Don’t Force Employees Into the Dark
While your internal AI system may check all the boxes, have you forced employees to outsource actually how to use it?
Event / Source
Real-world workflow discovery · Global enterprise employee · September 2026
What it was about
A colleague at a global shoe brand came to me with an increasingly common problem: her employer had built an internal AI system and encouraged employees to use it, but had provided very little guidance about how AI should actually change the way they work.
The AI can summarize meetings and generate transcripts. But employees are still manually checking Jira tickets, repeating information across meetings, chasing follow-ups and relying on individuals to remember what happened.
They had the technology but had not learned or understood how to make it a real part of their daily habits.
What I’m keeping
Giving employees AI is not the same thing as teaching them how to work with AI.
There’s actually emerging research around this problem. Researchers have begun using the term “shadow learning” to describe employees developing AI capabilities outside their organization’s formal learning systems.
That’s different from “shadow AI,” where employees use unauthorized tools.
An employee can be using the company’s approved AI and still have to figure out, on their own, how to make it useful.
3–5 things I learned
Access isn’t adoption. Providing an AI tool doesn’t tell employees which problems to use it for, when to use it or how their existing workflows should change.
AI training needs to be tied to actual work. A generic introduction to prompting isn’t enough. Someone managing projects needs examples involving meetings, Jira tickets, decisions, deadlines and follow-ups.
The workflow may be the real problem. If people routinely lose decisions, repeat instructions or avoid ownership, adding AI on top of that process doesn’t fix it.
Employees will build their own systems when the organization doesn’t. That can produce useful innovation, but the company may never capture or spread what its most capable employees learn.
AI enablement should include ways of working. The useful question isn’t only “How do I use this AI?” It’s “What should we now do differently because this capability exists?”
How this changes my work
This changed how I think about AI adoption.
I’m less interested in starting with “What can we automate?” and increasingly interested in asking people where work is breaking down.
Where are you repeating yourself?
What are you constantly searching for?
What requires someone to chase five other people before anything happens?
Sometimes the answer will be an AI workflow. Sometimes it will be automation. Sometimes it may simply be a better operating practice.
The best discovery is finding the friction first. Then the technology comes second.
The Federal Reserve’s 2026 research found exactly this gap: workers value AI training, but a significant share of employers still aren’t providing either adequate access or training. (1)
Workers who feel confident using AI use it 2.6x more frequently, and their weekly time savings nearly double. Yet 62% of surveyed state/local government workers had never received AI training, while another workforce survey found 62% rated their organization’s AI training average or poor. (2)
Researchers call the phenomenon “shadow learning.” Importantly, they explicitly distinguish it from shadow IT. Employees can be using an approved internal AI system and still engage in shadow learning because they’re developing the actual skills and ways of working outside the organization’s formal training structure. (3)
Sources
(1) Federal Reserve research on workplace GenAI and training. By Ali Hashim, Gizem Kosar, Wilbert van der Klaauw. July 14, 2026. https://www.frbsf.org/research-and-insights/publications/system-research-new-york-fed/2026/07/use-of-gen-ai-in-workplace-value-of-access-to-training/
(2) Making AI training work in government: A framework for tailored learning. Maximizing gains from AI calls for an AI-fluent workforce, which starts with government leaders matching training to employees’ roles, skills, and contexts. By Joe Mariani, Lucy Melvin, Chris Foss, Josh Rachford, Samantha Roberts 06 May 2026
Deloitte Center for Government Insights https://www.deloitte.com/us/en/insights/industry/government-public-sector-services/tailored-ai-training-for-government.html
(3) Shadow Learning and Knowledge Redistribution: How Workers Appropriate Generative AI Outside Formal Organisational Systems August 12, 2026 Eduardo Carlos Dittmar, Martin Sposato https://onlinelibrary.wiley.com/doi/full/10.1002/kpm.70145