How to Calculate ROI on AI Automation

Most companies can't answer a simple question: is our AI spend making us money or costing us money? Forrester found fewer than 1 in 3 AI decision-makers can tie AI value to the P&L. MIT research puts it more bluntly — 95% of GenAI pilots fail to demonstrate ROI.

The math isn't the hard part. Here's the formula:

Step 1: Calculate labor value recovered

Take the hours your automations save per week, multiply by 4.33 (average weeks per month), then multiply by your fully-loaded labor cost per hour. Fully-loaded means salary plus benefits, taxes, and overhead — typically 1.25 to 1.4 times base pay. If your team saves 40 hours a week at $50/hour fully loaded, that's 40 × 4.33 × 50 = $8,660 per month in recovered labor value.

Step 2: Subtract your total AI spend

Count everything: platform subscriptions (Zapier, Make, n8n, Power Automate), API usage fees (Anthropic, OpenAI), per-operation charges, hosting for self-hosted tools, and maintenance time converted to dollars. Against $2,000/month in spend, our example nets $6,660.

Step 3: Divide net value by spend

$6,660 ÷ $2,000 × 100 = 333% monthly ROI. Every dollar spent recovers $4.33 in labor value.

What counts as a good number?

Healthy automations typically return 200–500% monthly. Below 100%, the automation costs more than it saves and should be reviewed — not necessarily killed, but reviewed.

The three mistakes that inflate AI ROI

Leaving out API usage. Subscriptions are easy to count. Usage-based API bills grow quietly as adoption grows, and they're the most commonly omitted cost.

Ignoring maintenance time. Someone rebuilds the workflow when a vendor changes their API. That's labor cost, not free.

Counting gross instead of net. "This automation saves us $8,660 a month" is a gross number. The board wants net — what's left after the tools are paid for.

Why a one-time calculation goes stale

AI costs drift. Vendors reprice, API bills scale with usage, automations silently break. A number calculated in January is fiction by March. That's the argument for tracking ROI continuously rather than in an annual spreadsheet exercise.