Shadow AI Could Cost UK Businesses Millions in Unused AI Subscriptions

New modelling from Marketing Signals estimates that forgotten AI subscriptions could cost UK marketing departments £63.7 million a year, as employees increasingly buy AI tools outside approved software budgets.

Artificial intelligence is creating a new type of software spending problem inside businesses.

While “shadow AI” is usually discussed in terms of cybersecurity and data governance, new research from Marketing Signals suggests there is also a significant financial cost. Employees are increasingly paying for AI tools themselves, often through personal cards or individual expense claims, leaving finance and IT teams with limited visibility over what the business is actually using.

According to the company’s Subscription Creep Report, spending on individually purchased AI tools has increased by 108% year on year. These subscriptions include copywriting assistants, image generators, transcription services and other AI applications that often cost only $20 to $40 per month.

Individually, the charges may appear small. Across hundreds or thousands of employees, however, they can quickly become a much larger recurring expense.

Up to 71% of Employees Are Using Unapproved AI Tools

Marketing Signals points to research suggesting that between 50% and 71% of employees use AI tools that their employer has not formally approved. Separate research cited in the report also found that 98% of executives admitted bypassing IT when purchasing technology.

This creates a growing visibility gap.

Businesses may have formal contracts for major enterprise platforms while dozens of smaller AI subscriptions sit across expense claims, individual cards and departmental budgets.

In the underlying data examined by Marketing Signals, ChatGPT emerged as the most expensive individual software purchase, ahead of traditional SaaS tools.

The issue is not simply that companies are buying more AI. It is that many organisations may not know how many tools they are paying for, who is using them or whether several subscriptions perform the same function.

Forgotten AI Subscriptions Could Cost UK Marketing Teams £63.7 Million

Within UK marketing departments alone, Marketing Signals estimates that forgotten AI subscriptions cost businesses around £63.7 million annually.

The company gives a wider estimated range of £34 million to £112.1 million.

Those figures cover forgotten AI subscriptions rather than the wider software stack, meaning the total cost of unused and duplicated technology could be considerably higher.

The problem is also getting worse. Marketing Signals cites Vertice data showing that combined SaaS waste has increased from 62% to 65% of all licences over the past year, while spending on AI-native tools continues to accelerate.

For businesses trying to control technology costs, this creates an unusual contradiction: AI may improve productivity while simultaneously adding a new layer of poorly monitored spending.

Why Small AI Subscriptions Are Easy to Miss

Harry Nisbet, General Manager at Marketing Signals, argues that the low monthly price of many AI tools is part of the reason they escape scrutiny.

“Everyone talks about shadow AI as a security problem, and it probably is one, but almost nobody's counting how much this actually costs businesses,” Nisbet said.

A $20-a-month tool feels invisible on a personal expense claim, and that's exactly why it never gets reviewed or cancelled.

That spending becomes more significant when several employees purchase similar tools independently.

Nisbet said Marketing Signals encountered the same issue internally after asking employees which tools they were paying for themselves.

The company subsequently removed two tools and replaced three others with cheaper alternatives. According to Nisbet, monthly spending across those five subscriptions fell from £1,875 to £713.

That represented a saving of approximately £1,100 per month, or £13,200 over a year.

Shadow AI Is Also a Governance Problem

The financial cost cannot be separated entirely from the wider risks surrounding shadow AI.

When employees adopt tools without approval, organisations may lose oversight of the information being uploaded, where that data is processed and how external providers use it.

The same fragmented purchasing can also make compliance more difficult. One department may approve a platform after security checks, while another employee independently subscribes to a similar service with different privacy terms.

For larger companies, the issue therefore sits across several teams at once: finance, procurement, cybersecurity, IT and data governance.

However, Nisbet argues that solving the problem does not necessarily require another piece of software.

Instead, companies can begin with regular conversations about which tools employees actually use.

“It's a five-minute conversation, repeated regularly, with the people actually doing the work,” he said.

Businesses May Need to Audit Their AI Stack More Often

As generative AI becomes embedded into everyday work, annual software reviews may no longer be enough.

AI products can be adopted within minutes and abandoned almost as quickly. Employees may test several tools before settling on one, while recurring subscriptions continue after experimentation ends.

Regular software audits can help companies identify unused licences, duplicate capabilities and subscriptions that no longer justify their cost.

Businesses may also need clearer policies for employees who want to experiment with emerging AI platforms. An outright ban can push usage further underground, while completely decentralised purchasing makes costs and risks harder to control.

A more practical approach could combine approved AI tool lists with simple processes for testing new products and reporting individual subscriptions.

AI Adoption Is Moving Faster Than Corporate Procurement

The wider issue is that employee adoption of AI is moving far faster than traditional procurement processes.

Workers can now access sophisticated writing, coding, design and automation tools without waiting for an enterprise software rollout.

That speed creates opportunities for experimentation, but it also changes how businesses need to think about technology spending.

Shadow AI is therefore becoming more than a cybersecurity concern. It is also a procurement, finance and management problem.

For companies investing heavily in AI, the next challenge may not simply be deciding which tools to buy. It may be discovering how many they are already paying for.

About the Research

Marketing Signals developed the modelling as part of its Subscription Creep Report, examining software waste and the growth of individually purchased AI subscriptions.

The company has also created a free calculator that allows businesses to estimate the potential cost of subscription creep within their own organisation.