47% of organizations have one and 30% of workers admit to violating whatever policy exists.
Build targeted, ongoing AI training programs at every job level rather than one-off sessions, since self-reported AI experience does not match actual aptitude test performance.
Reinvest the time AI saves (workers report saving about six hours per week) into complex problem-solving, creative work, and skill development rather than assuming it disappears into extra output.

The contrarian takeDirectors and senior leaders reported the highest self-assessed AI expertise and were the most frequent AI users, yet they produced the most AI slop and scored no better than other job levels on an objective AI aptitude test. That suggests leadership's AI confidence is largely overconfidence rather than earned skill.

Take this back Monday

Do this for your team

Publish a plain-language AI use policy and walk through it with your team, since 30% admit violating unwritten or unclear rules.

Say this in your next leadership meeting

Our biggest AI risk isn't job loss, it's the 34% of employees pasting confidential data into unauthorized tools because we never published a policy.

Watch out for

Assuming leadership's self-reported AI expertise is accurate: directors and above overestimated their skill while actually producing the most AI slop.
Relying on negative consequences or informal, one-off training sessions to drive adoption, which the data shows are the least effective tactics.
Failing to communicate AI policy evenly across levels: only one in three individual contributors were informed about AI implementation beforehand, versus 74% of directors.

Fun fact · Kenny Pyle

Before HR tech analysis, Kenny Pyle's career spanned founding a startup, a Fortune 500 company, academia, and the entertainment industry.

If this landed, see these

↳ Go deeperReturn on Intelligence: Move AI ROI Beyond Cost to Competitive AdvantageTurns messy adoption data into an ROI framework for treating AI like a hire.⇄ The counterpointPause, Presence & Trust: Slowing Down to Speed Up in the Age of AIArgues AI adoption depends on human trust and pace, not just usage stats.✦ The unexpected oneAfter The Complaint: The Decisions That Create Retaliation LiabilitySame lesson as retaliation liability: the risk is what follows, not the original gap.