TL;DR:
- ROI = (Value Generated − Total Cost) ÷ Total Cost × 100 — the formula’s simple; getting the inputs right isn’t
- Most calculators omit implementation time, LLM API costs, and maintenance overhead — include all three
- Most automations still deliver strong returns when calculated honestly; 300% ROI beats the vendor’s 2,000% fiction, but it’s still excellent
Every vendor selling automation software has an AI automation ROI calculator on their website. These calculators share a common design philosophy: inflate the numerator, minimise the denominator. Hours saved are generous. Implementation costs are absent. LLM API costs don’t exist. This is the version that includes those costs.
The Basic ROI Formula
ROI (%) = (Value Generated − Total Cost) ÷ Total Cost × 100Value generated = tangible, measurable benefit: time saved at a monetary value, errors prevented, or additional revenue enabled.
Total cost = everything: tool subscriptions, implementation time, LLM API costs, and ongoing maintenance.
Always specify a time horizon — all calculations below use 12 months. Year 1 has worse ROI than Year 2+ because implementation cost is front-loaded.
Calculating the Value Side
Time Saved
Use fully-loaded hourly cost — not salary. Multiply annual salary by 1.3 to account for benefits, employer’s National Insurance, and management overhead. A £45,000/year employee costs roughly £58,500/year fully loaded, or around £28/hour.
One critical caveat: time saved only generates value if the employee reallocates it to higher-value work. Automating 3 hours of data entry for someone who fills those 3 hours with lower-priority tasks doesn’t improve your ROI.
Error Reduction
Calculate this as error occurrences per month × cost to fix each error × expected reduction rate (automation typically eliminates 80–95% of rule-based errors).
Example: 20 data entry errors/month × 45 min to fix × £28/hr = roughly £5,040/year in avoided rework.
Throughput Increase
The strongest ROI case: automation handles volume that would otherwise require hiring. If your support team handles 50 tickets/day manually and automation enables 80/day without new headcount, the value ceiling is the fully-loaded cost of the hire you avoided.
The Full Cost Side
Include every item below — this is where most automation ROI calculations fail.
- Subscription cost — price at your actual usage tier, not the free plan you started on
- Implementation time — hours × fully-loaded rate; a developer building a workflow at £75/hr × 20 hrs = £1,500
- Maintenance — budget 10–20% of implementation cost per year; APIs change, tokens expire, integrations break
- LLM API costs — measure tokens per run before deploying; at 1,000 runs/month with 2,500 tokens average on Claude Sonnet, that’s roughly $13.50/month
- Monitoring overhead — someone watches failed runs; at £25/hour, 30 min/week = around £650/year
Three Worked Case Studies
Case 1: Lead Enrichment — ROI 1,205%
10 sales reps each spending 3 hours/week on manual CRM data entry. Automation enriches leads automatically.
- Annual time value: 1,560 hours × roughly £24 = £37,440
- Conservative revenue lift (5% better conversion): ~£16,300
- Total value: ~£53,700/year
- Total Year 1 cost: ~£4,100 (Make Pro + Clearbit API + LLM API + implementation + maintenance + monitoring)
- ROI: ~1,200%
Case 2: Support Ticket Triage — ROI 1,449%
8-person support team triaging 100 tickets/day. An LLM classifies, prioritises, and routes each ticket.
- Conservative time savings (20% efficiency gain): 4,333 hours/year × £18 = ~£78,000
- Total Year 1 cost: ~£5,000 (n8n Cloud + LLM API + implementation + maintenance + monitoring)
- ROI: ~1,450%
Case 3: Monthly Financial Report — ROI 296%
CFO + analyst spending 8 hours generating the monthly report. Automation cuts that to 20 minutes of review.
- Annual time value: 92 hours × £80 blended rate = £7,360
- Total Year 1 cost: ~£1,850 (Zapier + LLM API + consultant + maintenance)
- ROI: ~296%
This is the lowest case because the process runs once a month. High-frequency automations generate more value because fixed costs amortise across more runs.
When ROI Doesn’t Pencil Out
Watch for these red flags before building. Fewer than 20 runs/month rarely justifies implementation cost unless time saved per run is substantial. High judgement requirement means most iterations need nuanced human input — error costs can exceed savings. If you can’t write down every step of the manual process, you’re not ready to automate it. And sometimes the right answer is eliminating the process entirely, not automating it.
Your ROI Template
| Line Item | Your Value |
|---|---|
| Hours saved per week × 52 × fully-loaded £/hr | _____ |
| Error reduction annual value | _____ |
| Total Annual Value | _____ |
| Platform subscription (annual) | _____ |
| Implementation hours × hourly rate | _____ |
| LLM API cost/run × monthly runs × 12 | _____ |
| Maintenance (15% of implementation) | _____ |
| Monitoring (hrs/month × 12 × £/hr) | _____ |
| Total Cost Year 1 | _____ |
| ROI = (Value − Cost) ÷ Cost × 100 | ___% |
Year 1 ROI below 100% is marginal. Below 50%, build a stronger case before proceeding. Above 300%, the economics are clear — prioritise implementation.
Bottom Line
Build the spreadsheet before you build the automation. When you include real implementation costs, real LLM API costs, and real maintenance overhead, most automations still deliver strong returns — just 300% ROI instead of the vendor’s 2,000%. That’s still an excellent investment. Know what return you expect before you start, and measure it for the first 3 months after launch.