
The AI Confidence Test 2026: Closing the Gap Between Perception and Capability with Adam Lacey and James Wann (Assemble You)
When assessing workforce AI readiness, tracking software adoption rates and surveying employee sentiment will only tell you half the story. To truly understand organisational risk and opportunity, businesses must measure the gap between how well employees think they understand AI and what they can actually demonstrate.
This week, Brigid McCormack sits down with our very own Co-Founder, Adam Lacey, and Head of Product and AI, James Wann, to unpack the startling findings from the AI Confidence Test 2026. Before building the AI Confidence Method here at Assemble You, Adam and James worked together at Filtered, pioneering machine learning in EdTech long before generative AI was part of the everyday conversation. Now, they have surveyed 1,000 full-time employees to uncover exactly how AI is being used, and misunderstood, in the modern workplace.
Adam and James break down the massive discrepancies revealed in our new Alignment Index report, noting an average 22-point gap between perceived and demonstrated AI capability. They reveal why daily power users barely outscored people who had never used AI before, why the IT sector displayed the largest capability gap (81% confidence versus 42% demonstrated capability), and why 1 in 4 employees admitted to pasting confidential company information into public AI tools.
They also explore why younger generations (Gen Z and Millennials) are twice as likely to take data security risks compared to Gen X, why simply knowing a rule isn't enough to change employee behaviour, and why L&D teams must pivot from measuring software engagement to actively testing foundational AI capability.
What you'll learn in this episode
- Why daily AI usage is a flawed metric for capability: daily users scored 46% on tests, while non-users scored 44%
- The dangerous gap in technical understanding: why only 19.7% of users can accurately explain how ChatGPT generates answers (next token prediction) despite 72% claiming they could
- How industry and age impact risk: why IT professionals and younger generations are significantly more likely to paste confidential data into public AI tools
- Why traditional compliance training fails to stop risky AI behaviour without explaining the "why" behind the technology
- Three actionable steps for L&D: testing true capability, auditing confident teams first, and building on positive workforce sentiment
- Details on Assemble Youās offer to run the Alignment Index for 50 employees at your organisation for free
Resources mentioned in this episode
- Read the AI Confidence Test 2026 report: https://www.assembleyou.com
- You can also find it on the Assemble You homepage at assembleyou.com
- Put 50 of your people through the Alignment Index for free: https://www.assembleyou.com/alignment-index-free-test
- Try a free demo of our AI capability training: https://www.assembleyou.com/book-an-ai-confidence-call
The Assembly DebriefĀ
A short, practical email unpacking the best ideas from each episode, designed for modern learning leaders who want insight without the noise. https://eu1.hubs.ly/H0xWSjW0
Come and connect with us!Ā
We're always keen to hear what's resonating or what you're seeing in your world.
Adam Lacey - https://www.linkedin.com/in/adamlacey/
Brigid McCormack - https://www.linkedin.com/in/brigidmccormack/
James Wann - https://www.linkedin.com/in/james-wann/
More from The Assembly & Assemble YouĀ
Learning that feels more like great content than training. Explore podcast-style lessons, video shorts, and series designed for real work, plus the AI Confidence Method, a workforce AI change programme that measures what your people really think about AI and then changes it. https://eu1.hubs.ly/H0xWSkk0
Privacy & DataĀ
This podcast is hosted by Transistor. For more information on how your data is handled, please visit: https://www.transistor.fm/privacy/
More episodes from "The Assembly"



Don't miss an episode of āThe Assemblyā and subscribe to it in the GetPodcast app.








