AI ROI calculator & methodology
Turn AI investments into numbers your CFO defends.
A practical AI ROI framework for SMBs — four levers, five steps, and a free assessment that scores your actual stack against ranked opportunities. No vague productivity-uplift percentages.
Built for 30–300 person businesses · Microsoft / Google / Salesforce stacks
The framework
Every AI ROI rolls up to one of four levers
Mixing them is how business cases get killed in board meetings. Naming them is how they get approved.
Hours redirected
The cleanest, fastest-to-measure ROI lever. Hours your team spends on a workflow today × (1 - automation rate) × loaded hourly cost. Track via Copilot dashboard, Power Automate run history, or simple before/after diary studies.
Revenue accelerated
Pipeline that closes faster because proposals are drafted in minutes, RFP responses are pre-populated, or sales reps spend more time selling. Hardest to attribute, biggest upside when you can.
Costs avoided
Headcount you didn't need to add as volume grew. Outsourced work pulled back in-house. Tool subscriptions retired because Copilot covered the gap. Usually the most defensible line in a board pack.
Risk reduced
Fewer compliance escalations, fewer missed SLAs, fewer escaped errors in contracts and invoices. Harder to count, but cheap insurance — typically expressed as expected-cost-of-incident × incident-rate-reduction.
The methodology
Five steps to a defensible AI ROI number
Same framework we use inside every client engagement. Take it, use it.
Inventory the workflow
For each candidate AI use case, write down the workflow in plain English: who touches it, how often, how long it takes today, and what 'done' looks like. If you can't describe it in 5 lines, the use case isn't scoped enough yet.
Pick the right ROI lever
Map the workflow to one of the four levers above. Most workflows have a primary lever ('this saves 4 hours per rep per week') and a secondary one ('and reps close 8% more pipeline because of it'). Be explicit; mixing them hides assumptions.
Build the baseline, then the target
Baseline = current cost or revenue, with no automation assumptions. Target = honest, defensible end-state — usually 40-70% workflow automation, not 100%. A 50% number you can hit beats a 95% number you can't.
Subtract the all-in cost
Licence cost (per user/month × users × 12), implementation cost (one-off, amortised over year 1), and ongoing run cost (1 FTE part-time for adoption, training, prompt curation). Net annual benefit = (lever × outcome) - all-in cost.
Validate inside 90 days
The model is only as good as the next 90 days of measurement. Build the telemetry into the rollout itself — Copilot Dashboard, Power BI usage, before/after KPIs. Re-forecast quarterly with real data, not the original spreadsheet.
Worked example
Copilot rollout, 120-person services firm
Real numbers from a typical Microsoft 365 Copilot pilot. Round numbers; same shape as most of our client cases.
Inputs
- 120 knowledge workers × $90/hr loaded cost
- Target persona: sales, ops, finance, exec assist
- Baseline ~5 hrs/week spent on "Copilot-eligible" tasks
- Honest end-state: 35% of those hours redirected
Outputs
- Hours saved/year: 120 × 5 × 0.35 × 48 = 10,080 hrs
- Gross benefit: 10,080 × $90 = $907,200
- All-in cost: licences + rollout = ~$280,000 year 1
- Net year-1 ROI: ~3.2× · Payback ~4 months
Assumes Microsoft 365 Copilot at list price for 120 seats and a 4-6 week pilot-to-rollout engagement. Actual results depend on persona mix, baseline hours, and adoption discipline. The bigger lever for most services firms is actually revenue accelerated via faster proposals and RFPs — typically another 1.5-2× on top of the hours line. We model both in the assessment.
The tool
What the Star AI ROI Lab will compute
Currently in active build. In the meantime, the free AI Readiness Assessment already produces the inputs the Lab will consume.
Workflows inventoried
Opportunities scored
Year-1 ROI ranges modelled
Microsoft tools mapped per use case
FAQ
Common questions on AI ROI
What's a realistic AI ROI for a small business?
For a well-scoped Copilot or Power Automate rollout, most SMBs see 3-7× return inside year one once you net out licence and implementation cost. Custom-built AI agents on Azure OpenAI can hit 10× or more, but with longer payback (4-6 months instead of 4-6 weeks).
How do you calculate hours saved without surveying everyone?
Three sources that work in combination: (1) Copilot Dashboard usage telemetry, (2) Power Automate run history with average human-time-equivalent, (3) before/after diary studies with a representative 8-12 person sample. Skip "how productive do you feel?" survey questions — they're unreliable.
Should we include "soft" benefits like morale and retention?
In an internal business case, yes — but as a separate line, never bundled into the headline ROI. CFOs discount soft benefits to zero in the first cycle. Earn the headline number on hard levers, then add soft benefits as upside.
When is AI ROI negative?
When you pay for capability you don't adopt. The #1 negative-ROI pattern: buying 200 Microsoft 365 Copilot seats, no training, no adoption telemetry, 12% weekly active usage at month 6, renewal denied. The licence wasn't the problem; the activation plan was.
How is this different from a generic AI ROI calculator online?
Generic calculators ask for inputs you don't know yet ("expected productivity gain %") and produce numbers you can't defend. Our assessment maps your actual stack to scored opportunities first, then models ROI per opportunity using ranges grounded in real client benchmarks. You leave with a backlog, not just a number.
Get a personalised AI ROI shortlist in 4 minutes
Free assessment. Scored opportunity backlog. Year-1 ROI ranges per opportunity. No call required.