Nearly nine out of ten accountants using AI report positive returns. But another statistic is more troubling. Over half of accounting firms have experienced data breaches recently, yet fewer than half have guidelines for how AI handles sensitive financial data. The productivity gains are real, but so are the risks we’re ignoring.
Blake Oliver, host of the Earmark Podcast, recently sat down with David Jani, Senior Content Analyst at Capterra, to unpack Capterra’s 2026 Accounting Software Trends report. The survey of 500 U.S. accounting managers shows the profession has moved beyond testing AI and into territory where the gap between adoption speed and security readiness is becoming dangerous.
The Productivity Gains Are Real (With a Catch)
AI in accounting has crossed from experiment to standard practice. More than half of accountants now use AI in their accounting software, and it appears across all company sizes, not just enterprises with big tech budgets. As David noted, “We’ve gone beyond the point of it being companies testing the water with this stuff.”
The most common uses for AI are chatbots and AI assistants, followed by data entry automation and fraud detection. AI is also making headway in predictive analytics, cash flow forecasting, smart invoicing, and bank reconciliation. David described it as “a coalescence around analytics and process-driven tasks.”
The 89% positive ROI figure comes from two main benefits. Half of respondents cite productivity gains, and nearly as many report reduced errors. So firms see real time savings and quality improvements.
But 48% of accountants manually check every single AI output. Not spot-checking, but checking everything. And about a third catch errors in their AI outputs more than half the time.
How do you square 89% positive ROI with error rates that high? David’s practical take is AI is “creating some gains in some areas, creating some extra work in others,” but the net result stays positive. Even when you add review time, firms come out ahead. But he cautioned, “It’s important that businesses still keep a close eye on the ROI of these situations and confirm it is delivering those gains.”
Meanwhile, plenty of work remains manual. More than half of respondents still handle financial reporting through spreadsheets or manual processes. Accounts payable and receivable, billing, invoicing, and payroll are all heavily manual. And yes, 51% of accountants still use Excel or Google Sheets for financial data. As Blake observed, spreadsheets have survived 40 years and aren’t going anywhere soon.
The Security Gap No One’s Taking Seriously
While firms celebrate productivity wins, the security picture is alarming, and almost nobody seems concerned enough to act.
Consider 52% of accounting managers surveyed have experienced a data breach in the last two years. That’s more than half. While David doesn’t have data linking these directly to AI, what he found about AI and sensitive data should worry every firm leader.
“Most companies don’t have clear guidelines on how they use AI tools with sensitive data,” David revealed. Fewer than half (49%) have guidelines for employee and payroll information. Coverage of bank reconciliation and customer billing data is even lower.
The perception gap is striking. Nearly half view AI cybersecurity risk as “minor,” another 12% as “insignificant,” and only 3% as “critical.” This might be “why so many people don’t have guidelines. Unfortunately, they just don’t perceive the risks at play,” David said.
Blake painted a scenario that’s probably happening now. Someone uploads payroll reports into free ChatGPT, where the terms of service may allow the vendor to train on that data. “We really need to step up,” he said.
The risks go deeper. Blake raised the issue of prompt injection, which involves hidden text in documents that manipulates AI agents into leaking data or changing payment information. It’s sophisticated and hard to defend against. As David acknowledged, “It’s a very new and rather sophisticated way of extracting information from a company. We still don’t really know enough about it.”
David didn’t sugarcoat his advice. “Guidelines around this don’t seem like much, and obviously, everyone is rushing to get AI tools. But it’s a huge risk factor we need to address.”
AI Is Raising the Bar
If AI makes accountants more productive, you’d expect fewer jobs. But the data tells a different story, and it came as a surprise to David.
“Despite a lot of reports predicting the end of accountants, it’s not really what we found,” he said. Companies are adopting AI, but “it’s not necessarily affecting hiring decisions in the same way. A lot of companies are actually more focused on upskilling.”
Blake offered a historical perspective. The same panic hit when VisiCalc and Excel arrived 40 years ago, yet accounting jobs grew. When cloud computing transformed the industry, client accounting services didn’t shrink. Instead, it’s grown year over year for a decade.
The talent shortage persists, with 73% of firms reporting trouble with retention and hiring. The hardest roles to fill are mid-career positions. About a third struggle to find financial analysts, with specialized accountants (tax and cost accounting) close behind.
The paradox is AI actually increases the need for experienced professionals. Someone must review those AI outputs that are wrong half the time. Someone must understand the AI well enough to catch mistakes. Someone must manage the security implications. All that requires judgment and experience, and that’s exactly what’s hardest to hire right now.
The data backs this up. Upskilling existing staff is the dominant strategy at 40%, double the 21% using AI to fill staffing gaps. Traditional hiring sits at 31%, with graduate programs at 23%. The profession is betting on people, not automation, to solve its workforce problem.
Looking Ahead: Challenges and Choices
What keeps accountants up at night? Budgeting and forecasting in an uncertain economy tops the list, followed by figuring out how to use AI effectively. As David put it, firms are trying to understand AI “in a way that makes sense.”
David has specific advice for where firms should invest their AI dollars. Map investments to your particular needs rather than chasing trends. For general guidance, he pointed to data entry automation and predictive modeling tools, especially cash flow forecasting and analysis dashboards, as areas delivering the most value.
When asked to predict what might change by the 2027 survey, David hopes to see more firms with updated security guidelines. “I think as these tools become more mature, more people will update their guidelines, especially for handling sensitive data like payroll and cash flow,” he said.
A Gap Between Speed and Safety
The Capterra data shows the profession is getting AI both right and dangerously wrong. The 89% positive ROI is genuine. Firms are saving time and reducing errors, even after factoring in review burdens. But that headline obscures the fact that over half have experienced breaches, fewer than half have AI data guidelines, and most dismiss the cybersecurity risk as minor, even with threats like prompt injection that the profession barely understands.
AI isn’t solving the talent crisis either. It’s raising the bar for what accountants need to know, making experienced reviewers more critical while the mid-career talent shortage intensifies.
Firms must build guardrails, write guidelines, and invest in upskilling their people to successfully work alongside technology that’s powerful but imperfect.
Want to dig deeper into these findings? Listen to Blake’s full conversation with David on the Earmark Podcast, and earn free NASBA CPE while you’re at it.