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). |
CPA ALBERTA began a three-part <a href="https://www.cpaalberta.ca/Professional-Development/Certificate-Programs/Introduction-to-Generative-AI-for-CPAs” rel=”nofollow noopener” target=”_blank”>Introduction to Generative AI for CPAs certificate program on August 19, giving accountants hands-on exposure to current tools and practical accounting use cases. The program emphasizes both automation and the profession’s role as trusted advisers.
That combination points to the next problem accounting firms need to solve. Learning to produce an AI-assisted draft faster matters. Being able to reconstruct why that draft says what it says matters more.
Consider the ordinary work now moving into AI-assisted workflows: drafting a client memo, summarizing a technical document, explaining a variance, organizing research, generating a spreadsheet formula, preparing a meeting brief, or turning notes into a first-pass analysis. A fluent answer can look finished long before the underlying reasoning
Canadian Accountant has already been examining AI governance, ethical use, and the accountant-in-the-loop opportunity. Firms can turn those principles into daily practice with a simple evidence ledger.
For one recurring workflow, run a 30-day test. Each time AI contributes to the work, record the task, theny correction or exception, and where the output goes next. If the output informs a client recommendation, <a href="https://bitcomme.com/chi-rho-financial-appoints-daniel-hoinacki-head-of-capital-formation/” title=”Chi-Rho Financial Appoints Daniel Hoinacki Head Of Capital Formation”>financial decision, filing, report, or other consequential action, record who owns the final judgment as well
The point is not paperwork for its own sake. The ledger answers questions firms otherwise discover too late. Whichy need correction? Where do reviewers spend the most time? Which exceptions recur? Which uses can be standardized, and which still depend heavily on professional context?
That evidence also produces a better measure of AI value. A tool that saves ten minutes on drafting but creates eight minutes of checking, correction, and downstream repair has produced a very different result from a tool that saves ten minutes and requires one minute of review. Both may look equally impressive in a demonstration.
The ledger gives managers something more useful than usage statistics. It shows where AI removes work, where it merely moves work to a reviewer, and where it introduces a new control burden. It can also reveal where a workflow needs a clearer template, betterman
Accounting firms already understand the value of traceability. AI adoption should use the same instinct. When a workflow becomes routine, someone should be able to explain what information went in, what the system produced, what a person verified, what changed, and who accepted responsibility for the result.
That discipline should come before scale. Start with one workflow and one month. If the evidence shows reliable gains with manageable review, expand it. If the same exceptions and repairs keep appearing, redesign the workflow before automating more of it.
AI can make accounting work faster. An evidence ledger helps firms determine when faster also means dependable.
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). Title image: Stock photo ID: 1050855182. Author photo: courtesy Gleb Tsipursky.
