AI in financial consolidation automates the mechanical work of a multi-entity close: mapping subsidiary accounts to the group chart, matching intercompany transactions, applying FX rates and drafting variance commentary. Finance staff approve every mapping, adjustment and narrative. For Australian groups with three to fifteen entities, the right consolidation tool depends on entity count, ERP mix and currency complexity.
Key takeaways
- AI is good at pattern matching across ledgers: account mapping, intercompany matching, missing FX rates and first-draft commentary.
- Policy choices, adjustments above materiality and statutory sign-off stay with people.
- Financial consolidation tools fall into four tiers: tightened spreadsheets, off-the-shelf tools, a custom AI layer over existing ERPs, and enterprise platforms.
- Consolidation data is highly sensitive, so settle where it is processed, what is logged and who can see outputs before choosing a tool.
- A well-scoped rollout usually takes 12 to 16 weeks, including at least one close run in parallel with the existing process.
Why is multi-entity consolidation so slow?
Ask a CFO at an Australian group with five or more entities about month-end. The books close on the 1st, the first consolidated draft lands well into the second week, the board starts asking questions, auditors want workings, and the team is already starting the next close.
In between sits a wall of spreadsheets: a mapping file translating each subsidiary's chart of accounts to the group, an intercompany matrix that never quite balances, an FX tab that breaks when someone adds a row, and commentary stitched together from emails with subsidiary controllers.
The ERPs exist. Xero, MYOB, NetSuite, SAP Business One and Dynamics 365 all offer some consolidation capability. The problem is that consolidation is still a judgement-heavy task that needs someone who understands the business to interpret the numbers. AI earns its keep as an assistant for the mechanical part, so people can focus on judgement.
Why do mid-sized groups feel this hardest?
Single-entity businesses run one ledger. The largest corporates run enterprise consolidation platforms with dedicated teams. Groups with a handful of trading entities sit in the middle: too complex for single-entity accounting, too small for an enterprise platform. So they run on Excel and overtime.
Typical pain points:
- Mixed ERPs across entities. An acquisition runs MYOB, the original business runs Xero, the NZ subsidiary has its own Xero organisation, and the investor wants a weekly rolling view.
- Intercompany mismatches. Entity A books an invoice in March, Entity B receives it in April. Multiply by dozens of transactions and eliminations become detective work.
- FX translation. An NZ or Singapore subsidiary means monthly translation and tracking of the foreign currency translation reserve, often by hand.
- Management commentary. Variance explanations usually take several rounds of emails with subsidiary controllers.
- Audit trail. When the auditor asks "why did you adjust this?" six months later, the answer is often in someone's inbox.
What can AI automate in financial consolidation?
Across our consultants' careers on group reporting and ERP integration in large resources organisations, one pattern holds: AI is excellent at pattern matching across large volumes of structured and semi-structured data, and weak on policy decisions, unusual transactions and anything an auditor will want justified.
| Human Judgment Required | AI Can Automate | |
|---|---|---|
| Account mapping | Approving new mappings, policy choices | Suggesting mappings based on account name, historical usage |
| Intercompany matching | Resolving disputes between entities | Matching invoices by amount, reference, and date tolerance |
| FX revaluation | Choosing policy (closing vs average rates) | Applying rates, flagging missing rate data, reconciling the translation reserve |
| Variance commentary | Final narrative, tone, audit-facing explanations | First-draft commentary by pulling subsidiary context |
| Adjustments | Booking any adjustment above materiality | Proposing routine reclass entries for review |
| Audit evidence | Signing off workpapers | Assembling evidence trail, linking source documents |
What does an AI-assisted close look like?
- IngestPull trial balances from each entity's ERP via API
- MapAI maps subsidiary accounts to group chart, humans approve exceptions
- MatchIntercompany matching engine flags unmatched items with explanations
- TranslateFX translation applied, reserve calculated, missing rates flagged
- DraftAI writes first-draft variance commentary from source data
- ReviewCFO and controller review, adjust, and sign off before release
Where do the time savings come from?
The biggest win is usually the time spent chasing information, more than the final reports. In many groups, a large share of the close is waiting for subsidiary controllers to explain why a number moved. An AI layer with read-only access to subsidiary ledgers and prior-period commentary can draft that explanation, which the controller confirms or corrects. That one change, done properly, can take days out of the close.
Can AI handle financial consolidation across multiple ERPs?
Yes. This is one of the strongest cases for an AI layer. Trial balances are pulled from each system through its API, accounts are mapped to the group chart with AI suggestions and human approval, and intercompany items are matched across systems regardless of which ledger they sit in. The ERPs stay as they are, which avoids forcing a migration just to consolidate.
Which financial consolidation tools fit which groups?
How many entities do you consolidate, and what is your growth trajectory?
- 2 to 3 entities, stable, single currencyDo nothing new. Tighten Excel workflow, standardise chart of accounts.
- 3 to 8 entities, growing, single or dual currencyOff-the-shelf consolidation tool with AI features (e.g. Fathom, Spotlight Reporting)
- 5 to 15 entities, multi-currency, investor-backed or IPO trackCustom AI layer on top of existing ERPs with governance wrap
- 15+ entities, complex group, regulated reportingEnterprise consolidation platform (OneStream, CCH Tagetik, Workiva)
| Strengths | Limits | |
|---|---|---|
| Tightened spreadsheets | No new software, full flexibility | Key-person risk, fragile formulas, weak audit trail |
| Off-the-shelf tools (Fathom, Spotlight Reporting, Joiin) | Fast to start on Xero or MYOB, growing AI features | Struggle with unusual structures, JVs, complex intercompany |
| Custom AI layer over existing ERPs | Handles mixed ERPs, data stays in your environment, fits your structure | Needs a well-scoped build and an owner |
| Enterprise platforms (OneStream, CCH Tagetik, Workiva, Oracle) | Deep consolidation, regulatory reporting, controls | Long implementations, priced for large finance functions |
Off-the-shelf consolidation tools with AI features
Fathom, Spotlight Reporting and Joiin have been adding AI for variance commentary and anomaly detection. For groups of three to eight entities on Xero or MYOB, they work well. If you have unusual structures, material joint ventures or complex intercompany arrangements, they reach their limits.
A custom AI layer over your existing ERPs
For mixed ERPs or unusual structures, a custom layer is increasingly practical. It sits over Xero, MYOB, NetSuite or SAP and does four things: pulls trial balances on a schedule, handles mapping and intercompany matching, drafts commentary, and writes back adjustments once a person approves them. It can run on your infrastructure in Australia, so group financials never leave your environment. Our guides to consolidating multiple Xero organisations and Xero automation beyond basic sync cover the integration patterns, and our system integration work builds them.
Enterprise consolidation platforms
OneStream, CCH Tagetik, Workiva and Oracle's consolidation products are strong. They are also built for groups with dedicated finance transformation budgets and appetite for a long implementation. Many mid-sized groups are not there yet.
What data governance questions come first?
Consolidation data includes full wage bills, margin by product line, intercompany pricing a tax office may examine, M&A activity and forward-looking commentary. Answer these before choosing technology:
- Where is the data processed? If an AI tool routes trial balances through an overseas-hosted model, your group P&L has left Australia. This matters most for investor-backed groups and those approaching a capital event.
- What is logged, and by whom? If a model provider keeps prompts for training, intercompany pricing and margins could end up in someone else's training data.
- Who can see outputs? Access inside the AI system must mirror access controls in your ERPs.
- How are audit trails kept? Every AI-generated mapping, adjustment proposal and commentary draft needs timestamps and user attribution for the auditors.
Treat any AI consolidation tool like a new financial system, with a full security review before anyone uploads a trial balance. Our AI agent governance guide and Privacy Act compliance guide go further.
How long does an AI consolidation rollout take?
A well-scoped program typically runs 12 to 16 weeks from decision to first AI-assisted close, assuming the underlying data is in reasonable shape.
- Weeks 1 to 3Discovery and data auditMap every entity's ERP, chart of accounts, FX policy, and intercompany processes. Identify data quality gaps.
- Weeks 4 to 6Design and governanceDesign the AI layer, governance model, access controls, and audit trail requirements. Sign off data sovereignty decisions.
- Weeks 7 to 10Build and integrateBuild the integration layer, mapping engine, and commentary generator. Parallel run with existing Excel process.
- Weeks 11 to 13Parallel closeRun one or two full closes in parallel. Compare AI output to human output. Tune mappings and prompts.
- Weeks 14 to 16Go live and tuneFirst AI-assisted close as the primary process. Continue tuning based on controller feedback.
What payback should a CFO expect?
Build the case on your own close data.
- Finance team days saved per closeMeasure in parallel run
- Less external audit effort from cleaner workingsAsk your auditor
- Earlier close and faster board reportingQualitative
- Less reliance on one person's knowledgeQualitative
- Year oneOften around break-even
- Year two onwardGains compound
Year one is often around break-even once implementation is included. From year two, mappings, commentary style and intercompany patterns are established and gains build with every close.
What will AI not do in consolidation?
AI will not replace the group financial controller. It will not decide whether to book an adjustment above materiality or settle policy questions on consolidation method, non-controlling interests or hedge accounting. It will not sign the statutory accounts or sit across from the auditor.
It will reclaim hours spent on mechanical work, which in most groups is the difference between a finance team that reports the past and one that helps run the business. New presentation requirements such as AASB 18 income statement categories make clean, consistent mapping even more valuable. For adjacent use cases, see month-end close automation and the Xero reconciliation agent guide.
How Solve8 can help
Our consultants and engineers map where your close loses time, then build the consolidation layer over the ERPs you already run, on your infrastructure in Australia, with your controllers approving every mapping and adjustment. It runs alongside your existing process until it proves accurate. See our data, BI and analytics services or talk to us.
