Accounting Students Beat AI-Free Peers by 2x

An accounting professor at the College of Charleston did something most companies still haven't: he gave his students structured permission to use AI, then measured what happened. Roy Martin's students, who completed an EY-built AI prompting course before working a tax scenario with AI assistance, scored roughly twice as well on related exam questions as the cohort that took the class without AI the year before, according to a July 2026 report from Live 5 News. Ninety percent of his students rated the project close to a nine out of ten.

That's not a marginal bump. That's a professor accidentally running a controlled experiment on how AI literacy compounds learning, while most workplaces are still arguing about whether to block ChatGPT on the corporate network.

The reality check

Martin's framing is the useful part: AI does maybe 80% of the heavy lifting, but the remaining 20%, the judgment, the specificity, the "is this actually correct for this client" work, is entirely human, and entirely the point. His students weren't handed AI output to forward. They were required to submit their prompting strategy, the AI's output, and their own analysis of where it was wrong. That's a workflow, not a shortcut.

Marketing teams rarely build that workflow. Most AI adoption in growth organizations still looks like individual people quietly finding their own prompts, with no shared standard for review, no documented judgment layer, and no accountability for what ships. Martin said senior partners at major accounting firms are already telling universities they expect new hires to walk in AI-fluent. Marketing leadership should assume growth counterparts, clients, and competitors are having the identical conversation, minus the university's structure.

Where we land

This isn't a story about AI replacing anyone. Martin says explicitly he doesn't think it replaces the accounting profession, and student Ally Evans made the sharper point herself: AI is going to be wrong sometimes, "just like anyone coming out of college." The risk was never the tool. It's organizations deploying AI at speed with no equivalent to Martin's three-part structure, no one accountable for the last 20%, and no way to measure whether the output actually got better or just got faster.

If your team can't answer what your equivalent of Martin's grading rubric is, that's worth fixing before your next campaign cycle rather than after a client asks why the report reads like nobody checked it. Our growth strategy work starts exactly there, and if you want AI folded into your team the way this classroom folded it into a syllabus, that's what our AI marketing services are for.