Automation business cases have a credibility problem, and it is self-inflicted. The standard model multiplies a task's duration by its frequency, applies a fully loaded hourly rate, assumes 80% of that time disappears, and reports a return north of 300%. Finance approves it because the number is large, and eighteen months later nobody can find the savings in the general ledger.
The savings usually were real. The model just measured the wrong thing. Hours saved across forty people at ninety minutes a week do not show up as reduced cost unless headcount changes or those hours are redeployed to work that generates measurable value. Neither happens by default, and neither is what most business cases actually promise.
A defensible automation ROI model does three things differently. It counts the full cost of ownership rather than the license fee. It separates savings that will appear in a budget line from savings that will not. And it commits in advance to how each benefit will be measured, by whom, and against what baseline. That discipline produces a smaller headline number and a business case that survives contact with the CFO a year later.
The Cost Side: What Gets Left Out
Software cost is typically 30% to 50% of the three-year total. The rest is spread across categories that rarely appear in a vendor quote.
| Cost category | Typical share of 3-year TCO | Frequently omitted? |
|---|---|---|
| Platform licensing and usage | 30-50% | No |
| Implementation and configuration | 15-30% | Sometimes |
| Integration to existing systems | 10-25% | Often |
| Process redesign and documentation | 5-15% | Almost always |
| Change management and training | 5-15% | Almost always |
| Ongoing maintenance and support | 10-20% annually | Often |
| Internal staff time during rollout | 10-20% | Almost always |
Three of these deserve specific attention because they are the ones that turn a good business case into a bad outcome.
Internal time is a real cost. The process owner, the subject matter experts, the IT resource who builds the integrations — their time is not free, and during a rollout it is substantial. Costing this at zero is the most common single error in automation business cases, and it understates TCO by a meaningful margin.
Maintenance scales with brittleness. An integration built against a stable API needs occasional attention. Screen-level automation against an application that ships quarterly UI changes needs continuous attention. Ask, for any proposed approach, what breaks it and how often that thing changes.
Process redesign is not optional. Automating a process without fixing it first locks in its defects permanently. Budget for the redesign or accept that you are buying a faster version of the wrong process.
The Benefit Side: Two Categories, Different Rules
Every benefit belongs in one of two buckets, and they should never be added together into a single headline figure.
Hard savings: traceable to a financial statement
These are the benefits a controller can verify.
- Headcount avoided. Not reductions — avoided hires. If volume is growing 25% a year and automation lets the team absorb it without adding two people, that is two salaries not spent, and it is defensible.
- Vendor and license consolidation. Retiring a point tool the automation replaces.
- Early payment discounts captured. A direct, measurable cash benefit when cycle time drops below discount windows.
- Late fees and penalties eliminated. Interest charges, statutory filing penalties, missed deadline costs.
- Error correction cost. Duplicate payments recovered or prevented, incorrect payroll runs avoided, rework hours eliminated on a measured error rate.
- Overtime reduction. Particularly around close and payroll cycles, where the peak load is what drives the overtime.
Soft savings: real, valuable, and not bankable
These belong in the business case but in a separate section with explicit labeling.
- Capacity redeployed to analysis rather than processing.
- Faster cycle times improving customer or employee experience.
- Reduced compliance and audit risk.
- Better decision quality from more current data.
- Lower turnover in roles where the work was primarily manual.
Soft savings frequently exceed hard savings in magnitude. They are also the ones executives discount to zero when the results are reviewed. The correct treatment is to present them, quantify them where a proxy exists, and build the payback calculation exclusively on hard savings. A case that pays back on hard savings alone and delivers soft benefits on top is a case that gets approved twice — once at funding and once at review.
A Worked Model
Consider a 900-person organization automating accounts payable, processing 4,000 invoices per month.
Baseline. Fully loaded cost per invoice of roughly $14, covering four AP staff, approver time, error correction, and overhead. Annual processing cost near $672,000. Median cycle time of nine days. Early payment discounts captured on about 12% of eligible invoices. Two duplicate payments per year averaging $8,000 each.
Three-year cost of ownership. Platform and usage at $84,000 per year. Implementation, integration, and redesign at $110,000 in year one. Internal staff time valued at $45,000 in year one and $12,000 annually after. Total three-year cost of approximately $437,000.
Hard savings, year two onward. Cost per invoice falls to roughly $4.50, a saving of $456,000 annually — but only $228,000 of that is bankable, because two of the four AP roles are redeployed rather than eliminated and their salaries remain in the budget. Discount capture rising from 12% to 70% of eligible invoices adds $61,000. Duplicate prevention adds $16,000. Bankable hard savings: roughly $305,000 per year.
Payback. Year one costs of $239,000 against partial-year savings of about $130,000. Cumulative break-even lands around month sixteen. Three-year net benefit on hard savings alone is approximately $310,000, a return of roughly 71% — nowhere near 300%, and far more likely to be true.
The redeployed capacity, faster close, reduced fraud exposure, and improved supplier relationships are all real. They sit in the soft column, clearly labeled, and they are the reason the process owner wants the project. They are not the reason finance funds it.
Measuring Honestly
The measurement plan matters more than the model, because the model is a forecast and the measurement is what determines whether you get funded again.
Establish the baseline before anything changes. This is the step everyone skips. Once the new process is live, the old numbers are unrecoverable and every subsequent claim becomes an argument. Capture cost per transaction, median and 90th percentile cycle time, error rate, rework hours, and volume for at least one full cycle prior to go-live. Instrumented real-time financial reporting makes this considerably easier, since the baseline can be pulled from live data rather than reconstructed from memory.
Name an owner for every benefit line. Each claimed saving needs a person who agrees to be measured on it and a specific report it will appear in. Benefits without owners do not materialize.
Separate volume growth from efficiency. If transaction volume rises 20% while total cost stays flat, that is a 20% efficiency gain that a naive before-and-after comparison shows as zero. Always measure on a per-transaction basis.
Review at a fixed interval with the original numbers visible. Six and eighteen months, with the forecast side by side against actuals. Variance is expected; unexplained variance is the problem. This review is also where the soft benefits get revisited, and where the case for extending automation to the next process is either made or lost.
Where the Returns Are Largest
Return scales with transaction volume, the amount of manual data movement per transaction, error cost, and how much value is trapped by cycle time. Processes strong on all four — invoice processing, payroll, expense handling, reconciliation, onboarding — reliably produce defensible cases. Processes that are low-volume, highly variable, or already digital rarely do, regardless of how frustrating they feel to the people running them.
The second wave of return comes from reuse. The first automated process carries the full cost of platform selection, integration, governance, and organizational learning. The third carries almost none of it, which is why programs that plan a sequence outperform programs that fund one project at a time. The same effect appears in adjacent domains — the case for payroll automation becomes substantially stronger once approval routing, audit logging, and system integration already exist from an earlier project.
Building a case that holds up starts with a measured baseline and an honest split between what will appear in the budget and what will not. If you want help structuring that model against your own volumes and cost per transaction, our team can work through a business process automation assessment with you and produce numbers your finance function will recognize as their own.
