Resources

How to Work Out ROI with AI

A framework to calculate whether AI investment pays for itself.

What to measure

ROI for AI isn't just about cost. Value can come from several places:

Time saved: Hours per month that staff no longer spend on repetitive tasks. Multiply by hourly rate.
Revenue uplift: Faster response to leads, better conversion, more deals closed.
Risk reduced: Fewer errors, better compliance, fewer escalations. Harder to quantify but real.
Quality improved: Better summaries, more consistent outputs, fewer rework cycles.

Start with what you can measure. Time saved is usually the easiest to baseline. Revenue and risk come next.

Inputs

Cost side: Licensing (cloud APIs, SaaS, or self-hosted infra), implementation (setup, integration, training), and ongoing (support, updates, power if self-hosted). Don't forget implementation—it's often 20–40% of Year 1.

Value side: Hours saved × hourly rate. Or revenue uplift from faster sales cycles. Or a conservative estimate of risk avoided. Be realistic. It's better to under-promise and over-deliver.

Quick ROI calculator

A simple time-saved model. Enter your numbers to see monthly value, payback, and annual net benefit.

Monthly AI cost (£)

Hours saved per month

Hourly rate (£)

Common mistakes

Ignoring implementation cost: Setup, integration, and training can add 20–40% to Year 1. Include it.
Overestimating savings: "We'll save 10 hours a week" often becomes 3. Baseline first. Measure after.
Not baselining: You can't prove ROI if you don't know where you started. Capture current state before you change it.
Treating it as one-off: AI costs recur. Licensing, power, support. Model ongoing, not just Year 1.

When ROI is unclear

Sometimes the numbers aren't clear upfront. That's okay. Pilot first. Run a small-scale trial. Measure time saved, quality, or throughput. Then scale if the data supports it.