Pattern Chaser: Audit Analytics for Internal Audit

PATTERN CHASER

Audit Analytics for Internal Audit in Financial Services

The Replaceable Half

Half your audit work already runs without you. The real question is not whether the machine takes it, but which half you have spent your career becoming excellent at.

Hey Reader 👋

A reconciliation that used to take a junior the back half of a fieldwork week now runs on the full population before lunch. The break list comes out cleaner than the manual one ever did.

Nobody mourns the lost days. The work was real. It was also the part of the job a script was always going to take.

Half of audit work is already a script. Pulling the sample, tying out the recs, extracting the exceptions, filling the workbook. The other half is judgement, and it never showed up in a workbook. The first part is the replaceable half.

Most of us spent our careers being measured on the replaceable half, because it is the half that is easy to inspect. Nobody chose this. The system selected for it.

So the question is not whether you will be replaced. It is which half you are spending your week feeding. Today: what the split really is, the five skills that make up the half a script cannot reach, and an honest self-audit you can run before anyone runs it on you.

In today’s issue:

↳

The two halves of every audit job, and why you were trained to perfect the wrong one

↳

The five skills that make up the irreplaceable half, and how to spot your weakest

↳

The honest self-audit: which half is your week actually feeding

Examples are composites drawn from industry experience and peer conversations.

DEEP DIVE

Which Half Is Your Week Feeding?

The two halves

Pick any audit task. It falls on one side of a line. On one side: pulling the sample, tying out the recs, extracting the exceptions, populating the workbook. Mechanical, rule-checkable, the same every time.

On the other side: what the exception actually means, whether it is a timing difference or a control that stopped operating, what to test next, what to escalate, the story that makes the board act.

The first side runs without judgement, so it runs without you. The second side has never run without a person, because it is judgement.

Here is the uncomfortable part. The replaceable half is the half that felt like the job. It is visible, it is effortful, it fills a week and a workbook. The review process can see it, so the review process rewards it. Nobody chose to perfect the half that was always going to leave. The incentives chose it for us.

The five skills a script cannot reach

The irreplaceable half is not one thing. It is five skills. They do not move together, and most auditors are strong in two and quietly exposed in one. The exposed one is usually the one holding the others back.

Skill 1 · Data Capability

Working With Data, Not Just Having It

Strong looks like shaping and trusting the data yourself, before fieldwork starts, instead of auditing whatever extract IT hands you. The trap is starting every audit one step behind the data. The extract you are given has already had decisions made about it. On one review the population handed over had quietly excluded reversed transactions, as a courtesy. The reversals were the audit.

Skill 2 · Analytical Judgement

Turning Exceptions Into Decisions

Strong looks like forming the risk question first, so 200 exceptions become something you can answer instead of a list you have to clear. More exceptions are not better coverage. Sometimes they are just a bigger manual review problem. On one audit, 200 flags, 195 were timing noise. Five had re-drawn the scope by lunchtime. The skill was knowing which five mattered, not producing the 200.

Skill 3 · AI Fluency

Directing AI, Not Fearing It

Strong looks like being able to audit an AI-driven control on its methodology, and using the tools in your own work. This is not coding. It is knowing what to ask of a model and what to distrust. An AI risk score looked authoritative until someone asked what population it had learnt from. The answer reshaped the finding. Fluency is asking that question before the model’s output becomes your conclusion.

Skill 4 · Board Narrative

Turning a Finding Into a Board Decision

Strong looks like a finding that changes a decision, because it is framed as risk to strategy, not findings to file. Brilliant analysis that dies in a dashboard nobody reads changes nothing. The same finding, written two ways. One version went in the file. The other named the decision the committee had to make, and the committee made it.

Skill 5 · Learning Posture

Staying Current as the Half-Life Shortens

Strong looks like building new capability deliberately, on your own system, not while waiting for the profession to set a direction. Capability decays. The auditors who pull ahead are not the ones who did a course. They are the ones who picked one technique, applied it on a live audit and kept what worked. Moving only when everyone moves is, by definition, late.

If you are thinking you have tried this and it was not practical, read the five again. Not one of them needs a new tool, a budget or a line of Python. They are not projects. They are where you choose to spend the judgement you already have.

The honest self-audit

Here is the lens. Take your last fieldwork week and split it down the middle. How many hours went on producing the data, and how many on deciding what the data meant. Most honest answers come back heavily weighted to the first half. That is not a personal failing. It is what the review process rewarded.

Run it at the start of scoping, not mid-fieldwork, while there is still time to move your hours. Then look at the five skills and find your weakest. That is not the one to feel bad about. It is the one where the least effort buys the most ground. Lift it and the others move with it.

The question was never whether you will be replaced. It is which half you are feeding, and whether you can move your week into the half a script will never reach.

WHAT’S POSSIBLE

The First Pass Runs Itself

Most audits still spend their first half producing the data. Pull the population, clean it, match it, extract the exceptions. Only then does the judgement work begin, usually with the clock already half gone.

Now picture the mechanical half wired as standing infrastructure. Every period, the full population is pulled from source, reconciled, the exceptions extracted and risk ranked, before the audit even opens. You do not start fieldwork by producing the data. You open it at the judgement half. Here are the twelve exceptions that survived the ranking, which three change the scope.

The machine does the replaceable half, continuously, across the whole population. The human starts where judgement begins. Your scarcest resource stops being spent on production.

This is not a tool you install next week, and continuous monitoring has a long history of becoming one more dashboard nobody reads. The difference is the intent. The first pass is not the answer. It is the thing that hands your hours back to the half that needs them. The skill is knowing what the first pass should surface, and that skill is yours, not the machine’s.

The work is not standing still. The rule-checkable half is being automated first, and it is the half most auditors still spend most of their week on. Standing still is not holding position. It is drifting toward the work that gets designed out.

You did not learn this craft to pull samples and tidy workbooks.

You learnt it to find what others miss and say what others will not. A script can do the first half. It will never do the second.

The half a script can do is not the half worth defending.

Run the split on one week this week. Then hit reply and tell me which way it came out. I read every one.

Before you go

The self-audit, scored

The split tells you which half your week is feeding. The Future-Ready Auditor Scorecard tells you which of the five skills to move first. Two minutes, five dimensions, no audience. You find out instead of wonder.

Get your score

WHAT’S ON MY RADAR

Worth your time this week.

🔍 THE PATTERN

Why Are There Still So Many Jobs?. An MIT economist on the thing this whole issue sits inside: a job is a bundle of tasks, automation takes the routine ones, and the value moves to the tasks it cannot. The replaceable half is a task, not your job. Free, and the clearest 28 pages on why.

🧠 THE MIND

The Labor Illusion. People rate a service as more valuable when they can see the effort behind it, even when the result is identical. It is why the visible, effortful half felt like the job. You were rewarded for the labour being seen, not for what it found.

📊 THE PROOF

Radiology was supposed to be gone by now. In 2016 the godfather of AI said stop training radiologists. Since 2018 their caseloads are up around a quarter and demand keeps climbing. The machine took the read. The judgement, the integration, the call stayed human. The cleanest live proof that the second half survives.

🗞️ WHERE IT’S HEADING

The IIA’s AI Auditing Framework. Being able to audit an AI-driven control on its methodology is moving from specialist edge to baseline expectation. The third skill on this list is quietly being written into the profession’s floor. Worth reading before it is assumed of you.

Know a colleague who’d find this useful? Send it on.