If you’re reading this, there’s a non-zero chance you’re spending too much on software. You probably suspected as much, but a recent Spendesk report confirms it. European small and medium-sized enterprises are overspending on software and AI by an impressive €10bn per year.
€7bn of this stems from vendor-level price variance, meaning that companies of roughly the same size are paying different prices for the same products or services. This means that every company in Europe overspending because of variance is coughing up an extra €40,000 per year. But the extravagance doesn’t stop there. Europe’s businesses are also overspending €3bn on duplicate software, including functionally duplicate AI tools.
The response of business experts and consultants to this unfortunate situation is to call for as much centralisation over software usage and contracts as possible. Enterprises should also consider whether they can negotiate for discounts or more favourable volume pricing, including usage-based pricing. By taking such steps, and by ensuring greater oversight of spending, they might be able to spend all that extra money on something more productive.
Fragmented approaches to software procurement
Price variance has a number of sources. Spendesk’s European Spend Report 2026 indicates that perhaps the biggest is negotiation, or rather the lack of it. If a company goes too long between reviewing contracts, and if there’s a divergence between how many licenses it’s secured and how many it actually uses, then it can pay more than a more ‘disciplined’ peer.
And if no one in the finance department gets to look at a contract and question it? Overspending could continue for years.
As its name suggests, Spendesk is a budget management platform, with access to what companies are actually spending on software. For the report, it tracked 2,500 European businesses with headcounts of between 50 and 250. These 2,500 firms overspent to the tune of €69m. When extrapolated to all 250,670 similarly sized companies in the EU (according to Eurostat Structural Business Statistics), this scales to around €7bn.
Yet Faram Medhora, a Principal Analyst at Forrester, points out that this €7bn figure needs “a narrower label” than total European software overspend. “It estimates the annual gap in per-seat SaaS prices paid by European companies with 50 to 250 employees,” he notes. “It does not measure total software overspend across all European enterprises.”
This point aside, Medhora agrees that software rarely has a single visible market price. He affirms that what a company ends up paying is usually a reflection of its negotiating strength, the package chosen, user count, contract term, and timing. As such, some variation is “reasonable,” yet things can become unreasonable when buyers are unable to see what they own, what their employers actually use, or what other similar customers pay.
“Software providers negotiate with a complete view of the account,” he adds. “Buyers often negotiate with contracts in one system, usage in another, and renewal dates in someone’s spreadsheet.”
Medhora uses the example of a company renewing 1,000 licenses for a product when only 650 employees actually use it. Even if procurement drives a hard bargain and secures a significant discount, it may still ultimately sign a bad deal. A similar judgement applies in cases where departments buy software separately, or where a contract renews without renegotiation. To prevent such possibilities, Medhora advocates for greater centralisation of procurement.
“Companies should bring contracts, invoices, usage, and renewal dates into one view, then start major renewals early, compare equivalent deals, and remove unused licenses before negotiating,” he says. “SaaS and software asset management tools can connect contractual entitlements with cost and actual use.”
Duplication more than doubles in a year
The report also highlighted the top five software vendors in terms of price variance and opportunities for savings. The 2,500 companies surveyed are spending €11.7m more than needed on Hubspot, putting it in first place, followed by Atlassian, with a total overspend of €11.3m. Rounding out the top five were Salesforce (€8.3m), Adobe (€4.4m) and Notion (€3.3m). While Notion came last out of the five, Spendesk reports that it has the widest variance per head, with the “most disciplined buyers” paying only €0.6 per user and the median company paying €5.5.
Overspending is also getting worse. Last year, the 2,500 surveyed companies outlayed €64.9m more than needed, compared to €69m this year. Spendesk’s data also suggests that variance is worse in Germany and France than it is in the UK, which has “the narrowest spread and fewest extreme overpayers.” By contrast, France and Germany boast lower medians but suffer from a longer tail of enterprises paying at the extremes, while Spendesk estimates that France alone spends €1.6bn too much on software each year.
And while Europe’s SMEs as a whole are overspending €7bn because of price variance and related issues, they’re also spending an additional €3bn as a result of tool duplication. Much of this comes from AI, with 72% of AI-adopting enterprises paying for at least two competing LLMs. Yet the problem is also evident in other areas, with 64% of design tool-using companies guilty of duplication and 44% of firms using dev tools.
“In many organisations, fragmented teams and organisational structure cause duplicative buying,” says Stephen Elliot, Group VP of Software Development and IT Operations at IDC.
What stands out in Spendesk’s data is that duplication of AI services has grown substantially in the past year. In 2025, only 29% of AI-adopting SMEs used two or more similar tools, but now almost three quarters of European SMEs are in this bracket. However, Spendesk concedes in its report that the increase “tracks with models specialising by discipline,” implying that some of the reported duplication may not actually be duplication in the strictest sense.
Learning where value resides
There is an expectation that, even if no single AI provider will dominate, duplication will shrink over time. “Enterprises will cut obvious overlap,” Medhora says. “If four assistants can summarize the same meeting, search the same files, and draft the same email, finance will eventually ask why it is paying four bills.”
On the other hand, Medhora also observes that many large enterprises will continue to use several AI providers, given that their tools perform different jobs. “A company might use Microsoft 365 Copilot as an employee assistant, a Salesforce agent to update an opportunity, ServiceNow AI to resolve a support request, and SAP AI to investigate an invoice,” he says
Forrester expects AI usage to eventually create an “agentic business fabric,” whereby agents will coordinate work across multiple applications rather than remain boxed inside any single interface. The process of learning where value resides in such an approach could be “messy” and might take several years, but Medhora expects that existing “applications, built-in AI, standalone assistants, and custom agents will run side by side while companies learn which ones earn their keep.”
Stephen Elliot also argues that there will be no single provider for AI. On the contrary, he suggests AI is going to be “everywhere soon,” including in every piece of ‘non-AI’ software an enterprise uses. This could make avoiding duplication and redundancy difficult, but he advises tech leaders to focus on outcomes and value, and to define the metrics that matter.
“Compare how the human-based operational models work today and map the changes in an agentic-led model,” he says. “Finally, recognize this is a marathon not a sprint.”
If you’re reading this, there’s a non-zero chance you’re spending too much on software. You probably suspected as much, but a recent Spendesk report confirms it. European small and medium-sized enterprises are overspending on software and AI by an impressive €10bn per year.
€7bn of this stems from vendor-level price variance, meaning that companies of roughly the same size are paying different prices for the same products or services. This means that every company in Europe overspending because of variance is coughing up an extra €40,000 per year. But the extravagance doesn’t stop there. Europe’s businesses are also overspending €3bn on duplicate software, including functionally duplicate AI tools.
The response of business experts and consultants to this unfortunate situation is to call for as much centralisation over software usage and contracts as possible. Enterprises should also consider whether they can negotiate for discounts or more favourable volume pricing, including usage-based pricing. By taking such steps, and by ensuring greater oversight of spending, they might be able to spend all that extra money on something more productive.
