CPA firms already know artificial intelligence matters. The harder problem is execution. FloQast’s 2026 research found that 85% of accounting teams call AI a strategic priority, while only 10% use it extensively. The same research found that just 17% of teams feel ready to put rising AI investment to work. That gap is why firms need an AI adoption playbook built around human behavior as seriously as they build around technology.
The technology itself is already moving into real accounting workflows. A 2026 CPA.com and Blue J survey of more than 1,000 tax professionals found that 60% now use AI-powered tax research at least weekly, up from 33% the prior year. That rapid accounting AI adoption shows what can happen when a tool solves a specific problem and professionals understand how it fits their work. The challenge is getting from pockets of successful use to consistent adoption across tax, audit, advisory, and firm operations. Recent middle-market research shows that successful AI pilots still run into scaling barriers such as data quality, integration, security, and workforce readiness.
Diagnose the resistance before prescribing more training
Firm leaders often assume uneven adoption means employees need another demonstration or prompt-writing session. Sometimes they do. More often, people already understand enough about the technology to have formed an emotional judgment about what it means for them.
One group fears economic displacement. If employees believe that using AI means helping the firm eliminate their jobs, enthusiasm from leadership can sound threatening rather than inspiring. RSM’s 2026 Middle Market AI Survey found that 85% of respondents said executive leadership is more enthusiastic about AI than employees are. That leadership-employee gap makes firmwide AI adoption a trust problem before it becomes a skills problem.
Where leadership can credibly do so, firms should make a job-neutral commitment around productivity gains. Explain that AI-created capacity will first support growth, client service, higher-value work, and redeployment rather than become an automatic headcount-cutting target. Then make the career logic explicit: professionals who learn to supervise AI, test its output, and use it responsibly become more valuable as workflows change.
A second group experiences AI as a threat to professional identity. Accountants have spent years developing expertise in research, analysis, writing, review, and judgment. When a tool suddenly performs part of that work in seconds, telling professionals that AI will make them more efficient can miss the source of resistance. They may hear that the work proving their competence matters less. That reaction deserves attention because accounting leaders are already describing AI as a way to shift routine effort toward higher-value work, including strategy, relationships, and client trust.
The better response is to involve them in workflow redesign. Ask which parts of a process feel repetitive and low-value, which parts require judgment, and where clients benefit most from human involvement. CPA Practice Advisor has emphasized the value of starting with focused use cases, clarifying oversight, and tying AI learning to real engagement scenarios. Those practical AI guardrails help professionals see a path where AI removes friction while their expertise moves closer to exceptions, recommendations, and client conversations.
Make responsible use easier than secret use
A third pattern appears when employees see useful AI applications but do not know what the firm actually permits. They may experiment quietly, hide how they produced a draft, or use unapproved tools because the approved process feels slower than the work itself. CPA Practice Advisor reported that one-third of lawyers, accountants, and compliance professionals were using AI their organizations had not approved, with the rate rising among those who thought their organization was moving too slowly. That shadow AI is a predictable response to ambiguity.
Firms need rules employees can apply during an ordinary Tuesday afternoon. Name the approved tools. Define which client and firm data can enter them. Specify when outputs require source checking or second-person review. Explain what work should never be delegated to AI without human judgment. Create an escalation path for uncertain cases. This kind of governance control gives teams a way to expand AI use while preserving visibility, accountability, and review.
The professional obligations already point in this direction. Circular 230 establishes standards of competence, diligence, and conduct for tax professionals practicing before the IRS. A workable responsible AI use policy should translate those duties into concrete AI behaviors, including verification, confidentiality, documentation, and supervision.
Psychological safety matters here because people need permission to surface uncertainty. Amy Edmondson’s foundational research linked psychological safety with learning behavior in work teams, exactly the kind of speaking up and error discussion firms need during AI experimentation. An employee should be able to say, “I used the approved AI tool, this result looks plausible, and I am not sure it is right” without fearing that responsible experimentation will be treated as incompetence. That response gives managers a chance to teach judgment before a questionable output reaches a client.
Use respected peers to change everyday behavior
Firmwide adoption will not spread through policy documents alone. People watch colleagues they trust. A skeptical audit manager is more likely to reconsider AI after seeing another respected audit manager use it well on a familiar workflow than after watching a generic vendor demonstration.
Choose peer champions for credibility, not enthusiasm. They should understand the work, know where AI fails, and be willing to show both useful applications and mistakes. Their role is to model habits inside actual tax, audit, advisory, and administrative workflows, then feed recurring problems back to leadership. The profession already has examples of firms pairing enterprise deployment with AI enablement at scale, including role-relevant training paths and explicit guardrails. CPA Practice Advisor has reported that skills shortages are already delaying AI and automation initiatives, which makes practical, role-based AI training a capacity issue as much as a learning issue.
Managers then reinforce the behavior. Ask in reviews where AI helped, where it failed, what the professional verified, and what should change next time. That turns AI use from a side experiment into part of how the firm learns.
Measure behavior before declaring victory
License counts, logins, and training attendance tell leaders whether people had access. They do not show whether work changed. Thomson Reuters’ 2026 AI in Professional Services Report found that only 18% of professionals said their organizations track AI return on investment, while another 40% did not know whether ROI was measured. Better measurement starts with the AI behavior firms actually want to see.
Track whether approved tools appear in the workflows selected for adoption, whether employees verify outputs as required, whether review corrections rise or fall, how cycle time changes, and where saved capacity goes. Add short pulse questions about confidence: Do employees know which tools they can use? Do they know what data is off-limits? Do they feel safe admitting an AI-assisted draft needs help?
The goal is to create a system where experimentation becomes normal, mistakes surface early, and useful practices spread. That emphasis on measurable implementation already shows up in accounting-industry recognition of firms that pair AI deployment with governance and replicable outcomes. CPA Practice Advisor recently argued that an AI-first culture depends on everyday habits and human review rather than simply buying tools. That focus on AI at work is the difference between isolated productivity gains and a firm that actually changes how work gets done.
CPA firms have strong incentives to move quickly, but speed without behavioral design creates a patchwork: enthusiasts sprint ahead, worried professionals hold back, and quiet users invent their own rules. A practical approach to AI adoption at work addresses each of those behaviors directly. Give people credible reasons to trust the change, preserve the parts of professional identity that matter, make guardrails usable, let respected peers demonstrate good practice, and measure whether behavior and business results change together.
That is how AI moves from scattered experiments to firmwide capability.
ABOUT THE AUTHOR:
Gleb Tsipursky, PhD, is a behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026). https://disasteravoidanceexperts.com/aibook
Photo credit: BoliviaInteligente/Unsplash
Sign in to get access to this free resource, and all of our whitepapers and reports.
Download this content today!
Register Now Already registered? Click here to Log In
