BY NUMBER7 AIThe base labor equation
For one measurement period:
Transaction labor cost = V × H × T × M × C
Where:
- V = in-scope transaction volume;
- H = share of transactions with at least one human touch;
- T = mean touches per handled transaction;
- M = mean active hours per touch;
- C = loaded labor cost per hour.
If minutes are used, divide by 60:
Transaction labor cost = V × H × T × minutes per touch ÷ 60 × loaded hourly cost
The equation is more diagnostic than a single average cost per invoice. It shows whether value comes from fewer handled transactions, fewer back-and-forth touches, faster decisions, or a different labor mix.
Complete ROI model
Annual net value = labor value + capacity value + quality/control value + SLA/working-capital value − software − implementation − change − ongoing oversight − residual operating cost
ROI = (annual net value ÷ total investment) × 100
Payback months = upfront and annualized investment relevant to payback ÷ monthly realized net benefit
Do not double count. If avoided hiring is used as capacity value, do not also treat the same hours as cash payroll reduction.
Worked example: multi-client accounting operation
Assume an illustrative firm processes 20,000 eligible invoices per month.
Current state
- transaction-touch rate: 42%;
- touches per handled transaction: 2.4;
- active minutes per touch: 3.2;
- loaded labor cost: $32/hour.
Current monthly transaction labor:
20,000 × 0.42 × 2.4 × 3.2 ÷ 60 × $32 = $34,406
Modeled future state
- transaction-touch rate: 18%;
- touches per handled transaction: 1.6;
- active minutes per touch: 2.4;
- same loaded labor cost.
Future monthly transaction labor:
20,000 × 0.18 × 1.6 × 2.4 ÷ 60 × $32 = $7,373
Modeled gross monthly labor value is about $27,034, or $324,408 annually. If total first-year software, implementation, change, and oversight investment is $120,000, the simplified first-year ROI is:
($324,408 − $120,000) ÷ $120,000 = 170%
Simplified payback is approximately 4.4 months.
This is an illustrative scenario, not a Number7 result or market benchmark. A real business case must validate whether the released capacity produces cash reduction, avoided hiring, additional client revenue, faster close, improved SLA, or another outcome.
Capacity economics for accounting firms
Accounting firms often should not begin with staff reduction. Their opportunity may be to serve more clients, absorb seasonal peaks, improve turnaround, or move experienced people from repetitive processing into review and advisory work.
Model capacity explicitly:
Additional monthly transaction capacity = released productive hours ÷ future human hours per in-scope transaction
Then convert only the portion that can be operationally sold:
Realizable capacity value = additional supported client volume × expected contribution margin × utilization factor
The utilization factor prevents the model from valuing theoretical capacity as guaranteed revenue.
The finance-BPO capture factor
Gross productivity value does not automatically belong to the service provider.
BPO retained value = gross productivity value × provider capture factor
The provider capture factor depends on the contract:
| Contract model | Typical value path |
|---|---|
| Fixed fee/output | Provider may retain productivity value until repricing |
| FTE/time-and-materials | Client may expect headcount or rate reduction; provider capture can be low |
| Gainshare | Value is divided by the agreed method |
| SLA/outcome based | Provider may earn margin from efficient delivery if outcomes hold |
| Minimum volume/step pricing | Value changes at volume thresholds |
The model should also include transition obligations, redeployment, bench cost, retraining, quality governance, and contract-renewal effects.
Value categories and proof requirements
| Value claim | Required evidence |
|---|---|
| Reduced labor expense | Payroll/FTE/overtime change tied to the workload |
| Avoided hiring | Approved demand plan and actual hires avoided |
| Added client capacity | New volume served without proportional labor, plus realized margin |
| Lower rework | Reopen/correction time and root-cause trend |
| Faster cycle | Receipt-to-terminal timestamps using the same population |
| Better SLA | Contractual performance and penalty/bonus impact |
| Lower duplicate/error loss | Verified incidents and controlled counterfactual; avoid speculative totals |
| Improved control | Audit/control outcome and exception evidence, not a vague risk percentage |
Counterexample: 10,000 hours “saved,” no return
A project estimates ten minutes saved per invoice but does not measure whether people actually stop performing the task. Staff continue reviewing every item because confidence and controls are unclear. The spreadsheet shows large savings; the operating model does not change.
The corrected approach measures transaction-touch rate, active minutes, and terminal-state outcomes before and after implementation. Value is recognized only when the work or business capacity changes.
Sensitivity analysis
At minimum, run conservative, expected, and upside cases for:
- volume;
- baseline and future touch rate;
- touches/minutes;
- adoption ramp;
- exception mix;
- loaded labor cost;
- implementation time;
- software/oversight cost;
- capacity utilization;
- BPO provider capture factor.
Show which assumption drives the result. If the business case works only with the upside case, it is not yet robust.
Limitations
This model does not quantify every risk or strategic effect. Loaded labor costs, time observations, and future touch rates can be uncertain. Automation may shift work to implementation, master data, exception governance, or suppliers. Currency, tax, accounting, and contracting assumptions require qualified review.
ROI is decision support, not a guaranteed result.
Sources
- IBM: straight-through processing — general context for touch-free transaction flow.
- Number7 Research: human-remainder and touch-rate economic model, revision 1.0.
How to cite
Number7 Research. “How to Calculate AP Automation ROI for Accounting Firms and Finance BPOs.” Number7AI, revision 1.0, 7 September 2026. https://number7ai.com/research/ap-automation-economics.
Revision history
| Revision | Date | Change | Reviewer |
|---|---|---|---|
| 1.0 | 7 September 2026 | Initial touch-based economics, capacity model, BPO capture factor, and example | Number7 AI editorial review |
Next step
Build the business case from a measured current-state action log, not assumed minutes per invoice.