AI software audits for Hungarian SMEs: lessons learned

European finance departments are working with fewer software tools while more and more of them are using artificial intelligence. This is shown by Spendesk’s international survey published in September. Among Hungarian small and medium-sized enterprises, AI is also widely present, but the cost-side benefits are less visible for now. We have gathered what lies behind the difference, and how reducing unnecessary subscriptions can free up budget.

Two-thirds already use AI in finance

The research titled Top CFO Tools Report 2026 by Spendesk and CFO Connect is based on the responses of 215 European finance leaders. According to the two organisations’ own report, 67 per cent of finance teams use artificial intelligence, which marks a rapid rise: based on Spendesk’s data, the same figure was still 31 per cent in 2024 and 56 per cent in 2025. The results were also reported by FF News in a summary.

The other striking result is the decline in the number of software tools. According to the Spendesk and CFO Connect report, more than 90 per cent of finance leaders use no more than six software tools for finance work, while 41 per cent work with just one to three tools. According to the report, among the AI models in use, Anthropic Claude is the most popular, with 41 per cent usage.

Why are they narrowing their software stack now?

At many companies, subscriptions have built up almost unnoticed in recent years: separate tools have been used for invoicing, expense management, approvals and reporting. This phenomenon is often called SaaS sprawl. A single subscription may seem like a small item, but together they represent spending that is hard to oversee, while data also becomes scattered across multiple systems.

According to an analysis by The SaaS CFO, AI’s share in IT budgets rose from 12.1 per cent to 14.2 per cent, and 45 per cent of chief information officers (CIOs) finance new AI tools by reallocating existing SaaS licence budgets. Based on the analysis, a significant share of spending on AI therefore comes not from new budget, but from trimming old subscriptions.

As for how much room there may be in this, the cost-optimisation company CloudZero gives an estimate in its blog: according to the company, eliminating unused or overlapping SaaS licences can save as much as 20 to 40 per cent of IT costs. This is a provider’s own estimate, and actual results may differ significantly from one company to another.

Where does AI pay off quickly, and where is it uncertain?

According to Gartner’s survey of 160 finance leaders, automation of data extraction and the processing of supplier and customer invoices (AP/AR) pays back on average within 9 to 10 months. These are well-defined, repetitive tasks where the labour time saved and the reduction in errors can be measured relatively easily.

Gartner’s statement also points out, however, that 57 per cent of finance leaders say the real return on individual AI projects is still unclear. The research firm therefore urges a more disciplined investment approach in finance AI development.

Another survey draws attention to the predictability of costs. According to data from SpendHound published on Business Wire, 57 per cent of finance leaders are not sure they are paying a fair price for AI tools, while 46 per cent exceeded their planned AI budget, primarily because of a surge in usage-based fees, according to the report. If a tool’s fee depends on the volume of use, the monthly bill is harder to plan for than a fixed subscription.

Hungarian SMEs: widespread use, few rules

According to a report by Store Insider, 73 per cent of Hungarian SMEs use some kind of AI tool, but only 25 per cent have regulated corporate frameworks in place. According to the publication, in the absence of a strategy, AI has not yet replaced software subscriptions at domestic small businesses.

A 2025 analysis by VOSZ Port paints a similar picture. According to it, 82 per cent of employees are open to AI, but the main barrier to system-level implementation is the lack of expertise, which was identified by 88 per cent.

The difference, then, is not in the level of interest. According to international surveys, at larger European finance teams AI forms part of software consolidation, while based on Hungarian reports, at smaller companies it is often just another tool alongside the existing ones. In such cases, spending tends to rise rather than fall.

Software audit step by step

The steps below provide a general starting point; the exact order and depth depend on the company’s size and activity.

• Full inventory: it is worth gathering every subscription in a spreadsheet, including the monthly fee, the responsible person, the renewal date and the number of actual users.
• Actual usage: many tools’ admin interfaces show who logs in and how often. Licences that have not been used for months are the first candidates for cancellation.
• Identifying overlaps: it is common for two or three programs to perform the same task, for example expense management or document storage.
• Designating a central platform: following the European trend, it is worth examining whether finance processes can be organised around a few central systems.
• A narrow, measurable AI entry point: based on Gartner’s data, invoice processing and data extraction are the areas where return can be measured the fastest.
• Cost cap: with usage-based pricing, it is advisable to set a monthly limit and an alert, if the provider allows this.
• Internal rules: it is worth defining what data may be uploaded into AI tools, and who may decide on a new subscription.

Before cancelling a subscription, it is worth reviewing the contractual terms and notice periods, and consulting the accountant on how vouchers and data stored in the old system must be retained.

Order first, then automation

The international data suggests that the companies profiting most from AI are those that first put their software estate in order, then direct the freed-up budget towards targeted, clearly measurable tasks. For Hungarian SMEs, the first step may therefore not be buying yet another tool, but reviewing the ones they already have. This article provides general information and does not constitute financial, legal or IT professional advice; before making a specific decision, it is worth consulting a specialist.