Key Takeaways
- Artificial intelligence (AI) can analyze accounting system data to reconstruct a departed controller’s recurring tasks, workflows and processing schedules.
- Audit logs, journal entries, bank rules and reconciliations can reveal accounting patterns and support documented procedures, calendars and stronger internal controls.
- A controller transition creates an opportunity to strengthen segregation of duties, access controls and tax compliance while addressing accounting and reporting risks.
Every finance team has one person like this. 10 years in the seat, never wrote anything down and the job runs because it’s in their head. Then one day they’re gone — no notice, no handoff, no binder of “here’s how I do things.”
Most companies respond by panicking or by hiring an expensive consultant to sit through weeks of discovery meetings that nobody who’s still there can actually answer, because nobody else ever needed to know.
What many business leaders don’t realize is that in most accounting positions, the job wasn’t only in that person’s head. Their work in QuickBooks Online or whatever solution the business is using, contains a critical trail of data: repetitive processing cycles, transactions tied to invoices, email requests or other documentation. With the added power of artificial intelligence, you can quickly recreate, document and structure knowledge about completed work, ongoing procedures, processing schedules and other key information into a project management roadmap for moving forward.
The Reports That Actually Tell You What Someone Did
Most of the answers you need won’t come from the standard financial reports. Those reports show you outcomes, not the behavior. The reports that reconstruct a job are the ones that show pattern and cadence.
One of the most important reports in the system is the audit log. It timestamps every action, by every user, going back as far as your data does. You can filter it to your departed controller’s name to give you a literal timeline: what they touched and on what day of the month, every month. From this information, you can create a working calendar — reconciliations always closed by the 4th, AR follow-up ran every other Tuesday — what days of the month the bills were paid and more.
Recurring transactions and memorized reports also take a lot of the guesswork out of the position. Anything memorized to auto-fire — a bill, an invoice, a journal entry, a report — is something that person built specifically so they wouldn’t have to remember it. Those tasks can be layered into the future calendar, providing key additional pieces of the puzzle.
Additional clues can be easily found by running a year of journal entries, a month-over-month balance sheet and P&L looking for accounts that reset on a schedule, AP and AR aging for payment rhythm, reconciliation history for what’s already gone stale since they left and bank rules for the judgment calls they encoded as if/then logic. These reports don’t just tell you what the job included — they tell you when it happened and why.
Where AI Turns This From Tedious Into Fast
Pulling those reports by hand and reading them is a real project — usually days of cross-referencing spreadsheets, squinting at dates and hoping you didn’t miss the quarterly task buried between two monthly ones. This is exactly the kind of work AI is built for.
A year of audit log activity can be thousands of rows. AI clusters it by day-of-month and day-of-week instantly, surfacing the calendar a person would take a week to piece together by eye.
A year of manual journal entries usually runs a few hundred. Most of them look similar on the surface — an account pairing, a memo, a dollar amount — and a person scanning them will lump the real recurring accrual in with the one-time correction that only happened once. AI can classify each entry by whether it repeats on a schedule with a consistent amount, which is the difference between a useful procedures guide and one bloated with tasks you’ll never do again.
Bank rules are decision logic nobody documented. AI reads a wall of categorization rules and turns them into plain language — this vendor under this amount gets coded here, over that amount gets coded there — so you inherit the reasoning, not just the automation.
From Reconstruction to a Real Manual and Calendar
Once the patterns are classified, the same data becomes the operating procedures manual — not a rewrite of the old process, but the “should be” version. Every recurring task gets written up with what triggers it, which report to pull, the steps to complete it and where the output goes.
The calendar falls out of the same work. Daily is usually bank feed review, deposits and AP entry. Weekly is AR follow-up and payment runs. Monthly carries the heaviest load — reconciliations, accruals, statement review, sales tax, payroll entries. Quarterly picks up estimated payments and board or ownership reporting. Annual is 1099s, W-2 reconciliation, year-end close and tax return support. None of it is a guess at best practice — it’s the actual cadence the business has been running on, made visible.
The Point of the Reset Isn’t What Was Hidden — It’s What Was Accepted
None of this is really about uncovering secrets. Everyone already knew, at some level, that one person doing everything for 10 years meant no real separation between who requests, approves, posts and reconciles. That wasn’t hidden — it was a trade the business made for continuity and trust and it worked, until the day it didn’t. A departure like this is the one moment nobody has muscle memory to defend the old way and it’s worth using it for more than damage control.
Best practices to implement moving forward:
Segregation of Duties: assign every recurring duty to at least two people, with a threshold above which nothing posts without a second signature. Split reconciliation from posting into different hands. Turn the audit log from a forensic tool into a standing one — someone other than whoever’s doing the work checks it monthly.
Quality control and automated review assistance: the same scans that reconstruct the job — duplicate vendors, categorization sitting unresolved, stale reconciling items — aren’t an audit of the past. They’re the baseline for the next chapter. Fix what’s flagged, then set an exception alert for each one.
Access: treat the departure as the reset — audit every user and permission level down to least privilege, enforce MFA everywhere, standardize roles that were once created for one person and put an annual access review on the calendar now.
Don’t Skip the Tax Side
The reconstructed calendar and error findings feed directly into tax planning, not just bookkeeping hygiene. Once the sales tax filing cadence, the 1099 vendor list, depreciation schedules and any multi-state or multi-entity activity are visible, nexus exposure and filing completeness can actually be checked instead of assumed. The gaps found in the error scan often carry tax consequences too — a missed accrual shifts taxable income timing, a misclassified expense affects a deduction. Reconstructing the year cleanly isn’t a separate exercise from tax planning — it’s the foundation tax planning has to sit on.
The Point
The knowledge didn’t actually walk out the door. It was sitting in the system the whole time, in a format no one had a reason to read closely until now. The job isn’t rebuilding what was lost — it’s reading what was already there and using the one moment nobody’s defending the old way to fix what everyone already knew needed fixing.
If you’re staring down a sudden departure and a system full of undocumented history, that history is more recoverable — and more useful — than it feels right now.
Contact Us
PKF O’Connor Davies helps businesses and family offices navigate controller transitions, reconstruct accounting procedures and modernize their financial operations using AI-native tools and processes. To learn more, please contact your client service team or:
Jennifer Katrulya, CPA
Partner
jkatrulya@pkfod.com