Finance operations: where the hours actually go.
We analysed time-and-motion data across 112 finance transformation engagements between 2023 and 2026. The pattern of where the hours land is remarkably consistent — and so is the mismatch between where the hours are and where the automation budget gets spent.
Methodology
The data comes from process diagnostics we run at the start of finance transformation engagements. In each, we time-stamp every meaningful activity in the month-end close and the days bracketing it, across roles from staff accountants to controllers. The 112 engagements cover financial services, public sector, telco, energy, and consumer industries, in companies ranging from $200M to $40B in revenue.
The numbers are averages. Individual finance teams vary, but not as much as you'd expect — the variance between a $500M telco and a $20B bank is smaller than the variance between two business units inside the same bank.
The breakdown
Reconciliations (31%)
The single biggest sink. Bank recs, intercompany recs, sub-ledger to GL recs, balance-sheet recs. The work has three components, in this order of effort:
- Matching — finding the line on one side that corresponds to the line on the other. Boring, voluminous, and overwhelmingly automatable.
- Researching the breaks — when something doesn't match, figuring out why. Less voluminous, more variable, but still highly patternable.
- Posting the corrections — relatively small effort.
What we see in benchmarks: reconciliation tools have been deployed in 70%+ of the firms we work with, but they typically automate the matching layer only. The break-research layer is still done by humans reading explanations from other humans. This is exactly the work the new generation of document-understanding and reasoning agents handles well.
Journal entries & accruals (22%)
Recurring entries, accrual estimates, prepayment amortisations, intercompany allocations. Most of the volume is recurring. Most of the recurring volume could be parameterised and posted automatically with simple controls. In practice, we still find finance teams in the largest banks manually posting the same 600+ recurring entries every month.
The blocker is rarely technology. It's that the entries were originally set up by people who left, the supporting calculations live in spreadsheets, and nobody has the appetite to clean up and re-platform.
Variance analysis (17%)
The "why did this number change?" work. Half of the hours go into finding the variances. The other half go into explaining them — which means an analyst reads a P&L, identifies the outliers, traces them to source, and writes commentary.
The finding half is largely a solved problem with modern FP&A tools. The explaining half is where AI is now changing the economics. Agent-based "first-pass commentary" — where the agent reads the variances, pulls context from prior periods and the underlying transactions, and drafts a paragraph for the analyst to review — saves 60–70% of the analyst's time per close cycle in the engagements where we've deployed it.
Management reporting (14%)
The packs. The board pack, the exec pack, the divisional pack. The hours go into chasing inputs from across the business, reconciling versions, and reformatting. Almost none of it is genuinely analytical work — it's data plumbing dressed up as analysis.
The way teams that have solved this report differently: they treat the management pack as a product, with a fixed schema, a single source of truth, and a generation pipeline. The analyst's job becomes commentary and exception spotting, not assembly.
Audit & controls support (9%)
Sample pulls, walkthroughs, evidence preparation, follow-up. Highly seasonal, lumpy across the year. The volume per audit cycle has crept up as control environments have tightened. Nobody has materially automated this end-to-end, but the ingredients are now there: structured controls libraries, evidence pipelines from source systems, and AI-assisted sample analysis.
Other (7%)
Tax provisions, treasury reporting, regulatory reporting, ad-hoc analysis, and project-related work. Distributed across many small activities. Less automation upside per activity, but real cumulative effect.
The mismatch
Now overlay the time distribution above with where finance teams typically spend their transformation budget over the last three years.
| Activity | Share of hours | Share of transformation spend |
|---|---|---|
| Reconciliations | 31% | 22% |
| Journal entries & accruals | 22% | 11% |
| Variance analysis | 17% | 9% |
| Management reporting | 14% | 34% |
| Audit & controls support | 9% | 7% |
| Other | 7% | 17% |
The mismatch is striking. Management reporting absorbs 34% of the transformation spend but represents 14% of the hours. Journal entries absorb 11% of the spend for 22% of the hours. The reason is unglamorous: reporting is visible to executives, journal entries are not. Spend follows visibility, not effort.
What to do about it
1. Re-budget against the time map
Take the time-and-motion data for your own team. Compare to your transformation budget. Where the budget under-weights the hours, that's where the next dollar of investment has the highest return.
2. Attack the journal-entry mass
The fastest large win in most finance teams is parameterising recurring entries with controls and review. It's not glamorous, it doesn't demo well, and it's high single-digit FTE recovery in most teams. Start here, fund the next thing with the savings.
3. Move reconciliations from match-only to break-resolution
If you already have a matching engine, the next layer is automating the research of the breaks. Document-understanding agents do this work well. The hours saved per break are small individually, but cumulative.
4. Re-imagine variance commentary
First-pass commentary drafted by agents, reviewed by analysts, signed off by managers. The analyst's role shifts from data archaeology to judgement. Most analysts prefer the new role.
5. Treat the management pack as a product
One schema, one source of truth, one generation pipeline. The reorganisation effort is the work — the technology is comparatively cheap.
What changes when this works
In the engagements where the full re-architecture has landed, the finance function looks different in three ways:
- Close timeline compresses from 8–10 working days to 3–5.
- FTE mix shifts: fewer junior analysts on plumbing, more senior analysts on commentary, controllers spending more time on partnering, less on chasing.
- Audit cost drops 15–25% because evidence is generated by the system, not assembled on request.
The dollars saved are real, but the bigger change is that finance becomes a function people want to join again — because the work is mostly judgement and partnering, not data assembly.