Is there an AI for accountants?

The Problem: Accountants Drowning in Manual Work and an Overwhelming AI Market

Every month, the same story plays out in accounting firms: closings that stretch late into the evening, reconciliations repeated by hand, and deadlines that slip despite everyone's best effort. Meanwhile, an avalanche of new AI tools promises to fix it all — if only you could figure out which ones actually work.

The scale of the manual burden is real. Stanford and MIT researchers, surveying 277 accountants and analyzing task-level data from 79 firms, found that AI adopters spend 8.5% less time on routine back-office processing — which tells you just how much of the average accountant's week is consumed by tasks a machine could handle. Nearly half of surveyed accountants said AI helped them meet deadlines more reliably, confirming that deadline pressure is one of the profession's defining pains.

Yet the same research reveals deep anxiety about handing work over to machines. 62% of accountants worry about AI-generated errors, while 43% fear data-security risks. Those concerns aren't hypothetical — junior staff who accept AI outputs uncritically see smaller gains than seniors who review everything, which means a bad tool can quietly create more review work than it eliminates.

And the tool landscape keeps exploding. Adoption statistics vary widely depending on how you define "using AI" — 98% of accounting professionals report using AI tools per Intuit's survey, while Thomson Reuters finds just 21% of tax firms use GenAI — but every measure points the same direction: up. One CPA, Tricia Katebini of GRF CPAs & Advisors, captured the feeling in the Journal of Accountancy: "There are so many new tools... It's nice to see, but it's overwhelming."

The options now span every corner of the profession:

• Tax research assistants like Ask Blue J and Thomson Reuters CoCounsel
• Audit automation suites from Trullion, Deloitte, EY, and PwC
• Bookkeeping agents such as Dext AI Assist and Intuit Intelligence
• General-purpose tools — ChatGPT, Claude, Gemini, Power Automate — repurposed for accounting work

That's the trap: more tools don't automatically mean more value. As Wolters Kluwer observes, the firms seeing the most benefit aren't the ones adopting the most tools — they're the ones embedding the right AI into how work actually gets done. The question isn't whether AI for accountants exists. It's which AI genuinely helps, and which one just moves the bottleneck from data entry to double-checking a confident-sounding machine.

The Answer: Yes — AI for Accountants Is Near-Universal and Measurably Working

Yes — and by a wide margin. The question is no longer whether AI for accountants exists, but which tools a firm should adopt, because the profession has already answered it en masse.

The adoption numbers are striking. A 2024 Intuit QuickBooks Accountant Technology Survey found that 98% of accounting professionals report using AI tools. Broader economic data confirms the trend: Federal Reserve research shows professional services — the sector that includes accountants — is among the top AI-adopting industries in the U.S., with roughly 33% of firms using AI and 62% of professional-services workers reporting work-related generative AI use.

Does it actually work? The most credible evidence comes from Stanford and MIT researchers who surveyed 277 accountants and collected task-level data from 79 small and mid-sized firms. The results were measurable, not hypothetical:

• AI users finalize monthly statements 7.5 days faster than traditional methods
• They spend 8.5% less time on routine back-office processing
• Reporting granularity rises by 12%, giving firms richer detail at no extra cost

Nearly half of surveyed accountants said AI helped them meet deadlines more reliably, and almost two-thirds named automation of routine tasks as the biggest benefit.

The consistent finding across academic research, trade press, and vendor surveys is that AI augments rather than replaces accountants. As MIT Sloan's Chloe Xie puts it, the technology "is not here to replace the human being — it's here to augment the experts who are already in place." Stanford GSB's Jung Ho Choi agrees: "that final judgment call — that's still a human decision." Employment data backs this up — the Bureau of Labor Statistics projects accountant and auditor jobs to grow 5% from 2024 to 2034, faster than average.

That makes human-in-the-loop review the profession's operating standard, not a temporary safeguard. Real-world case studies show firms like Eventus reviewing every AI-drafted memo before it goes out, and CPA.com identifies human-in-the-loop verification as a core strategic theme — AI handles the first pass while accountants focus on judgment and exceptions. This matters because 62% of accountants cite AI-generated errors as a top concern, and the research found junior staff who accept AI outputs uncritically see smaller gains than seniors who treat AI as a supervised collaborator.

For teams extending AI beyond off-the-shelf tools, agent skills — instruction packages that give coding agents like Claude Code or Cursor domain-specific capabilities — are an emerging path for automating accounting workflows. SkillzBench is building a free directory that security-scans and benchmarks these skills in sandboxes before deployment, so firms can verify what actually works.

The AI Tools Accountants Are Actually Using, by Task

Walk into most accounting firms today and you'll find AI quietly embedded in the daily grind — tax research, audit evidence, memo drafting, even the mail. But the tools accountants actually use vary sharply by task, and the market itself is mid-shift.

Tax research and audit automation

Purpose-built research tools lead adoption. Ask Blue J answers federal and state tax questions with citations — one CPA quoted in the Journal of Accountancy calls its factual answers "95%" reliable — while Thomson Reuters CoCounsel serves as an agentic assistant across tax, audit, and accounting. On the audit side, Trullion automates evidence search, and Deloitte deploys GenAI for document review, with EY running a unified AI platform across 160,000+ engagements.

Bookkeeping agents and general-purpose tools

Bookkeeping is where agentic AI is landing hardest. Dext AI Assist behaves, per one analysis, "less like a chatbot and more like a junior bookkeeper embedded in your workflow," while Intuit Intelligence handles instant answers and routine tasks inside QuickBooks. Firms also build their own pipelines with ChatGPT, Claude, Gemini, and Power Automate.

The documented ROI is concrete:

• A custom Power Automate + LLM mail pipeline saves roughly 4 hours of staff time per week at no added cost
• Eventus's memo bot, built with ~60 hours of development, cut memo drafting from 4 hours to 30 minutes
• Invenio Advisors eliminated outsourced coding work costing up to $150/hour

The shift toward purpose-built agents

Here's the telling statistic: according to Thomson Reuters research, 52% of tax firms using GenAI still rely on open-source tools like ChatGPT, versus only 17% using industry-specific solutions. That balance is expected to flip as vendors ship embedded GenAI — moving the profession toward what Wolters Kluwer describes as "review-ready, touchless tax returns."

Every case study shares one guardrail: humans review the output. Eventus checks every AI-drafted memo, and researchers at Stanford and MIT consistently find AI augments judgment rather than replacing it.

For firms extending beyond off-the-shelf tools, agent skills — instruction packages for tools like Claude Code and Gemini CLI — offer another route to custom automation. SkillzBench is building a free, independent directory of these skills with security scanning and sandboxed benchmarking, so teams can verify what actually works before deploying it.

How to Implement AI in Your Accounting Workflow: A Practical Roadmap

The gap between firms dabbling in AI and firms profiting from it comes down to implementation. A Wolters Kluwer report puts it plainly: the most successful firms won't be the ones that adopted AI first, but the ones that integrated it most thoughtfully. Here's how to join them.

Start with high-ROI, low-risk workflows. Roughly 40% of firms concentrate their AI use on tax research and document summarization, per industry data, while the top GenAI use cases for smaller firms include tax research, document summarization, and transaction categorization. These tasks deliver measurable returns without touching sensitive judgment work. One CPA-built mail-automation pipeline saved about 4 hours of staff time per week at no additional cost, and an AI-assisted coding setup eliminated outsourced work costing up to $150 per hour.

• Tax research and document summarization — the entry point for ~40% of firms
• Transaction categorization and bookkeeping automation
• Routine reconciliations and data extraction

Keep humans in the loop — and train your juniors. Every documented case study in the profession emphasizes human review of AI outputs, and 62% of accountants cite AI-generated errors as a top concern. The Stanford/MIT research found a critical wrinkle: senior accountants treat AI as a collaborator with oversight, while junior staff accept AI outputs uncritically and see smaller gains. Make review skills an explicit part of early-career training, not an afterthought.

Redesign the process, don't bolt on the tool. Early success comes to firms that rebuild their entire workflows with AI in mind rather than tacking AI onto legacy processes. The firms seeing the most value are not adopting more tools — they are embedding the right AI into the core of how work gets done. That also means implementation is a leadership responsibility, not an IT project: leadership owns data governance, review guardrails, and staff training. If you're extending beyond off-the-shelf tools with custom agent workflows, verify what you install — SkillzBench is building a free, independent directory of agent skills with security scanning and sandboxed benchmarking, plus a free MCP tool to connect chosen skills to any agent.

Track your results as you go. Firms using AI finalize monthly statements 7.5 days faster and spend 8.5% less time on routine back-office processing, so baseline your current cycle times now. When a workflow stalls, review the guardrails before blaming the technology — thoughtful integration, not speed of adoption, is what separates the winners.

Going Further: Equipping AI Agents with Verified Accounting Skills

Off-the-shelf accounting AI gets you partway there — but the firms seeing the biggest gains are the ones that extend it. CPAs are already building custom pipelines, like a Power Automate + LLM setup that saves roughly 4 hours of staff time per week at no additional cost. The next frontier for that kind of customization is agent skills: SKILL.md instruction packages that teach coding agents like Claude Code, Codex, Cursor and Gemini CLI how to handle specific workflows — reconciliations, transaction categorization, document summarization, and more.

Here's the catch: agent skills are proliferating fast, and quality varies wildly. Published research found that roughly 26% of skills carry at least one vulnerability — a serious concern in a profession where 62% of accountants already worry about AI-generated errors, per Stanford/MIT research. Dropping an unvetted skill into a workflow that touches client financial data is a risk no firm should take blindly.

That verification gap is exactly what SkillzBench is being built to close. It's a free, independent directory of agent skills designed around a simple principle: before you deploy a skill, you should know what it does, whether it's safe, and whether it actually helps. The approach rests on a few pillars:

• Commit-pinned versioning — every skill version is pinned to a commit SHA, so you always know exactly what you're running.
• Security scanning and license checks — each skill is validated and scanned before it earns a place in the directory.
• Sandboxed benchmarking — skills are tested in a sandbox with the same model, with and without the skill, so you see the measured Lift in percentage points rather than marketing claims.
• Ongoing tracking — upstream changes are monitored, re-scanned and re-benchmarked, so a skill that was safe last month doesn't silently become a liability.

A free MCP tool then connects the skills you choose to any agent, with per-agent loadouts so an agent only sees what you've deliberately selected — a useful guardrail when you're running accounting automations alongside other work. Everything on the platform, including the directory, benchmarks and alerts, is free; scores are built from security, quality, license, maintenance and community signals, and money never touches them.

The timing matters. The market is shifting from generic chatbots toward applied agents that operate within defined rules and create auditable decision trails, as industry analysis notes — and 52% of tax firms using GenAI still rely on open-source tools rather than industry-specific ones. Skills are how those open tools become purpose-built for your practice.

SkillzBench is pre-launch, with the product actively being built and a live waitlist open now. Join the waitlist at skillzbench.com to get early access to the directory, the free MCP connector, and benchmark results as accounting-automation skills are verified. Skills, proven — before you deploy them.