Accounting and tax firms: technology and AI trends report

The biggest accounting firm technology trends are the spread of AI tools into daily work, a shift of staff time from data entry toward client work, and a widening gap between firms that train and govern their AI use and those that do not. Budgets reflect it: Thomson Reuters' 2025 and 2026 State of Tax Professionals reports both found AI was the only technology spending category growing year over year.

This report compiles Karbon's State of AI in Accounting reports for 2025 and 2026, Thomson Reuters' 2025 Generative AI in Professional Services report (as covered by CPA Practice Advisor), the Thomson Reuters Institute's 2025 and 2026 State of Tax Professionals reports, and a Stanford and MIT study reported by the Journal of Accountancy.

Nearly all accounting professionals now use AI in some form, though firm-level adoption of generative AI in core work is lower. Karbon surveyed 593 accounting professionals across six continents in October 2025, most in the United States. It found 98% use AI, up 15 points from the prior year. 74% use it daily and 55% several times a day, while 7% use it rarely. Karbon notes its respondents come from its own network and may be more progressive than the industry as a whole.

Thomson Reuters measured firm implementation, a stricter test. Its 2025 Generative AI in Professional Services report, covered by CPA Practice Advisor, found the share of tax firms implementing generative AI rose from 8% in 2024 to 21% in 2025, with another 25% planning to. 79% expect significant integration by 2027. Among firms using it, 44% use it daily or multiple times a day.

Accounting firms mostly use AI for communication, research and documentation, with tax-specific uses growing. Karbon's 2026 report and the Thomson Reuters 2025 professional services report show these patterns:

Self-reported savings average about an hour a day, and an independent study found a smaller but measurable shift. Karbon's 2026 survey found accounting professionals save 60 minutes a day on average with AI, about 21 hours a month. Savings rose with skill: beginners reported 48 minutes a day, intermediate users 61 and advanced users 82.

The Journal of Accountancy reported in August 2025 on a study by accounting professors at Stanford and MIT. They surveyed 277 accountants at 79 small and midsize firms over three consecutive workweeks and analyzed transaction data from an AI accounting software provider they partnered with. Generative AI users shifted about 8.5% of their time from routine data entry to work such as client communication and quality assurance, roughly 3.5 hours in a 40-hour week. They also supported 55% more clients per week than non-users and closed the books 7.5 days sooner.

The two figures measure different things: Karbon's is total self-reported time saved, the Stanford and MIT figure is time measurably moved between task types. Both point the same way.

Yes. Training is the clearest dividing line in the data. Karbon's 2026 report found people at firms that offer AI training save 28% more time than those at firms that do not. Advanced users were concentrated in firms that train: 67% of advanced users work at firms offering or developing training. Yet only 46% of firms invest in AI training, up 9 points from the prior year.

Karbon's 2025 report found the same gap a year earlier: 85% of respondents were optimistic about AI, while 37% were actively investing in training. The most common training methods in 2026 were peer learning (26%), internal workshops (22%) and self-guided learning (19%).

Yes. Karbon's 2026 report found clear differences by firm size. Firms with 21 to 50 staff have been the most excited about AI for three years running (60%) and the least fearful or skeptical (8%), even after a 7-point dip in excitement. Firms with 4 to 10 staff were the only size group where excitement rose in the past year, up 8 points.

The smallest and the in-between firms showed more strain. At firms of 1 to 3 people, skepticism rose 6 points and excitement fell 5, which Karbon attributes to the owner acting as implementer, trainer and user at once. Firms with 11 to 20 staff saw skepticism rise 11 points; Karbon describes them as too large for informal adoption but too small for dedicated roles to standardize and train. For firms in that range, a written process and a named owner for each tool tend to matter more than the choice of tool.

Data security is the top concern, and policies lag behind use. Karbon's 2026 report found 83% of respondents are concerned about data security, up 7 points and 19 points higher than three years earlier. Thomson Reuters' 2025 professional services research found 70% of firms had no formal policies governing generative AI.

Staffing is the other tension. Karbon found 76% of firm owners are excited about using AI to forgo hiring in the next one to two years, while concern about job displacement in the field rose to 22%. Thomson Reuters' 2026 tax report describes a weak new-talent pipeline and senior retirements, with midsize firms feeling the squeeze most.

Many firms are moving capacity into advisory and new services. Karbon's 2026 survey found 70% of firms using AI report improved efficiency and productivity, 16% are offering new services and 42% are considering it. Thomson Reuters' 2026 report says clients increasingly want advisory help beyond tax preparation, and its 2025 report found firms expanding into tax strategy and business consulting. The same 2025 report found average revenue rose 21.3% in 2024 and profit 25%, and 49% of respondents said technology figures heavily in their profitability plans.

The data suggests the advantage now comes from how a firm adopts technology, not whether it does. Steps that follow from the research:

• Write an AI policy before expanding use. Cover which tools are approved, what client data may be entered and how output gets reviewed. With 70% of firms lacking one (Thomson Reuters), this is a gap most firms can close quickly.
• Train deliberately. Karbon's data ties training to 28% more time saved. Short peer sessions on real firm tasks are the most common method and cost little.
• Start with communication and documentation. These are the most common uses and the lowest risk: client email drafts, meeting notes and SOPs.
• Automate the workflow around the work. Client document requests, reminders, engagement letters, e-signatures and status updates are rule-based and repeat every season.
• Measure where hours go. Track time on data entry, chasing documents and client communication before and after a change, so savings are visible.
• Decide where saved capacity goes: more clients, advisory services or less overtime during busy season.

Start by mapping one recurring process, such as onboarding a new tax client or collecting year-end documents, from first contact to completion. Mark each manual handoff and reminder, then automate the ones that follow a fixed rule and keep review steps with staff.

Monvell's Automation and Operations teams do this kind of process mapping and build the automated reminders, intake and document collection around a firm's existing practice management software.