AI accounting for a group of companies

Most groups do not have a group finance function. They have three or four companies that each close their own books, and one person who assembles the result afterwards.

That assembly step is where group accounting actually lives, and it is almost always the least systematised part of the whole operation — a workbook, a set of manual eliminations, and knowledge that exists in one head.

Why the group close is a separate process

Each entity can close cleanly and the group still fail, for reasons that belong to the group rather than to any company in it.

The entities close at different speeds. The group waits for the slowest, so group reporting is set by the weakest close in the set rather than the average.

Charts of accounts diverge. Two companies record the same cost under different accounts, mapping is done by hand at consolidation, and the mapping is redone from memory each period.

Intercompany balances do not agree. One entity has recorded a charge the other has not, or has recorded it in a different month. Every group has a persistent difference somebody has stopped trying to explain.

Eliminations are manual and undocumented. They live in the consolidation workbook, not in any ledger, which means the group numbers cannot be traced back to a system by anyone who did not build the workbook.

What automation genuinely helps with

Making the entity closes finish together. Most of the delay in a group close is one entity waiting on reconciliation. Automating reconciliation at entity level compresses the slowest close, and the group date moves because of it. This is indirect and it is the largest effect.

Consistent coding across entities. A shared chart applied automatically at the point of entry removes the mapping step entirely rather than making it faster. See intercompany transactions without the spreadsheet.

Intercompany matching. Comparing what one entity recorded against what the counterparty recorded is exactly the kind of mechanical comparison that is tedious by hand and continuous when automated. Differences surface in days rather than at the year-end audit.

Recording the eliminations. Moving them out of a workbook and into a system with a stated basis is a control improvement independent of any time saved, because it survives the departure of the person who understood them.

What it does not resolve

Whether the entities should have different charts. Sometimes they genuinely should — different businesses, different reporting needs. Automation applies a decision; it does not make one.

The judgement in consolidation. Ownership percentages, minority interests, translation of a foreign subsidiary, whether an arrangement is at arm's length. These are accounting decisions requiring a qualified view.

Group-level policy. If two entities recognise revenue on different bases, consolidating them produces a group figure that means nothing. That is a policy problem and no system will notice it for you.

The intercompany discipline that fixes most of it

One rule prevents most group pain: an intercompany transaction is recorded in both entities in the same period, or in neither.

Simple, routinely broken, and the source of nearly every persistent group difference. A management charge raised in December by one company and recorded in January by the other creates a difference that survives into the audit and consumes hours to explain.

The automated version of the discipline is a matching check that runs before close rather than after, which converts a year-end problem into a five-minute monthly one.

Where a group should start

Entity-level reconciliation at the slowest-closing company. Not at the largest, and not at the group. The group date is set by the slowest entity, so the group benefit comes from fixing that one.

Then intercompany matching. Then the elimination structure. Consolidation mechanics last, because consolidating unreliable entity figures faster is not an improvement.

Common questions

Why is a group close harder than an individual company close?

Because it is a separate process with its own failure modes rather than the sum of the entity closes. The group waits for the slowest entity, charts of accounts diverge and get mapped by hand, intercompany balances disagree, and eliminations usually live in a spreadsheet rather than in any ledger.

How does automation shorten a group close?

Mostly indirectly, by compressing the close of the slowest entity — since that entity sets the group date. Directly, it removes the manual mapping between divergent charts of accounts, matches intercompany balances continuously instead of at year end, and puts eliminations into a system where their basis is recorded.

What is the single most useful group discipline?

Recording every intercompany transaction in both entities in the same period, or in neither. Most persistent group differences come from one side recording a charge in a month the other did not, and an automated matching check run before close turns a year-end investigation into a routine monthly task.

What should a group automate first?

Reconciliation at the entity that closes slowest, rather than at the largest entity or at group level. The group reporting date is determined by the last entity to finish, so that is where the group-level benefit comes from. Consolidation mechanics should come last, because consolidating unreliable figures faster does not help.

Related: intercompany transactions without the spreadsheet · month-end without the scramble · AP automation for a multi-entity group