AI Project Business Case Template
A section-by-section template for building an AI automation business case that wins approval and holds up under scrutiny.
Why This Matters
The most common reason AI automation projects fail to get approved is not the idea — it is the business case. Reviewers want to see: what problem does this solve, what will it cost (fully loaded), what will it return (conservatively modelled), what are the risks, and how will we know if it works. This template structures all of that in the order reviewers expect to see it.
Section 1: Executive Summary
A one-page summary of the entire business case. Written last, placed first. Should be self-contained — a reviewer who only reads this page should understand the investment being requested.
Prompts for this section
- What is the specific problem being solved? (1-2 sentences)
- What is the proposed AI solution? (1-2 sentences, non-technical)
- What will it cost? (total investment over 18 months)
- What will it return? (conservative scenario metric improvement)
- What is the payback period?
- What is being requested from the approver?
Example language
“We currently process 800 supplier invoices per month manually, taking an average of 3.2 days per invoice and producing an 8% error rate that generates £18,000 per month in rework costs. This project proposes an AI-automated invoice processing system that will reduce processing time to under 4 hours and reduce the error rate to below 2%. Total 18-month investment: £95,000. Conservative annual saving: £140,000. Payback period: 8 months. We are requesting approval of the Phase 1 budget of £45,000 for system build and pilot.”
Section 2: Problem Statement and Current State
A precise description of the problem being solved, backed by measured data. This section should quantify the cost of the current state — making the case that the problem is worth solving before the solution is introduced.
Prompts for this section
- What is the process being automated, and what is its current volume? (units per day/week/month)
- What is the current processing time per unit? (measure it — do not estimate)
- What is the current cost per unit? (staff time × loaded hourly rate)
- What is the current error rate, and what does each error cost to remediate?
- What is the total monthly/annual cost of the current state? (volume × cost per unit + error remediation)
- What are the secondary costs not captured in the above? (staff frustration, opportunity cost, customer experience)
- What does the trend look like — is volume growing, and is the problem getting worse over time?
Section 3: Proposed Solution
A non-technical description of what the AI system will do, how it will work, and what the implementation looks like. Avoid technical jargon — the goal is for any reviewer to understand what is being built.
Prompts for this section
- What will the AI system do, in plain language? (one paragraph)
- What does the workflow look like after automation? (before → after comparison)
- What will humans still do? (address the human oversight question directly)
- What systems will the AI connect to?
- What are the major implementation phases, and what does each produce?
- What does the pilot look like, and how will it validate the solution before full deployment?
Section 4: Financial Model
The quantified cost and benefit analysis. Build two scenarios: conservative (the case you are committing to) and realistic (the expected case). Never present optimistic as the primary scenario.
Prompts for this section
- Build costs: engineering time (hours × day rate), design and architecture, testing and quality assurance, data preparation, integration work
- Deployment costs: infrastructure setup, vendor onboarding, security review, training
- Year-1 operation: monitoring, maintenance, model evaluation, user support, bug fixes
- Ongoing annual costs: infrastructure, vendor licence, maintenance engineering time, annual model updates
- Conservative benefit: what metric improvement can you commit to with high confidence, at what adoption rate?
- Realistic benefit: what is the expected improvement, and what does that represent in £/year saved or revenue generated?
- Payback period calculation: total investment ÷ annual net benefit (conservative scenario)
Section 5: Risks and Mitigations
An honest assessment of the key risks, their likelihood, their potential impact, and the mitigation plan for each. Reviewers who see risks hidden or minimised lose confidence — reviewers who see risks clearly articulated with credible mitigations gain confidence.
Prompts for this section
- Data quality risk: what data quality issues have been identified, and what is the plan to address them?
- Technical risk: what are the most uncertain technical components, and how will they be validated?
- Adoption risk: what if the team resists using the system or works around it?
- Model performance risk: what is the plan if the AI does not reach the target accuracy?
- Vendor risk: what happens if the AI vendor changes pricing or deprecates the API the system depends on?
- Timeline risk: what are the most likely causes of delay, and what is the contingency?
Section 6: Success Criteria and Evaluation Framework
The explicit criteria by which the investment will be evaluated at each milestone. Defined before approval, not after the results are in.
Prompts for this section
- Pilot success criteria (at 8-12 weeks): what metric thresholds must be met to proceed to full deployment?
- Phase 1 success criteria (at 6 months): what does the system need to achieve to be considered successful?
- Full deployment success criteria (at 12 months): what is the standard the investment will be evaluated against?
- What are the failure thresholds — below what performance level would you recommend discontinuation?
- Who will evaluate success, how, and when?
Section 7: The Ask
What is being requested from the approver, specifically. Be explicit: amount, authorisation for what phase, by when, and what the next decision point is.
Prompts for this section
- What budget amount is being requested, and for which phase?
- What authority or resources (beyond budget) are required to proceed?
- What is the timeline to the next review or decision point?
- What decision will the approver need to make at that next review point?
AI Systems Architect
Want help applying this to your business?
A strategy call is where the framework meets your specific situation, team, and goals.