Guest Blog: Why AI and SaaS spending is spiralling, and how CFOs can regain control

I recently worked with a SaaS company running Claude and Gemini in parallel. Claude had strict usage and spend controls in place. Gemini on the other hand, with a few thousand euros spent per month, had none.

So, when I read Spendesk's 2026 European Business Spend Report and noticed it suggests this is extremely common, I thought “finally, some data around this.” 

Turns out, 72% of companies using AI tools pay for two or more competing products, usually without any specific spending guardrails. The report is built on twelve months of actual transactions from 2,500 European companies rather than survey responses, which is why I take its conclusions seriously.

Key takeaways

• AI spending is rising faster than financial controls. Spendesk's research found AI spend grew 340% in 12 months, while monthly costs fluctuated significantly.

• Duplicate subscriptions are common. 72% of companies using AI tools pay for two or more competing products.

• SaaS overspending isn't just a procurement issue. Finance teams need leadership accountability, better cost visibility, and regular software reviews.

• Traditional fixed budgets don't work for usage-based AI. Driver-based forecasting, spending limits, and real-time alerts can help manage unpredictable costs.

• Visibility matters more than eliminating every duplicate tool. Different products can serve valid purposes, provided their costs and ownership are understood.

Why AI spending is becoming harder to forecast

The volatility in AI spending came as no surprise, either. Total AI spend grew 340% in twelve months, and the typical company's monthly AI bill moves by 61% from one month to the next. I see the same pattern with many of the companies I work with: a usage-based cost still budgeted as a flat monthly amount or flat amount per user, with the variance only surfacing at month-end close. It is also why I have been writing a LinkedIn series on keeping AI costs within budget.

What SaaS overspending reveals about financial governance

The first was how little scrutiny software receives compared to headcount. An unplanned €40,000 net new hire would require a business case and CFO approval. The report puts the average overpayment on software pricing at about €40,000 a year per overpaying company, and in most businesses that spend renews without anyone on the executive team reviewing it.

In more than ten scale-ups, I have never seen software spend or tool utilisation oversight on a departmental scorecard. It becomes a priority only when revenue falls short of target and FP&A is asked to find savings. By that point, employees are attached to their tools, and I often find one team paying for two products that do the same job because an individual prefers their own. With licence utilisation at <50% on external benchmarks, the report's estimate of €3 billion a year in recoverable spend from duplicate tools across Europe is easy to believe.

The second was that overpayment does not correct itself over time. Companies of the same size pay three to six times more per head for the same software, with HubSpot ranging from €6 to €24 per employee per month and Atlassian from €2 to €12.

Some of that gap is normal and expected. Larger companies within a size band qualify for volume tiers and bundle discounts, and annual prepayment typically earns a better rate. Regulated businesses also pay for enterprise security features. I would treat each gap as a question for the budget owner before calling it waste. What concerns me more is that the total overpayment stayed flat year on year while median SaaS rates fell. The market became cheaper, and the companies already overpaying did not benefit from these rate reductions.

€7 billion in SaaS overspend. Is your business paying more than it should?

Read the Report

How CFOs can control AI and SaaS spending

Software spend needs to be an organisational goal set by the C-suite, and reviewed regularly by the leaders of the organization, whatever the revenue picture. FP&A is well placed to identify duplicate and underused tools, but accountability belongs with the leadership team and the department heads who own their budgets. That means putting software in the operating plan next to headcount, and giving each department head a software budget and a utilisation KPI that sits on their scorecard.

The upfront contract deserves the hardest negotiation. Although most scale-up companies don’t have dedicated procurement teams, having software spend oversight as an active KPI for the company could encourage teams to negotiate more strategically when adding new software to their tech stacks. Renewal terms often start with the current year price so any lack in financial discipline in the first negotiation will result in compounding software costs over time.

For AI, I would replace the flat monthly budget with a driver-based forecast built on active users, usage per user and unit price, with a ceiling set by the budget owner. Spending limits and real-time alerts at the point of spend then flag an unusual month while it is happening. Those controls need to cover every AI tool in use. For the company I mentioned, controls on Claude covered only part of the exposure while Gemini ran uncapped and caused expense variability.

Spendesk's CFO, Pauline Bellee Babel, sets the standard well in the report: "not zero duplication, but zero duplication without visibility." I agree, and I would take it one step further. That visibility has to reach the leadership team, because tool spend gets managed when the people setting company goals are the ones reviewing it.