Direct an AI tool to a folder of client files and get back a referenced, first-draft workbook a few minutes later. Days of keying figures and chasing them across a spreadsheet avoided. For anyone who has spent a career doing audit work manually, this experience is both frightening and exhilarating.
The exciting part is what the auditor now gets to spend their time on. With the workbook building itself, the week's focus shifts to what the job was always meant to be about: understanding the client's business, weighing the evidence and forming an opinion. The part of the job that you had to work so hard to reach can now become the whole job.
The grind was always the toll
The lived experience of an audit is the grind: tie out the trial balance, agree it to underlying schedules, select a sample, obtain evidence and cross-reference. Build the workbook, have a client change a figure, build it again. When the hours go into building this scaffolding for the judgement call, the grind can easily look like the work itself.
The grind was the toll you paid to reach that judgement, and until now there was no way around it. You couldn't form a view on revenue without keying in every figure behind it, so the keying was always needed for the judgement. Once the workbook builds itself, the judgement is what remains.
The bottleneck just moves
Here's an old idea from factory floors: every process has a single bottleneck, and the moment you remove it, a new constraint appears somewhere else. Speed up the slow machine and there's a new slow step that makes itself known. Improving a process is a matter of chasing each new constraint.
Audit's constraint has always been execution. There was simply more support to work through than there were hours and people to do it. That is why audits tend to stall in the middle, with sound planning and a sound conclusion on either side.
Remove that constraint and the bottleneck moves. When a first-draft workbook arrives in minutes, the slow step becomes the review: deciding which flagged exception actually matters and which is noise. The bottleneck moves upstream, from the hands to the head — consider the importance of understanding the client's business well enough to see what should have been flagged and was missed.
The skill that gets more valuable
The skill that survives this shift is the one auditors already have and an LLM is worst at: knowing what looks wrong, when to investigate.
A good auditor reads a set of accounts, understands it and spots the one number that doesn't seem right. The number that doesn't sit well with the story of the business. They notice the margin that moved for no reason, the accrual that's suspiciously round, or the cut-off that's a little too convenient. This instinct is not mere pattern recognition, but borne of a deep understanding of the business. It comes from a wealth of experience, from conversations with the client, from knowing the levers that drive the company and incentives at play. Professional scepticism is a trained sense for what's wrong.
AI can compare this year to last year, tie a figure to its contract, surface the three places the numbers disagree. But it's up to the auditor to understand why they disagree, the significance thereof and to decide whether it's worth pursuing. Out of the countless ways a misstatement can hide, detecting and choosing the one worth chasing is the judgement.
Learning is what to protect
Now, what about trainees learning how to audit? If juniors never build the workbook by hand, how do they learn?
Internship works by providing exposure to how a real business works, and building workbooks is simply how that exposure has always been delivered. Until now, this manual work was the only way in which juniors could simultaneously add value and be within arm's reach of the business. But the workbooks were never the point; contact with the client was.
The learning is the thing to protect. If we were to strip out the manual work and do nothing else, we risk raising a generation that can run a tool, but is blind to the business. Where would this leave audit? The answer is to bring juniors to the judgement in their first week, where years of keying used to delay their exposure to what review looks like. We show them: here are the three flagged exceptions, here is the client, what do you think and why? A good week leaves the auditor sharper. The understanding and judgement is what a firm is really building in a trainee.
The billable hour stops measuring the work
This is the habit audit will find hardest to give up.
No one ever believed the engineer who wrote the most lines of code was the best engineer; the best ones tend to make a system smaller. Lines of code measured typing volume. Audit runs on the same kind of proxy measure: the billable hour. Time on the file is the measure of value, because when execution is the constraint, time and value are correlated. More hours genuinely meant more work done.
An auditor who catches the one misstatement that matters in five minutes has done the most valuable thing on the engagement. A timesheet can't record that. If work that took three days now takes an afternoon, measuring by hours reads as the work getting less valuable. The part that's left is worth the most, but it's not what takes the most time. The firms that keep trying to measure time will be measuring the one thing that has stopped mattering.
The one thing that cannot move
If almost everything in the job relocates, one thing stays exactly where it is, and it keeps the auditor at the centre of all of this.
Accountability. When IRBA opens the file, it's your name on the opinion, not a vendor's and not the LLM's. Accountability can be shared with your team, but it can never be handed to the tech, because a tool can't be held accountable. Accountability is what an audit sells. That is the anchor. It is why the manual work can be automated all the way down and the auditor still stays essential. The signature is load-bearing, and only a person can sign it.
A workbook that appears in minutes is a strong start and nothing more, in the way a working prototype is still a long way from a finished product. The gap between something that looks good and something you'd put your name to is the critical one. The first draft is where judgement starts. Checking the references, weighing the exceptions, deciding it's true enough to sign, is the work no tool can do for you. The good ones won't try. They do the grind, show their working, and leave the conclusion where it belongs, with the person whose name goes on the file.
So, what will an auditor do?
What they have always done. The grind was just loud enough, and slow enough, to drown it out.
An auditor is for judgement. For understanding a business well enough to know what should worry them. For choosing the one risk worth chasing out of the thousand that aren't. For being willing to put their name to a conclusion and answer for it. All of that stays. What's changing is the grind.
The job is finally coming into focus. Less mindless manual work, more review and understanding. Fewer hours, higher output. Less grind, more judgement.