One of the more interesting questions surrounding AI in accounting isn't simply whether it saves time.
It's whose time it saves — and where the work moves next.
Tools such as ChatGPT and Claude can help experienced professionals draft review comments, summarize information, organize questions and work through complex material more quickly. For a partner or reviewer, that can create meaningful efficiency. But consider what happens downstream. If an AI-assisted review generates more detailed questions, identifies additional inconsistencies or produces a longer list of potential issues, the preparer may now have more items to investigate, verify and resolve. The partner may have saved twenty minutes. Did the preparer just gain forty? That doesn't mean the technology failed. It means the firm needs to evaluate AI at the workflow level, rather than measuring productivity one person at a time.
A useful question becomes:
Did AI reduce the total amount of work required to move the return from preparation to final approval — or did it simply move the work somewhere else?
There is another consideration. AI can assist with identifying issues and organizing information, but professional judgment and review remain human responsibilities. That means firms need to determine not only where AI can accelerate work, but where verification belongs in the process. As firms experiment with AI during preparation and review, I think this will become increasingly important. Instead of asking only: “How much time did AI save me?” It may be worth asking “What happened to the entire workflow after I used it?” That's a much more useful measure of efficiency.