May 15, 2026 · 3 min read

How to Calculate AI ROI — A Framework for Engineering Leaders

Your team spends $5,000/month on AI tools. Is it worth it? Most companies can’t answer this question because they don’t measure AI ROI. Here’s a framework that works.

The formula

AI ROI = (Value Generated - Total AI Cost) / Total AI Cost × 100%

The hard part is measuring “Value Generated.” Here’s how.

Measuring value: AI coding tools

For AI coding assistants (Claude Code, Cursor, Copilot):

Time saved

Hours saved/week = (Tasks completed with AI - Tasks without AI) × Avg hours/task

Measure this by:

Track how long tasks take WITH AI assistance (2 weeks)
Compare to historical data WITHOUT AI (or run a control group)
Multiply by developer hourly cost

Example:

Developer costs $75/hour (fully loaded)
AI saves 5 hours/week per developer
10 developers on the team
Value: 5 × $75 × 10 × 4 weeks = $15,000/month

Code quality improvement

Harder to measure, but track:

Bug rate before/after AI adoption
Code review turnaround time
Production incidents related to code quality

If AI reduces bugs by 20% and each bug costs $500 to fix (developer time + user impact), that’s measurable value.

Developer satisfaction

Developers who use AI tools report higher satisfaction and are less likely to leave. Developer turnover costs $50-150K per person (recruiting + onboarding + lost productivity). Even a small retention improvement has massive ROI.

Measuring value: AI in production

For AI features in your product (RAG, chatbots, AI search):

Metric	How to measure
Revenue impact	A/B test: conversion rate with AI vs without
Support cost reduction	Tickets deflected by AI chatbot × cost per ticket
User engagement	Session length, retention, NPS with AI features
Time to value	How fast users achieve their goal with AI assistance

Measuring costs

Direct costs

Cost	Monthly estimate
AI coding tools (per developer)	$20-40
LLM API costs	$100-5,000
GPU infrastructure	$0-2,000
Observability	$0-100
Hosting for AI features	$14-200

Indirect costs

Cost	Estimate
Engineering time building AI features	Hours × hourly rate
Testing and evaluation	10-20% of build time
Security and compliance	5-10% of build time
Training team on AI tools	2-4 hours per person

Example ROI calculation

AI coding tools for a 10-person team

Costs:

Claude Code Pro: $20/dev × 10 = $200/month
Cursor Pro: $20/dev × 10 = $200/month (some devs prefer this)
Total: $400/month

Value:

Time saved: 5 hrs/week × $75/hr × 10 devs × 4 weeks = $15,000/month
Bug reduction (20%): 10 bugs/month × $500/bug × 20% = $1,000/month
Total: $16,000/month

ROI: ($16,000 - $400) / $400 = 3,900%

Even if the time savings are half what you estimate, the ROI is still 1,900%. AI coding tools are the easiest AI investment to justify.

AI chatbot for customer support

Costs:

LLM API: $500/month
Hosting: $50/month
Engineering (build + maintain): $2,000/month (amortized)
Total: $2,550/month

Value:

Tickets deflected: 500/month × $15/ticket = $7,500/month
Faster resolution: 200 tickets × 10 min saved × $0.50/min = $1,000/month
Total: $8,500/month

ROI: ($8,500 - $2,550) / $2,550 = 233%

When AI ROI is negative

Not every AI investment pays off. Common money losers:

AI features nobody uses — built it because it’s cool, not because users need it
Over-engineered solutions — used Claude Opus when DeepSeek would suffice
No cost optimization — paying 10x more than necessary for API calls
Pilot that never ships — spent 6 months building, never reached production

See our build vs buy guide for avoiding the most expensive mistakes.

Presenting ROI to leadership

Keep it simple:

We spend $X/month on AI tools.
This saves Y hours of developer time worth $Z/month.
Net benefit: $Z - $X = $W/month.
ROI: W/X × 100 = N%.

Include one concrete example (“Last week, Claude Code wrote the entire test suite for the payment module in 20 minutes. That would have taken a developer 4 hours.”).