AI Assistants in Practice: How CPAs Are Actually Using Them Day-to-Day

By Barry S. Graham, CPA, CMA

In a recent article of mine, I focused on moving “beyond the buzz” of artificial intelligence. Since then, the most common follow-up questions I’ve received are simple: What tools are you actually using, and how are you using them in your day-to-day work?

For most CPAs, the answer starts with general AI assistants like ChatGPT, Claude or Microsoft Copilot. While these platforms are often discussed in broad, futuristic terms, their real value today is much more practical. They are most effective when used to support the core activities we already perform: drafting, summarizing and thinking through engagements.

One of the most immediate benefits is in drafting communications. Whether it’s emails to clients, internal follow-ups or status updates, AI assistants can help organize thoughts quickly and professionally. For example, I often provide a few bullet points — key facts, tone and objective — and use AI to generate a first draft. This is particularly helpful during busy season when responsiveness matters. The output is rarely final, but it consistently reduces the time it takes to get to a polished message.

Similarly, AI tools are highly effective for drafting memos and disclosures. In a review or audit context, we are frequently documenting conclusions around areas like revenue recognition, related party transactions or unusual balances. AI can help structure these memos by organizing the background, issue, analysis and conclusion in a logical format. It can also help refine language to ensure clarity and consistency, especially when adapting prior-year documentation to current-year facts.

That said, the key is to use AI as a starting point — not a substitute for technical judgment. The accountant still needs to ensure the content aligns with GAAP, firm methodology and the specific facts of the engagement.

Another area where I’ve found AI assistants useful is in reviewing workpapers and understanding engagement context. For example, when onboarding onto an engagement or revisiting an area after some time, I may input summaries of key documents — such as engagement letters, prior-year memos or risk assessments — and ask the AI to highlight key terms, obligations or risk areas. This can help quickly orient the reviewer to what matters most.

In risk assessment, AI can also be used to pressure-test thinking. For instance, after identifying significant risks or areas of judgment, I may ask the tool to suggest additional risks or considerations based on the facts provided. This doesn’t replace professional skepticism, but it can surface angles that might otherwise be overlooked, particularly in complex or unfamiliar industries.

There are also practical applications in analyzing fluctuations or unusual items. By summarizing account activity or explaining variances, AI can help frame initial inquiries or draft questions for management. This is especially useful in the early stages of fieldwork when building an understanding of the business.

Of course, there are important limitations. Data security is critical — sensitive client information should not be entered into public tools without proper safeguards. In addition, AI outputs can sound confident even when incorrect, so all content must be reviewed and validated. These tools are most effective when paired with strong professional judgment, not used in isolation.

For firms looking to adopt AI, the best approach is to start small. Focus on one or two workflows — such as drafting emails or summarizing documents — and build from there. The goal is not to overhaul your process overnight, but to incrementally improve efficiency and consistency.

AI will not replace CPAs. But CPAs who learn to use these tools effectively will be better positioned to manage workloads, respond to clients and focus their time on higher-value analysis. In that sense, the advantage is not theoretical — it’s already showing up in day-to-day practice.