A Deloitte survey of 25,000 workers finds that 31% have used AI tools without their employer's knowledge. Sixty-three percent knowingly use generative AI for work. Half have had no formal workplace training.
Most organisations will read this and reach for a policy document. That response misreads what is actually happening.
When people use tools covertly, they are solving real work problems without official support. The tools work well enough to justify the friction. People are finding what works and using it without institutional backing. The question to ask: what has the organisation built that makes reviewed, consistent AI use easier than shadow use? For most organisations, the honest answer is: nothing yet.
That structure has three parts.
Context. AI produces generic output when it gets generic input. What transforms a general-purpose tool into something useful for your team is what you put in front of the prompt: your organisation's terminology, your content standards, the type of document or deliverable you produce. This is a professional standards step that most AI rollouts skip entirely.
A review process. The LearnUpon 2026 AI in L&D survey found that 86% of learning and development professionals agree human review is essential. Only 26% have a consistent vetting process. Adoption without review creates exposure rather than efficiency. The human judgment step is not only about catching errors. It is about maintaining the professional standards the organisation is responsible for.
Shared output standards. Without a shared definition of what good output looks like, every person on the team runs their own quality bar. This is how you get AI-assisted work that varies widely even when the brief is the same.
These three things are the adoption layer that most enterprise rollouts never build. The tools get introduced. The guidance arrives too late, too general, and too thin to change how people actually work.
An L&D team redesigning onboarding content for more than 400 employees came to AI Content Ops at exactly this stage. The tools were already in use. The gap was the review layer. We rebuilt their content production workflow around AI-assisted drafting with a structured manager sign-off step, reducing per-module production time while keeping SME review intact. The programme ran over two days, and they left with a framework they now own and operate without us.
If you are at the stage where adoption is running ahead of governance, AI Content Ops works specifically on the layer that makes adoption stick: context, standards, and review built around your team's own tools and content. aicontentops.online
What is the one thing your team is doing consistently with AI, and the one thing that keeps varying?
Read the original report here: GenAI Workforce Survey