AI Business Case and ROI: Costs, Savings and Payback Your Board Can Check
An AI business case states what an AI project costs, what it should save or earn, its risks and payback period. Remolda builds it from your processes and data, starting with the two-week AI Opportunity Audit ($2,900 CAD + HST).
In short
- A usable AI business case has four parts: full cost (build, licences, usage, data work, training, running), measurable benefits, risks with owners, and a payback period with the assumptions written out.
- Numbers come from your own workflows: time per task, volumes, error rates and loaded staff cost, collected in interviews and from system data.
- Canadian lines that are easy to miss: OpenAI and Anthropic price API usage in US dollars, Québec's Law 25 requires a privacy impact assessment for new information systems that handle personal information, and GST/HST (and QST in Québec) apply to services.
- The two-week AI Opportunity Audit ($2,900 CAD + HST) produces the inputs: 10+ use cases ranked with ROI ranges and risk notes, and the budget for the first pilot. A detailed case for one project is quoted on the call.
- After launch, the six-week AI Pilot Sprint ($9,800) measures results against the plan and ends with a go, adjust or stop decision.
Your situation
A business case is needed when someone has to sign for the money. Typical moments:
- The board wants numbers before approving an AI budget. They ask for cost, benefit and payback for each project.
- Finance pushed back on a vendor's ROI slide. The figures came from the vendor's averages, and nobody checked them against your volumes.
- A funder or lender asks for a plan. IRAP advisors, BDC and Investissement Québec want scope, budget and expected results in writing.
- A pilot worked and now needs a scale decision. You need to know what the next step costs and what it should return.
Statistics Canada figures cited in the AI for All strategy show that 12% of Canadian businesses used AI to produce goods or services between mid-2024 and mid-2025.
What goes into an AI business case
| Part | What it contains |
|---|---|
| One-time cost | Discovery, design, build and integration, data cleanup, security and privacy review, staff training |
| Running cost | Licences per seat, model usage per token or per call, hosting, monitoring, maintenance, human review time |
| Benefits | Hours saved on named tasks times loaded staff cost, shorter turnaround, fewer errors and rework, revenue that was being missed |
| Risks | Data quality, staff adoption, vendor terms and price changes, privacy obligations, what happens when the AI is wrong |
| Payback | One-time cost divided by net yearly benefit, with cautious and optimistic cases |
| Measurement | The two or three numbers the pilot will be judged on, and their current baseline |
Canadian lines that change the numbers
US-dollar billing. Model usage from OpenAI and Anthropic is priced in US dollars per million tokens. A CAD budget needs an exchange-rate margin.
Privacy work. Québec's Law 25 (s. 3.3) requires a privacy impact assessment for any project to acquire, develop or overhaul an information system that handles personal information. It is a cost line, and it shapes the timeline.
Tax. Consulting and software services carry GST/HST, and QST in Québec, according to the place of supply; the business case shows amounts before and after tax.
Funding. Where a program such as IRAP, SR&ED, BDC LIFT or ESSOR fits, the business case shows the project with and without it. See AI funding in Canada.
How the numbers are calculated
Each use case goes through the same five steps, so finance can rerun the numbers:
- Baseline. Time per task, monthly volume and error rate, from system data or sampled with the people who do the work.
- Share handled by AI. What part of the task the AI does, and how much review time a person still spends.
- Yearly benefit. Hours saved times loaded hourly cost, plus measurable gains in errors or turnaround.
- Full cost. One-time and yearly lines from the table above.
- Payback and range. Payback in months under cautious and optimistic assumptions.
The formulas and a worked example are in our guide how to calculate ROI on AI projects. For a first estimate from your own inputs, use the AI ROI calculator.
What Remolda delivers
The inputs for the business case come from the AI Opportunity Audit:
- Interviews with 6–10 people and a walk-through of the real workflows, to get volumes and time per task.
- A ranked map of 10+ use cases with ROI ranges and risk notes.
- A data and systems check showing what blocks each use case.
- A 12-month roadmap with the recommended first pilot and its budget.
When the pilot runs, the AI Pilot Sprint agrees the measurement plan up front and ends at week 6 with a results review and a go, adjust or stop decision.
What it costs
| Package | Price (CAD + HST) | Time | Business case output |
|---|---|---|---|
| AI Readiness Review | $490 | 1 week | 3–5 candidate use cases with rough effort |
| AI Opportunity Audit | $2,900 | 2 weeks | ROI ranges for 10+ use cases, first pilot budget |
| AI Pilot Sprint | $9,800 | 6 weeks | Measured results against the plan |
What you pay for a smaller package is credited toward a larger one. Details are on the pricing page.
Why Remolda for an AI business case
- Your numbers. Volumes and times come from your workflows and systems, and every assumption is written down.
- Vendor-neutral. Tools are compared on your tasks, your data rules and your costs. See AI vendor selection.
- Canadian cost lines included. Currency, privacy assessments, tax and funding programs are in the model.
- Published prices. The cost of building the case is known before the first call.
- Measured after launch. The pilot is judged on the same numbers the business case promised.
How the work runs
- Call. 30 minutes: the decision to be made and who signs it.
- Audit. Two weeks of interviews, workflow walk-throughs and data checks.
- Business case. Costs, benefits, risks and payback per use case, with the recommended first pilot.
- Decision. Leadership and finance review the assumptions; a detailed case for one project is quoted on the call if needed.
- Pilot. Six weeks, measured against the plan, then a go, adjust or stop decision.
Frequently asked questions
How does the business case work out payback?
It compares the yearly benefit (hours saved times loaded cost, plus measurable gains such as fewer errors or faster turnaround) with the full cost: one-time build and data work, and yearly licences, usage and maintenance. Payback is the one-time cost divided by the net yearly benefit. Every input is written down so finance can check it; the formulas are in our guide on calculating ROI on AI projects.
What costs do AI business cases usually miss?
Data cleanup, integration with existing systems, staff training, time spent reviewing AI output, monitoring after launch, per-use model charges that grow with volume, currency exposure when the vendor bills in US dollars, and privacy work such as a Québec privacy impact assessment.
How long does an AI project take to pay back?
It depends on volume and on how much of the task the AI can handle safely. High-volume, repetitive work such as email triage or document extraction tends to pay back faster than low-volume expert work. The business case shows the payback for your numbers, with a range for optimistic and cautious assumptions.
What if the numbers do not work?
Then the business case says so and names what would change the answer, such as a larger volume, a cheaper model tier or a different use case. Stopping before the build is a valid result and costs far less than a failed pilot.
Can the business case support a grant or loan application?
Yes. Funders and lenders such as NRC IRAP advisors, BDC and Investissement Québec ask for scope, budget and expected results. The Opportunity Audit gives you those in writing; see the AI funding page for which programs fit.
How is this different from your free ROI calculator?
The ROI calculator gives a first estimate from a few inputs you type in. The business case is built from interviews, real volumes and system data, covers risks and full costs, and is written for a board, CFO or funder to review.
Do you guarantee the ROI?
Results depend on your data, adoption by staff and vendor pricing, which we do not control. We make the assumptions explicit, measure the pilot against them and recommend stopping or adjusting when results fall short.
Sources
- ISED — Canada’s National Artificial Intelligence Strategy: AI for All (adoption figures)
- LégisQuébec — Act respecting the protection of personal information in the private sector (CQLR c P-39.1), s. 3.3
- OpenAI — API pricing (USD per million tokens)
- Anthropic — Claude API pricing (USD per million tokens)
- Office of the Privacy Commissioner of Canada — AI, privacy, and your business
Facts checked:
Related services
Approach phases
Related insights
Talk to an AI transformation consultant
A 30-minute call: you describe the situation, we tell you what to do first and what it would cost.
Book a 30-min call30 minutes. English or French.