Shadow AI is a signal employers should act on

British workers are spending an estimated £958 million a year of their own money on generative AI tools for work, according to Deloitte. Nearly a third of those using the technology do so without their employer's knowledge.

That should prompt employers to examine what their staff are missing. People are finding tools that help them get through their workload and paying for access themselves. There is demand here that organisations have an opportunity to meet.

There are also risks. When employees use personal accounts for business tasks, employers can struggle to see what information they share, how they use the outputs and which decisions depend on them.

Training and clear policies help. Employers also need to provide approved alternatives that people find useful. If the official option is difficult to access or poorly suited to the task, staff have a reason to keep using their own tools.

The controls should reflect the work. Drafting a routine email requires a different level of oversight from reviewing sensitive customer records or authorising a payment. Organisations need to define what information AI can access, what actions it can take and when a person must approve its next step.

Those rules need to operate within the systems doing the work. Businesses should be able to trace actions, investigate mistakes and stop a process when it exceeds its authority.

Sensitive workloads may also require infrastructure under the organisation's control.

At Deliverance AI, we see these requirements as part of making agents useful in everyday operations. Clear boundaries give businesses greater confidence to put them to work.

The practical starting point is to understand where employees already find value. Build approved workflows around those needs, make them easy to use and measure the results. Employers can then expand access with evidence of what works and appropriate oversight.