Partner Ad Feature is produced by an advertiser with the specific intent to promote a product and is not produced by the City AM team.

Over the last few years, the pressure has been on for business leaders to drive rapid, widespread AI adoption in their organisations.

Sometimes that pressure has come from an awareness that being slow to adopt could quickly lead to being outcompeted. Sometimes it’s rooted in the idea that broad experimentation is the fastest way to uncover ways of using the technology that work well for the business. And sometimes it is simply wanting to demonstrate to the market and investors that AI is at play.

For all of these motivations, leaders have tended to reach for one metric above all others to achieve this aim: measuring pure usage as a performance target.

The financial pitfall of pure usage

That’s an understandable tactic. AI has emerged and evolved remarkably quickly, and in the face of significant uncertainty it makes sense to default to the simplest KPI. Today, though, the flaws in that approach are now becoming increasingly apparent.

Pleo recently surveyed 3,500 European finance leaders to explore how AI is being implemented and whether it is having the right impacts. We found that targeting usage metrics has, in one sense, worked strikingly well, with 98% of respondents saying that they use AI. The picture of actual benefits, however, is more ambiguous: 49% of UK respondents said that managing AI is more of a problem than adopting it, 47% that they have over-invested in multiple AI tools, and 61% that the ROI of some AI investments is simply unknown.

To see coin-toss success rates on important indicators of value is concerning for a technology that has entered effectively every medium or large business, and this issue will only become more acute as the financial reality of AI evolves.

In its graduation from an experimental phase to being a core business tool, AI is increasingly being costed in more robust ways by technology vendors. Internal Pleo data provides evidence to support predictions that software costs will grow significantly as the role of AI expands, with OpenAI and Anthropic now placing as the 3rd and 7th biggest recipients of British businesses’ technology spending. This year, we have seen those AI giants move towards tying the cost of AI directly to usage, posing a real financial risk to businesses that have taken a usage maximisation strategy.

AI finance solutions for AI financial risks

That risk is also, however, an opportunity for finance teams to help enable the benefits of AI on a fundamental level.

The costs of AI can balloon in two ways: through ratcheting subscription costs as vendors redefine their pricing, or through runaway employee usage of non-fixed pricing plans. In both cases, it is finance that can establish the visibility, insight, and control needed to mitigate the risk. Finance leaders will need to build ways of understanding which teams are using which tools, on which pricing models and to what extent, in real time so that they can inform their organisations about what adoption truly looks like and how it is evolving.

That means committing resources to the task, however, and Pleo’s survey of finance leaders found that the addition of AI is not yet freeing them up to make that possible.

Over half devote more time to policing spend than to shaping strategy, and likewise over half say that spend-control workflows place an excessive burden on their teams. As a result, they are spending days out of each week on fundamental, routine workflows at a moment when their businesses need them to innovate solutions to brand new challenges.

This is the more essential problem with measuring usage alone when it comes to AI adoption – not just for finance teams, but for businesses more generally. Expediting a broken process by adding AI on top might ease some pressure, but that’s not the kind of transformational shift that businesses want or need. If finance functions are going to meet the emerging challenge of AI, they first need to reassess their own approach to AI.

In practice, that means rethinking workflows from the ground up in this new context. 

Something as simple as an expenses claim, for example, typically triggers a series of manual steps to capture the employee’s request, match it against a receipt or invoice, apply spend policy and fraud detection to it, and finally approve it. A truly integrated, AI-powered finance solution could eliminate much of that workflow by fetching receipts directly from the employee’s inbox and automating the work of assessing it against company policy, looping in the finance team only when exceptional circumstances demand human intervention.

By pioneering the move from simply using AI to targeting it precisely at areas that cause friction, finance leaders can achieve two things. They will model to the wider business what effective implementation looks like, solving challenges that only AI can answer. And, perhaps just as importantly, they will free up vital time for the strategic finance thinking that businesses will rely on as AI transforms more and more of their day-to-day work.

That’s the kind of work that finance professionals are passionate about – and which will generate a far greater upshot than simply asking employees to use AI.