Consult · Executives
How to build an AI business case
An AI business case is a document that justifies an AI project by comparing the expected benefits with the full costs and risks, using a measured baseline. A strong AI business case names the problem, the owner, the measures of success, the running costs after launch, and the pilot result that would justify further investment.
- Reading time
- 8 min
- Chapters
- 6
- Last reviewed
- November 10, 2026
- Author
- Lingows AI Team
- Difficulty
- Intermediate
Before you start
- A candidate process with a named owner
- Access to current performance data or permission to measure the work
- A finance contact who can supply loaded cost rates
The steps at a glance
- Step 01
Define the problem
Describe the process, the people involved, and the outcome that needs to improve.
- Step 02
Measure the baseline
Record the current value of the outcome using existing systems or a time sample.
- Step 03
Estimate benefits as ranges
Translate the expected change into business terms with low, expected, and high cases.
- Step 04
Cost the full lifecycle
Include build, running, people, and vendor costs.
- Step 05
Assess risks and alternatives
List risks with mitigations and compare the AI option with simpler fixes.
- Step 06
Propose a pilot gate
Ask for pilot funding with written success criteria that open the next stage.
Chapter 01
Why does an AI project need a business case?
An AI project needs a business case because AI spending competes with every other investment and carries running costs that continue after launch. A business case forces the team to define the problem, measure the starting point, and agree what result would justify more spending, before money and staff time are committed.
AI projects often start from excitement rather than need. A business case slows the excitement just enough to ask useful questions: which problem, how big, for whom, at what cost, and how anyone will know the project worked.
A business case also protects the project. When budgets tighten, a project with a measured baseline and a documented result survives review. A project justified by enthusiasm alone is easy to cut.
Chapter 02
How do you define the problem and the baseline?
Defining the problem means describing the current process, the people involved, and the specific outcome that needs to improve, such as response time, error rate, or hours spent. The baseline is the measured value of the same outcome today. Without a baseline, no one can show whether the AI project changed anything.
Write the problem in operational terms. A statement such as using AI in customer service is a solution looking for a problem. A statement such as support agents spending a large share of each shift answering the same order status questions is a problem with a measurable outcome.
Measure the baseline before the project starts, using the systems the team already has. A short manual time sample is acceptable when no system tracks the work.
Baseline measures by process type
| Process type | Useful baseline measure | How to collect |
|---|---|---|
| Customer support | Time to first response, time to resolution | Ticketing system reports |
| Document processing | Minutes per document, error rate | Time sample and an audit sample |
| Sales operations | Hours spent on proposals or account research | Staff time log over a short period |
| Internal knowledge | Time spent searching for answers | Short staff survey and spot checks |
| Finance operations | Days to close, exceptions per cycle | Accounting system records |
Chapter 03
How do you estimate the benefits of an AI project?
Estimating the benefits of an AI project means translating the expected change in the baseline into business terms: staff hours redirected, faster response to customers, fewer errors, more capacity without new hires, or lower risk. Use ranges instead of single numbers, and state every assumption so reviewers can challenge the estimate directly.
Keep the arithmetic visible. A reviewer should be able to change one assumption and see the effect. The template below works for most time-based benefits; adapt the units for error reduction or revenue effects.
Hours redirected per year =
(minutes saved per task / 60)
x tasks per year
x share of tasks the system handles
Value of hours redirected =
hours redirected per year x loaded hourly cost
State every input as low, expected, and high.
Record the source of each input.The free ROI calculator applies this kind of arithmetic to a single process and shows the range.
Chapter 04
What costs belong in an AI business case?
An AI business case should include build costs, such as design, integration, and testing, and running costs, such as model usage, hosting, monitoring, maintenance, vendor subscriptions, and staff time for review. Training, change management, and the time of internal experts also belong in the total, because the costs are real even when unbudgeted.
Cost categories
| Category | Examples | When incurred |
|---|---|---|
| Discovery and design | Process mapping, use case scoping | Before build |
| Build and integration | Development, connections to systems, testing | During build |
| Model and platform usage | Model API usage, hosting, vector storage | Ongoing |
| Monitoring and maintenance | Quality review, updates when models or systems change | Ongoing |
| People | Training, review time, change management | Before and after launch |
| Vendor subscriptions | Licensed AI tools and platforms | Ongoing |
Running costs are the line most often missing. Model usage scales with volume, and model providers change prices and retire models on their own schedules. Budget a review of running costs at least once a year. The AI cost estimator gives a starting range, and the answer on AI automation cost explains the main cost drivers.
Chapter 05
How should a business case handle risk and alternatives?
A business case should list the main risks of the AI project, such as poor data quality, low adoption, model errors, and vendor changes, with a mitigation for each. The case should also compare the AI option with alternatives, including improving the process without AI, buying an existing product, or doing nothing.
Comparing alternatives builds credibility. Sometimes a simpler fix, such as a better form, a rule-based automation, or a clearer process, delivers most of the benefit at a fraction of the effort. A business case that shows the alternatives were considered is easier to approve when AI is still the best option.
Adoption risk deserves its own line. A technically successful system that staff avoid produces no benefit. Plan training, involve the team in design, and measure usage from the first day.
Common risks and mitigations
| Risk | Mitigation |
|---|---|
| Poor data quality | Fix the data needed for the pilot first, and test on real samples |
| Low adoption | Involve users in design, train by role, and measure usage weekly |
| Model errors | Keep a person reviewing output, and track error rates |
| Vendor changes | Write the project around outcomes, and avoid lock-in where practical |
| Scope creep | Fix the pilot scope in writing and defer additions to the next stage |
Chapter 06
How do you structure approval around a pilot?
Structuring approval around a pilot means asking leadership to fund a small, time-boxed pilot first, with written success criteria that open the next stage of funding. A pilot gate limits the downside, produces real numbers to replace estimates, and turns the business case into a series of smaller, evidence-based decisions.
Checklist
Business case checklist
After the pilot, update the business case with measured results instead of estimates. The guide on measuring AI adoption covers what to track during the pilot, and the answer on measuring AI ROI covers the return calculation. Lingows AI builds business cases inside AI consulting engagements, alongside the AI roadmap.
01Summary
Key takeaways
What to remember
- Start the business case from an operational problem with a measured baseline.
- Estimate benefits as ranges and show every assumption.
- Include running costs, not only the build.
- Compare the AI option with simpler alternatives.
- Fund a pilot first and tie further funding to measured results.
02FAQ
Questions about this guide
What ROI should an AI project show?
No universal ROI threshold applies to AI projects. The required return depends on the company's other investment options, the size of the project, and the risk involved. Many companies set the bar through the pilot gate instead: the pilot must move the baseline by an agreed amount before further funding is released.
Who should write the AI business case?
The owner of the process being changed should lead the AI business case, with help from finance for costs, IT for technical feasibility, and an executive sponsor for strategic fit. A business case written only by IT or only by a vendor usually misses either the business outcome or the operational detail.
How detailed should the cost estimate be?
The cost estimate should be detailed enough to show every major category, including running costs, with a range for each. Precision matters less than completeness at the business case stage. The pilot then replaces estimates with measured costs, so the next funding decision rests on real numbers instead of assumptions.
What if the benefits are hard to quantify?
When benefits are hard to quantify, describe the benefit specifically, name an observable indicator, and agree how the pilot will judge the indicator. Better customer experience can be tracked through response time, complaint volume, or survey scores. Qualitative benefits still need a clear definition so reviewers can tell whether the benefit appeared.
Should the business case name a specific AI model or vendor?
The business case should describe the solution by capability and outcome, and treat the model or vendor as a choice made during design. Naming a vendor early can anchor the decision before alternatives are compared. Note any existing contracts or platforms, because existing agreements can change both costs and timelines.
Written by
Lingows AI Team
Engineering and strategy team, Lingows AI
The Lingows AI team has built AI systems since 2017 and software since 2012. Collective posts cover model landscape changes, tooling selection, and the direction of applied AI.
Put this guide to work with AI Consulting.
AI consulting turns a guide like this into a scoped plan for your company: where AI fits, where it does not, and what to build first.