Why AI adoption stalls, and what gets it moving.
Most companies have given their people AI tools. Far fewer have made using them a daily habit. This brief pulls together public research on what separates the two, and what we think enablement teams should do about it.
For AI enablement and people leaders
·
4 min read
·
Published Nov 3, 2026 · XTrace research team
Research brief
From access to a daily habit.
In short
The gap is between leaders and the frontline.
Leaders use AI often. Many frontline employees have stalled, and most say they weren’t properly trained.
Coaching beats courses.
Regular use was much higher where people got real training time and in-person coaching, and strong leadership support went with far more positive attitudes. (BCG, 2025)
Feeling faster isn’t proof.
Self-reported productivity gains are unreliable. You have to look at how the work actually went.
The adoption gap
Access isn’t the problem. Habit is.
In BCG’s 2025 survey of more than 10,600 people across 11 countries and regions, over three-quarters of leaders and managers said they use generative AI several times a week, while regular use among frontline employees had stalled at about half (51%). Only about a third of employees said they had been properly trained.
Microsoft’s 2025 Work Trend Index shows the same split: leaders report regular AI use far more often than employees (69% vs. 45%).
Regular use of AI at work, by role
Share of respondents
BCG · AI at Work 2025
Leaders and managers
Over 75%
Frontline employees
51%
Gap of 24+ points
Microsoft · Work Trend Index 2025
Leaders
69%
Employees
45%
Gap of 24 points
Redrawn by XTrace from published survey results. The surveys ask different questions, so compare the gaps, not the bars across surveys.
SOURCE BCG, “AI at Work 2025: Momentum Builds, but Gaps Remain”, June 2025 · Microsoft WorkLab, Work Trend Index 2025, April 2025.
What moves it
Training time, coaching and leadership support move the needle.
BCG found regular use is markedly higher among employees who get at least five hours of training and access to in-person training and coaching. BCG also reports that employees who said they had strong leadership support were far more likely to feel positive about generative AI (55% vs. 15%), yet only about a quarter of frontline employees said they had that support.
Managers expect this to become their job. In Microsoft’s 2025 survey, 51% said AI training and upskilling will be a key responsibility for their teams within five years, and 35% were considering hiring AI trainers in the next 12 to 18 months.
Employees who feel positive about generative AI
Each dot is 5% of employees
Without strong leadership support
15%
With strong leadership support
55%
Only about a quarter of frontline employees said they had strong leadership support. Redrawn by XTrace from published survey results.
SOURCE BCG, “AI at Work 2025”, June 2025 · Microsoft WorkLab, Work Trend Index 2025, April 2025.
Measuring it
How it felt isn’t how it went.
In METR’s early-2025 controlled trial, experienced developers believed AI had made them about 20% faster, while their measured task time went up. METR has since said those early-2025 results don’t reflect today’s tools, and its 2026 follow-up leans toward a speed-up, with low confidence. What still holds: METR treats self-reported speed-ups as unreliable.
Microsoft found people split on how they think of AI: 52% see it as a tool you give simple commands to, 46% as a thought partner. Usage counts can’t tell those two groups apart.
Perceived vs. measured speed, METR early-2025 study
Change in task time with AI
How fast it felt
~20% faster
How it went, measured
~19% slower
30% slower
No change
30% faster
Dated result, shown for the gap between perception and measurement. METR says its early-2025 numbers don’t reflect today’s tools; its 2026 update leans toward a speed-up, with low confidence.
SOURCE METR, early-2025 study of experienced open-source developers, July 2025 · METR, uplift update, Feb 2026 · Microsoft WorkLab, Work Trend Index 2025.
What this means for you
Four moves for an enablement program.
Measure the work, not the logins.
Track where AI work stalls and where people step in, not just who opened the tool.
Coach on real work.
Train on your teams’ own tools and tasks, not generic courses.
Help in the moment.
Bring guidance into the work itself, when people are stuck.
Lock in the gains.
Turn what works into shared team standards so progress doesn’t fade.
Sources
For anyone checking our work.
BCG, June 2025
“AI at Work 2025: Momentum Builds, but Gaps Remain”
Microsoft WorkLab, April 2025
Work Trend Index 2025: “2025: The year the Frontier Firm is born”
METR, July 2025
Early-2025 study of AI’s effect on experienced open-source developers
METR, Feb 2026
Uplift update
Put it into practice
See how XTrace puts this into practice.
XTrace shows where teams get stuck and which standards to roll out. The XTrace plugin coaches people during the work, inside their AI tools, and adoption progress is clear at a glance. Our hands-on workshops and expert support help teams move faster.