How to Calculate ROI for AI Automation (Step-by-Step)
A step-by-step framework for calculating real ROI on AI automation — with example scenarios for missed calls, follow-ups, and admin work.
Short answer
ROI for AI automation = (revenue gained + cost saved − total cost) ÷ total cost. The three biggest sources of return are recovered missed-call revenue, faster lead follow-up that lifts conversion rates, and admin hours eliminated. Most well-scoped builds for small service businesses pay back in 1–4 months and return 3–8x in year one.
The basic ROI formula
ROI = (gain from automation − cost of automation) ÷ cost of automation. To make it real, gain has two components: new revenue captured (deals you would have lost) and cost saved (hours you no longer pay for). Add both, subtract the all-in monthly cost, and divide by that cost.
Step 1: list what the automation is doing
- Answering calls (after hours / overflow)
- Texting back missed calls
- Qualifying inbound leads
- Booking appointments
- Logging conversations into the CRM
- Sending follow-ups and reminders
Step 2: estimate revenue gained
For each task, estimate how many additional deals it produces. The biggest levers are speed-to-lead and after-hours capture.
| Lever | How to estimate |
|---|---|
| Missed calls recovered | Missed calls/mo × close rate × avg job value |
| Faster lead response | Extra leads converted per month from <5-min response × avg job value |
| After-hours bookings | After-hours leads/mo × close rate × avg job value |
| Re-activated stale leads | Leads recovered/mo × close rate × avg job value |
Step 3: estimate cost saved
Multiply hours saved per week by the loaded hourly cost of the person doing the work today (salary + benefits + overhead, usually 1.3–1.5× base wage). Common targets: front-desk call handling, manual data entry into the CRM, follow-up email/SMS, appointment scheduling.
Step 4: total the cost of the automation
- Build / setup (one-time, amortized over 12 months)
- Monthly platform fees
- Telephony / SMS / AI usage
- Ongoing maintenance or support
Example scenarios
| Scenario | Monthly gain | Monthly cost | ROI |
|---|---|---|---|
| HVAC missed-call text-back (20 missed/mo, $500 job, 33% close) | $3,300 | $400 | 725% |
| Med spa AI booking after hours (15 leads/mo, $300 service, 40% close) | $1,800 | $600 | 200% |
| Law firm intake automation (5 cases/mo recovered, $3,000 avg) | $15,000 | $1,200 | 1,150% |
| Admin time saved (20 hrs/wk × $35/hr loaded) | $3,000 | $500 | 500% |
How to justify the investment internally
Frame it in payback period, not just ROI. 'This pays for itself in X weeks and produces $Y in year one' is far more persuasive than a percentage. Pair the number with one concrete example from the team — a real missed call, a real lost lead, a real night spent on data entry.
How HD² Labs approaches ROI
Every engagement starts with a baseline: current call volume, missed-call rate, average deal size, current conversion rate, and current admin hours. We build the ROI model with those numbers — not generic industry averages — and review it again 60 days after launch using real system data. If the math doesn't work, we say so before we build.
Frequently asked questions
How fast does AI automation usually pay back?
1–4 months for most small service businesses. Higher-ticket industries (legal, HVAC replacement, dental implants) often see payback in weeks.
What's the most common ROI mistake?
Counting only hours saved and ignoring revenue gained. The bigger number is almost always the deals you weren't capturing before.
Should I include intangible benefits?
Mention them, but don't rely on them in the model. Stick to dollars: revenue captured, hours saved, churn reduced. Intangibles (faster response, better customer experience) are the bonus.
What if my close rate isn't well tracked?
Use a conservative range (20–30% for most service businesses) and build the model against the low end. If the ROI works at the low end, the actual return will be better.
Can you give me a custom ROI estimate?
Yes. Book a strategy call and bring last month's call log, average job value, and current close rate — that's enough to build a realistic ROI model on the call.