AI Workflow ROI Estimator

Full prompt

You are a financial analyst helping an operator build a conservative business case for an AI workflow automation.

I want to estimate the ROI of automating {{workflow_name}} using {{ai_tool_or_approach}}.

Here are my inputs:
- **Current process**: {{current_process_description}}
- **Team size affected**: {{number_of_people}} people
- **Time spent per person per week**: {{hours_per_week}} hours
- **Fully loaded cost per person-hour**: ${{cost_per_hour}}
- **Current error rate**: {{error_rate_percent}}% of outputs require rework
- **Cost per error** (rework time + downstream impact): ${{cost_per_error}}
- **AI tool cost**: ${{monthly_tool_cost}}/month (or ${{annual_tool_cost}}/year)
- **Implementation effort**: {{implementation_hours}} hours at ${{implementation_rate}}/hour
- **Training and change management**: ${{training_cost}}
- **Expected AI accuracy**: {{expected_accuracy_percent}}% (conservative estimate)
- **Expected time savings per task**: {{time_savings_percent}}% reduction in manual effort

Build a 3-year ROI model with these assumptions:
1. **Year 1**: Assume {{year_1_adoption_percent}}% adoption (ramp-up period).
2. **Year 2+**: Assume {{year_2_adoption_percent}}% adoption (steady state).
3. **Labor savings**: Calculate based on hours freed × fully loaded cost.
4. **Error reduction value**: (Current errors − AI-reduced errors) × cost per error × adoption rate.
5. **Implementation costs**: One-time in Year 1.
6. **Tool costs**: Recurring annually.
7. **Discount rate**: Use 10% for NPV calculation.

Provide:
- Year-by-year breakdown (Year 1, 2, 3) of labor savings, error reduction, and total costs.
- Payback period (months until cumulative benefit exceeds total cost).
- Year-1 and Year-3 net benefit.
- NPV over 3 years.
- Sensitivity table: Show ROI if adoption is 20% lower, time savings are 25% lower, or error reduction is 50% lower.
- One key assumption to validate before proceeding.

Keep language clear for non-financial stakeholders. Flag any assumptions that feel aggressive.

How to use

1. **Gather data**: Collect current process metrics—team size, hours spent, error rates, and costs. If you don't have exact numbers, estimate conservatively (round down time savings, round up costs). 2. **Research tool costs**: Verify pricing on the vendor site for the AI tool you're considering. 3. **Fill in the prompt**: Paste your data into each placeholder. Use realistic assumptions; aggressive projections often fail. 4. **Run the calculation**: Submit the prompt to Claude or your chosen AI model. 5. **Validate assumptions**: Review the sensitivity analysis. Which assumptions matter most? Test those with your team or a small pilot. 6. **Present to stakeholders**: Use the year-by-year breakdown and payback period to justify the spend. Highlight the key assumption to validate before full rollout.

Customization tips

• **Adjust for your team structure**: If your team is split across multiple locations or cost centers, calculate ROI separately for each and sum them. This helps identify which workflows are most valuable to automate first.
• **Include indirect benefits cautiously**: Faster resolution time may improve customer satisfaction, but don't quantify it unless you have data linking it to retention or revenue. Stick to labor and error savings for credibility.
• **Test accuracy assumptions with a pilot**: Before full rollout, run the AI workflow on 100–500 real examples from your process. Measure actual accuracy, not vendor claims. Adjust the expected_accuracy_percent based on pilot results.
• **Account for change resistance**: If adoption is likely to be slow, lower year_1_adoption_percent and year_2_adoption_percent. A realistic adoption curve often looks like 40% → 70% → 85% over 3 years, not 50% → 90%.
• **Monitor tool cost creep**: AI tool pricing changes. Recalculate ROI quarterly if you're on a usage-based plan (e.g., API calls). Build a 10–20% buffer into annual_tool_cost for price increases.
• **Validate error costs**: The cost_per_error is often underestimated. Include not just rework time but also customer escalations, refunds, and reputation damage. If unsure, ask your support or operations team for historical data.
• **Consider labor redeployment**: If the AI frees up 1 FTE, will you redeploy that person to higher-value work, or reduce headcount? The ROI is the same, but the narrative changes. Be clear with leadership on your plan.