Who checks the checker? Review is recurring work too

Open any firm's task list and you will find "prepare VAT return", "reconcile bank accounts" and "produce management pack". Look for "review VAT return" and you will usually come up empty. The accounting review process is some of the most skilled recurring work a firm does, and it is the part most often left off the list. It happens in the gaps of a manager's week, it has no owner on paper and no checklist anyone can find, and it is the first thing to slip when the close gets tight. This post makes one argument: review steps are recurring tasks, and they deserve the same treatment as the work they check.

Why review steps vanish from the task list

Nobody decides to hide review. It disappears for ordinary reasons.

• It rides on the preparer's task. The list says "VAT return, Client A" with one tick box. Preparation and review share the tick, so the review has no row of its own.
• It is done by someone else. The preparer owns the task. The reviewer is borrowed for twenty minutes and is not named anywhere.
• It feels like supervision, not work. Managers often think of review as part of being a manager, so they do not count it, and nobody else counts it for them.
• It has no written method. Ask a reviewer what they check and the honest answer is often "I look at the variances and I know when something is off". That is real skill, and it lives in one head.
• It happens wherever there is time. Between meetings, late in the day, on the train. Work done in the gaps is invisible by nature.

The result is a task list that describes the doing and leaves out the checking. Capacity plans built on that list are wrong in a predictable direction.

What invisible review costs

Here is an illustrative calculation, not research. Take a hypothetical bookkeeping firm with 40 clients on monthly management accounts. Say each file takes a manager 20 minutes to review, and a quarter of them need a second look after corrections. That is about 16 hours a month of review, sitting on one or two people, appearing on no list and in no capacity plan.

Those hours carry three costs.

1. A bottleneck. If review is not on the calendar, nobody sees that 30 files land on one manager on working day 4. The close runs late, and the delay looks like the preparers' fault.
2. Key person risk. The reviewer is often the only one who knows what a normal month looks like for a given client. That is a single-owner task, even though no task list says so. We covered how to find those in what happens to your month end close if one person quits.
3. Lost learning. When review finds a mistake, the fix is made and the file moves on. The preparer may hear about it, may not. The same mistake comes back next month, and review catches it again.

Put the accounting review process on the list

The fix is not a new control framework. It is treating each review as a recurring task with the same few facts every other task carries. If you have already built your recurring task list, this is a second pass over it.

Give each review its own task

Split "VAT return, Client A" into two rows: prepare and review. The review row gets its own owner and a backup, and a calendar rule tied to the preparation, for example "two working days after preparation is done". Now review shows up on the calendar, it shows up in the workload, and you can see who carries it.

Write down what the reviewer checks

This is the hard part and the valuable one. Sit with the reviewer for one file and write down what they look at, in order. Not "check it looks right", but the actual checks: input tax against last quarter, any single purchase over a set amount, reverse charge items, the reconciliation of the VAT control account. Most reviewers find their own method clearer once it is written, and the backup can now do a real review rather than a nervous one.

Record what review finds

Keep a short log against each review: what was corrected, and why. Over a few months, patterns appear. The same client, the same kind of error, the same missing document. Each pattern is a line to add to the preparer's checklist, so the fix happens upstream next time instead of being caught downstream again.

Review matters more when agents draft

There is a forward-looking reason to get this right now. The sensible way to bring AI into a finance team is the ladder we described in drafts first, trust later: an agent drafts, a person approves, and the task earns more freedom from its track record. On that ladder, review is the human's job and the agent's teacher.

That has two consequences. First, if review time is not counted, any claim about hours returned is overstated, because the hours moved from preparing to checking and nobody recorded the checking. Second, the review log becomes the agent's improvement loop. A correction written down with its reason can become a better instruction. A correction made silently teaches nothing to anyone, person or agent.

This is how we are designing Beaks: review as a step with its own owner, checklist and history, rather than a tick on someone else's task. A spreadsheet with a separate review row and a corrections column will get a small firm most of the way.

What this means for the people who review

Said badly, "put review on the list" can sound like checking up on managers. It is the opposite. The people who review are doing skilled, recurring work that has gone uncredited for years. Making it visible gives them a fair workload, a backup who can genuinely cover their holiday, and a written record of the judgement they bring. It also makes plain where their time should go as routine preparation gets easier: into the reviews that need judgement, not into re-checking the same simple errors every month.

One thing to do this week

Pick one recurring deliverable, such as the monthly management pack for your largest client. Split it into prepare and review on your task list, name the reviewer and a backup, and ask the reviewer to spend fifteen minutes writing down what they actually check. If they cannot, or if nobody else could do it next month, you have found a single-owner task that was never on the list.