What does your first AI agent actually return?
Edison AI's ROI calculator estimates the annual savings, implementation cost, payback period, and three-year ROI of a first AI agent for an Australian SMB. Inputs in AUD. Methodology and assumptions are fully disclosed below the calculator. No email gate.
Inputs
Defaults reflect Edison AI's published baselines for a 4-person Australian SMB team rolling out a first agent. Override every cell to model your business.
How many people directly hand work to the new agent.
Salary × ~1.3 (super, leave, overheads) ÷ ~1,800 productive hours/year.
Edison AI baseline: 6-12 hours/week for a well-scoped first agent.
Australian default is 46-48 (excludes annual + public + sick leave).
Edison AI sprint fee for a single in-production agent.
Ongoing LLM API + infra + monitoring (Edison AI baseline: ~$6,000).
Share of saved hours that actually convert to capacity or revenue. Default 70%.
Output
Your first agent, in numbers.
- Annual savings
- $66,976
- Implementation cost (year 1 total)
- $28,000
- ROI. Year 1
- +139%
- ROI.3 years
- +402%
- Payback period
- 4.3 months
1,030 hours/yr × $65/hr
$22,000 build + $6,000 tooling
(savings − year-1 cost) ÷ year-1 cost
Build amortised across 3 years of tooling
Implementation ÷ monthly net savings
Illustrative model. Edison AI shipped agents typically deliver year-one ROI between 300% and 900% for well-scoped first builds. Variance is dominated by realisation factor and hours-saved accuracy.
Pressure-test estimates with usHow the model works (and where to push back).
1. Gross hours saved
affected staff × hours saved per week × working weeks per year. The unrealised upper bound. What the agent removes if every freed hour stayed productive.
formula: staff × hours/wk × weeks/yr
2. Realised hours
Multiply gross by a realisation factor (default 0.70). Most teams convert about 60-80% of saved hours into actual capacity or revenue; the rest is absorbed by slack, meetings, or context-switching.
formula: gross hours × realisation
3. Annual savings
Realised hours × loaded hourly cost. Use loaded cost (salary × ~1.3 for super, leave, overheads, ÷ ~1,800 productive hours/year) rather than gross salary. It's the true marginal cost of a staff hour.
formula: realised hours × $/hr
4. First-year cost
Implementation fee + one year of tooling (LLM API, infra, monitoring). Edison AI's published baseline for a single agent: $18-28k build, $4-9k/year tooling.
formula: build + tooling/yr
5. ROI (%)
(annual savings − annual cost) ÷ annual cost × 100. Reported for year one and across three years. Year one is dominated by the build cost, year three by ongoing tooling.
formula: (savings − cost) / cost × 100
6. Payback period
Implementation cost ÷ monthly net savings. Edison AI's published median across shipped agents is 4-9 months. Anything over 18 months suggests the workflow doesn't justify the build.
formula: build / (annual savings − tooling) × 12
What the model deliberately excludes
- Revenue lift. The model is hours-saved only. It doesn't price the upside of faster lead response, higher renewal capture, or better margin discipline. For most Edison AI builds, revenue lift is 1.5-3× the hours-saved value. Add it after the baseline holds up.
- Risk-mitigation value. Audit-readiness, file-note completeness, compliance comms consistency. Real value, hard to quantify in this model. Treat as a qualitative tiebreaker between candidate workflows.
- Change-management cost. The build fee covers Edison AI delivery; it does not cover internal change-management time (typically 20-60 hours of the operations lead's time across the first 90 days). Add to the cost side if you want a tighter total-cost picture.
- Failed pilots. Edison AI scopes for one agent in production, not three pilots hoping one lands. If your buying process is exploratory rather than committed, multiply the build cost by 1.5-2× and revisit.
Edison AI baselines (the published numbers we use)
| Variable | Default | Sensible range |
|---|---|---|
| Loaded hourly cost | $65/hr | $45 – $180 depending on role |
| Hours saved per week (per person) | 8 hrs/wk | 4 – 14 hrs/wk for a first agent |
| Realisation factor | 0.70 | 0.55 – 0.85 |
| Implementation cost | $22,000 | $18,000 – $28,000 |
| Annual tooling cost | $6,000 | $3,000 – $9,000 |
| Working weeks per year | 46 | 44 – 48 |
Take it to a director. Or take it to us.
Twenty minutes is enough to pressure-test the realisation factor, sense-check the first workflow, and price the actual build. Fixed-fee audit available if the numbers hold up.