Direct answer
Calculate a baseline for task volume, time, loaded cost, quality, delay, and outcome. Estimate theoretical benefit, discount it for adoption, review, exceptions, and actual capacity capture, add defensible revenue or risk value, then subtract implementation and ongoing operating cost. Report a range, payback, and the assumptions most likely to change the result.
Automation ROI bridge
Translate task-level time into captured annual value
Illustrative annual net value
$300$24,300 captured value from $54,000 theoretical capacity
Do not count all theoretical time as cash. Confirm how recovered capacity will actually be used.
Step 1
Measure the unit of work before changing it
| Baseline | Definition | Common mistake |
|---|---|---|
| Volume | Tasks, cases, requests, or records per period | Using a memorable busy week |
| Active time | Labor actually spent on the unit | Ignoring wait and handoff time |
| Loaded cost | Compensation plus relevant employment cost | Treating salary as the only cost |
| Quality | Errors, corrections, rework, or review score | Counting speed while quality falls |
| Outcome | Revenue, cycle time, customer result, or risk | Assuming activity equals business value |
Step 2
Bridge theoretical savings to captured value
Theoretical capacity equals volume multiplied by time per unit and loaded cost. Captured capacity discounts that amount for adoption, residual manual work, exceptions, review, and whether the organization can put the time to useful work.
Do not call returned capacity cash savings unless spending actually decreases. It may instead support more volume, better service, reduced backlog, improved quality, or avoidance of future hiring.
Step 3
Keep benefit categories distinct
| Benefit | Evidence | Conservative treatment |
|---|---|---|
| Capacity | Measured time reduction at actual adoption | Apply a capture factor |
| Revenue | Incremental qualified volume, conversion, or margin | Count contribution, not gross revenue |
| Quality | Reduced correction, rework, refund, or escalation | Use observed unit cost |
| Speed | Economic impact of shorter cycle time | Avoid double-counting capacity |
| Risk | Lower expected loss or stronger control | Document probability and consequence assumptions |
Step 4
Use total cost, not the model invoice
- Discovery and current-state measurement
- Process, experience, and control design
- Software, models, automation, and infrastructure
- Integration and data preparation
- Security, privacy, legal, and procurement review
- Evaluation, testing, training, and rollout
- Human review, support, monitoring, and improvement
- Internal owner and subject-matter expert time
Step 5
Present conservative, expected, and upside cases
Conservative
Lower adoption, more review, slower improvement, and full cost. Ask whether the project still deserves discovery.
Expected
Use the most defensible assumptions supported by current evidence.
Upside
Show reuse or stronger adoption as potential, not as the approval case.
Sensitivity
Identify which assumption changes ROI most, often volume, capture, adoption, or operating cost.
Post-launch review
Replace assumptions with measured values and decide whether to scale, redesign, or stop.
The value point
After this page, you should be able to decide:
Whether expected captured value justifies the complete investment under conservative assumptions.Your working output should be a task-to-value calculation, capture factor, cost model, confidence range, and measurement plan.
Questions business leaders ask
Frequently asked questions
What is the formula for AI automation ROI?+
First-year ROI is first-year net benefit divided by first-year investment. Net benefit includes captured capacity and other defensible benefits minus implementation and operating cost.
Should all time saved be counted as savings?+
No. Discount for adoption, review, exceptions, and whether the organization can use the returned time. Treat it as cash savings only when spending is actually reduced.
How should revenue benefits be counted?+
Use incremental contribution attributable to the system, not gross revenue. State conversion, margin, and attribution assumptions and avoid counting the same effect twice.
When should ROI be measured again?+
Measure during controlled release, after adoption stabilizes, and at planned operating reviews. Model behavior, workflow, prices, volume, and business conditions change.
Research anchors
Primary and authoritative sources
- U.S. Small Business Administration: AI for small business↗
- NIST AI Risk Management Framework↗
- NIST AI RMF Playbook↗
Examples and planning ranges are clearly labeled. Source terms, provider behavior, and regulations can change; verify current requirements for your organization and jurisdiction.
Prepared and reviewed by the Future Made Useful systems editorial team. Material guidance reviewed July 17, 2026.