Report results: which consolidated reporting steps to automate and which need a person

Software can translate currencies, run eliminations, roll up cost centre figures and draft variance commentary from clean ledgers. People have to judge unusual movements, sign off on what the group tells its board and regulators, and own the story. None of this works until entity data maps to a single chart of accounts.

Where the effort really sits

On paper, reporting results looks like a roll-up. Subsidiaries close their books, the group collects them, translates them, removes intercompany balances and publishes. In practice most of the effort goes into chasing late submissions, reconciling intercompany mismatches and rewriting commentary that arrived in the wrong format. The publishing itself is quick. Anyone planning to automate this process should start by finding where the hours disappear, because that is rarely where the process map says they do.

What software or AI can take over today

Currency translation is the clearest candidate. Once rate tables are loaded from an agreed source and the rules for average versus closing rates are set, a consolidation tool applies them without error. The same goes for foreign exchange conversions on intercompany payments, provided the payment records carry the original currency and transaction date.

Intercompany matching also suits automation. A system can pair receivables in one entity with payables in another, flag differences above a tolerance and propose elimination entries. It cannot decide which side is wrong, but it narrows the argument to a short list.

Cost reporting by cost centre, project or responsibility segment is largely mechanical once the hierarchy is stable. Allocations, roll-ups and comparisons to budget can run on a schedule.

Cash forecasting benefits from pattern-based models that learn from past receipts and disbursements. These produce a reasonable baseline projection. Large or unusual expected movements still need someone to confirm them with treasury.

Generative AI is now useful for first drafts of variance explanations. Fed with account movements and prior commentary, it writes a plausible paragraph. Treat that paragraph as a starting point for a controller, never as finished text.

What still needs a person

Judgement on materiality stays human. A model can show that a balance moved; deciding whether the movement matters to investors, lenders or a regulator is a professional call.

Topside adjustments belong with named owners. When the group corrects a subsidiary figure at consolidation, someone must be able to explain why and defend it to auditors.

The narrative that goes to the board needs an author. Results are a message as much as a dataset, and the finance lead decides what to emphasise, what to caveat and what to hold back pending review.

Disputed intercompany balances need negotiation between entity controllers. Software surfaces the gap. Closing it often involves a phone call and a decision about which ledger to correct.

Final sign-off and certification cannot be delegated to a tool in any jurisdiction worth operating in.

What has to be true about the data first

Every entity must map its local accounts to the group chart, and that mapping must be maintained by someone with authority to change it. Unmapped or loosely mapped accounts are the most common reason automated consolidations produce numbers nobody trusts.

Intercompany transactions need a shared reference that both sides record. Without it, matching falls back to amount and date, which breaks as soon as fees or timing differences appear.

Exchange rates must come from one agreed source, stored with effective dates, and applied by rule. Rates typed in by hand at entity level undermine everything downstream.

Cost centre and project hierarchies should change through a controlled request, not by editing a spreadsheet. Reorganisations that are not reflected in master data make trend reporting meaningless.

Receivable and collection data, including delinquent balances, should be reconciled to the ledger before it feeds group reporting. Where external parties such as credit agencies or tax authorities receive debt information, the reporting extract and the ledger must agree.

Finally, submissions need a fixed format and a fixed cut-off. Automation cannot absorb entities that send figures in whatever layout suited them that period.

Questions to ask the people who run the process

What documentation says and what controllers do often diverge. These questions tend to expose the gap:

• Which entity's numbers usually arrive last, and what are you waiting on when they do?
• What do you fix by hand after the system has run, and how long has that been happening?
• Where do you keep the adjustments that never make it into the ledger?
• When intercompany balances disagree, who decides which side changes?
• Which exchange rates do you actually use, and where do they come from?
• Is there a spreadsheet that sits between the consolidation tool and the board pack? Who owns it?
• Which commentary do you rewrite every period because the entity version is unusable?
• What would break if a particular person were away during close?

The answers usually reveal a shadow process of workarounds. Automating the documented version while that shadow process carries on simply adds a second system to reconcile.

Where these projects tend to stall

Most stall at master data. Teams buy a consolidation or AI tool, connect it to entity ledgers and discover that mappings, hierarchies and intercompany references were never consistent. Fixing that foundation is slower and less visible than configuring the software, yet it decides whether the automated output is used or quietly replaced by the old spreadsheet.

APQC's Process Classification Framework® (PCF) is an open standard developed by APQC, a nonprofit that promotes benchmarking and best practices worldwide. To download the full PCF or to view definitions and measures, please visit www.apqc.org/pcf.