Capterra's 2026 Software Buying Trends Report surveyed 3,385 buyers across 11 countries and 35 industries. In the Finance sector, 62 percent ended up disappointed: they experienced disruption, regret, or both (Capterra / Gartner Digital Markets). That is 22 percentage points above the cross-industry average. Over a third of those disappointed buyers only found the right tool after they had already paid for a different one. This article pulls together 25 data points to explain what is going wrong and what the firms avoiding these outcomes do differently. For the broader adoption and productivity data, the 75-stat roundup covers the full picture.
How deep does the regret run?
The 62 percent figure is not a one-off finding. A separate Capterra survey found that 70 percent of tech buyers at financial organisations regretted one or more purchases in the prior year (Capterra 2024 Tech Survey). Gartner's own research found that 60 percent of renewal and expansion decision buyers regret nearly every purchase they make, a 6 percentage point increase from 2020 (Gartner, June 2023). Software buying regret is not a new problem and it is not improving.
What makes it expensive is the follow-on damage. 57 percent of buyers who experienced regret said the bad purchase had substantial or monumental financial repercussions (Capterra / Gartner Digital Markets). When Gartner investigated root causes, two dominated: 33 percent cited higher-than-expected cost of ownership and 32 percent cited slow or complex implementation (Gartner). Neither of those problems is visible during a sales demo.
What is overwhelming firms?
The regret problem does not happen in isolation. It happens inside firms that are already stretched thin.
66 percent of accountants feel overwhelmed by technology at least weekly (Intuit QuickBooks Accountant Technology Survey, April 2025). The average firm runs 8 different applications, and the consequences of that stack are measurable: 44 percent report high subscription costs, 41 percent struggle to get tools to work together, 41 percent say data entry across systems is time-consuming, and 33 percent find training burdensome (Intuit / Firm of the Future).
85 percent of firms believe that failing to adopt new technology will hold them back (Intuit QuickBooks 2025). But only 28 percent say their training programmes fully equip employees with the skills they need (Intuit QuickBooks 2025). That is where most regret starts: firms feel pressure to buy, but they cannot prepare their teams to evaluate or implement what they are buying. Pressure without preparation is a recipe for a bad purchase.
Why do AI implementations fail?
Start outside accounting for a moment. A RAND Corporation study of experienced data scientists found that 80.3 percent of enterprise AI projects fail to deliver their intended value: 33.8 percent are abandoned outright, 28.4 percent deliver no measurable value, and 18.1 percent cannot justify their cost of operation (RAND Corporation, 2024). That is not an accounting-specific figure. That is AI projects across every industry.
Finance does slightly better, but not by much. The CPA.com 2025 AI Report, drawing on Gartner data, found that only 36 percent of finance AI use cases proved successful (CPA.com / Gartner). A benchmark study that tested current AI models against 101 real accounting workflows found that the best model still fails one in five tasks, hitting 79.2 percent accuracy (CFO.com / DualEntry benchmark, 2025).
Hidden costs make this worse. 85 percent of organisations misestimate AI project costs by more than 10 percent, and 24 percent miss by more than 50 percent. Software licences typically represent only 30 to 50 percent of the total implementation cost. Data readiness alone commonly runs $5,000 to $15,000 in unplanned time and consulting (CPA Practice Advisor, May 2026). A firm that budgets only for the licence fee is budgeting for roughly half the project.
Do most firms even have an AI strategy?
Two independent surveys landed on nearly identical findings. Thomson Reuters found that only 22 percent of organisations have a visible, defined AI strategy (Thomson Reuters 2026 AI Report). Karbon, surveying its own user base, found only 21 percent of accounting firms have an AI policy or strategy (Karbon State of AI 2026). Different research teams, different samples, same result: roughly four in five firms are buying AI tools without a documented plan for how to choose, implement, or measure them.
The difference it makes is measurable. Firms with formal AI strategies are twice as likely to report AI-driven revenue growth and 3.5 times more likely to see critical AI benefits (Thomson Reuters 2026). And yet only 18 percent of organisations measure whether their AI investments are actually working (Thomson Reuters 2026, reported by Certinia).
Only 25 percent of accounting firms have established AI governance policies, according to Wolters Kluwer's survey of 2,768 respondents across 14 countries (Wolters Kluwer FRA 2025). Governance, meaning documented rules for where and how AI is used, who reviews outputs, and what data it can access, is different from strategy. A firm could have a strategy ("we will adopt AI for tax research") without governance ("here is how we review AI-generated tax guidance before sending it to a client"). Most firms have neither.
What else gets in the way?
Trust is declining even as usage grows. The KPMG global trust survey, the largest dataset on AI trust at more than 48,000 respondents across 47 countries, found that less than half of people (46 percent) are willing to trust AI, even though 66 percent use it. Trust has declined since 2022, not risen. (KPMG "Trust, Attitudes and Use of AI," 2025)
Inside organisations, the resistance is specific. Gallup found that 47 percent of managers at organisations actively implementing AI are either ambivalent about or actively discouraging AI use among their teams (Gallup, 2025). When the people making daily workflow decisions are sceptical, executive enthusiasm does not change behaviour on the ground.
On the technical side, 53 percent of executives say difficulties connecting AI to legacy systems have derailed initiatives (IBM Institute for Business Value, 2025). And Thomson Reuters found that data quality has been the single greatest barrier to AI adoption in tax and accounting for two consecutive years (Thomson Reuters, reported by CPA Practice Advisor). If the data going into an AI system is incomplete, inconsistently formatted, or spread across eight different applications, the outputs will reflect that regardless of how capable the model is.
What do the firms that get it right do differently?
The firms that get past all of the above do measurably better. Firms with a documented AI strategy see twice the AI-driven revenue growth of those without one. Firms actively using AI report 37 percent higher revenue per employee (Rightworks 2025 Accounting Firm Technology Survey, n=494). And 81 percent of accountants say AI has boosted their productivity (Intuit QuickBooks 2025).
What separates them is not luck or budget. They treated tool selection as a process: figured out what their firm actually needed, matched tools to those requirements, budgeted for the full cost of adoption rather than just the licence, and tracked whether the investment was paying off.
If you are running eight apps, feeling the pressure, and unsure which tool fits which problem, start with a structured evaluation before you spend anything. The five-question framework covers what to ask before you buy. The matchmaker quiz matches tools to your firm's specific requirements in about two minutes.
Twenty-five data points, and they point the same way: most firms end up disappointed because they bought without a plan and underestimated the real cost. The ones that avoid that treated tool selection as a process, not a purchase. For the full data behind these trends, the 75-stat roundup has 27 primary sources in one place.
