Closing the AI Readiness Gap in Audit and Accounting
Many firms have already purchased AI solutions. But not all are seeing the return they expected when they signed the contract.
The problem is usually the gap between adopting a tool and preparing the people who have to use it. Licenses get assigned. Training sessions get scheduled and completed. Then six months later, adoption has stalled, staff have drifted back to their old workflows, and nobody can prove the investment worked.
That gap has a name: the AI readiness gap. It doesn't affect every role the same way. A first-year auditor learning to write a decent prompt is dealing with a different problem than a manager reviewing AI-assisted work, who is dealing with a different problem than a partner trying to govern the shift across the firm.
Training on the tool is not the same as training for the tool
A lot of firms treat AI readiness as a scheduling problem: book the vendor demo, run the workshop, check the box on a slide for the partner meeting.
A global IDC study of more than 1,000 audit leaders, called The Future of Audit and Accounting in the AI Era, found that only 28% of them describe their firms as very or extremely prepared to reskill staff for effective AI use. That statistic should worry anyone who signed off on an AI budget expecting adoption to follow naturally.
A one-day workshop can teach someone which buttons to click. But it doesn't teach them how to evaluate an AI system's output, or how to push back when something looks wrong. That second skill is the one that actually protects audit quality.
Three things seem to be driving the gap. AI is evolving faster than most firms can absorb, plan for, or budget around. Change management is underfunded relative to the tooling itself, so people are handed new systems with no real support for the disruption that follows. And speed of adoption keeps getting confused with quality of adoption. The missing piece is investment in the people expected to use the technology well.
The problem looks different at every level of the firm
AI literacy isn't evenly distributed across a firm, and neither is the way each role is meant to use it. A policy written for everyone usually ends up serving no one, and it's often the people you most want to keep who feel that friction first.
For junior staff, the issue is judgment. AI now handles a lot of the manual work that used to build that judgment through repetition: pulling data, reading contracts, flagging obvious anomalies. A junior auditor who never does that work doesn't automatically develop an eye for what looks off. They need to be taught not just what a procedure does, but why it exists, so they can supervise an AI agent doing the work instead of just watching it run. If you skip that step, then you risk your strongest junior people either checking out or leaving for a firm that's investing in them.
For senior staff, the review process has to change when AI is involved. Reviewing a person's work and reviewing an AI's output ask different things of the reviewer. The useful questions shift: What was the tool actually asked to do? How did it get to this conclusion? Is the reasoning behind that conclusion something a reviewer can defend to a regulator?
For partners and firm leadership, the job is governance, and governance is a skill that has to be practiced as the tools change, not a policy drafted once. It usually has to work at two levels at once: firm-wide standards that set the outer boundary of acceptable use, and practice-specific rules that account for how differently AI gets used in, say, tax versus assurance. A governance policy broad enough to cover an entire firm is rarely specific enough to catch what actually goes wrong inside a single practice.
Measure what happens at the desk, not who showed up to training
Once a firm decides to actually measure readiness instead of assuming it, the picture usually changes.
The AICPA and CIMA's Future-Ready Finance survey found that 61% of respondents rank on-the-job training as the most effective way to build new skills, ahead of formal courses or one-off sessions. That tracks with what most audit leaders already suspect: a class people sit through once doesn't stick the way daily, supervised practice does.
Attendance confirms a training session happened. It says nothing about whether staff can now evaluate an AI-drafted conclusion, or whether they're quietly still doing the task the old way because it feels safer.
A more useful set of measures looks at the actual work: how AI-supervised output compares in quality to work done without it, how confident staff say they are in catching a wrong AI conclusion, and whether AI use across the firm is staying inside the guardrails the firm set for it.
Readiness is a retention question before it's a technology question
The firms that pull ahead over the next few years will likely be the ones that built the judgment to use AI well at every level. The technology line item is the easy part of the budget to approve. The harder line item is the one that teaches a junior auditor why a procedure exists, teaches a senior manager to interrogate a machine's reasoning instead of a person's, and gives a partner a governance structure specific enough to actually catch a problem before it becomes a finding.
Firms that skip that part are behind on developing the people who were supposed to use the technology well.
Comments