AI investment is growing rapidly, but the majority of AI projects still fail to deliver measurable business results. The problem is rarely the technology.
McKinsey’s State of AI 2025 found that 78% of organisations are now using AI in at least one business function – up from 55% in 2023. But only approximately 6% qualify as high performers, defined as organisations that attribute 5% or more of their EBIT to AI. The gap between the majority of AI adopters and the high-performing minority comes down to one thing: the quality of the business case built before the first deployment.
This guide takes you through how to build an AI business case that survives leadership scrutiny, delivers on its projections, and creates the foundation for expanding AI investment across the business.
Why Most AI Business Cases Fail
The most common failure is vague ambition. A business case that says ‘we want to use AI to improve our operations’ is not a business case – it is a wish. It gives a leadership team nothing to approve because it has no defined scope, no measurable outcome, and no accountability structure.
The second failure is technology-first thinking. Businesses that start with a tool (‘we want to implement GPT-4’) and then look for a use case build backwards. The tool then shapes what they measure, rather than the business problem shaping what they build.
McKinsey’s research is clear on this point: 60% of organisations that have deployed AI haven’t seen enterprise-wide EBIT impact. The biggest differentiator between those that do and those that don’t is whether they started with a clearly defined business problem and built the technology solution around it.
Step 1: Define the Problem in Financial Terms
Every strong AI business case begins with a problem that has a cost. Not a challenge, not an opportunity area, not a capability gap – a cost that shows up on the P&L.
Examples:
– Our customer support function costs £420,000 per year. We handle 3,200 tickets per month. 65% of those tickets are responding to queries our website FAQ should answer. That is 2,080 tickets per month at a cost of approximately £11 per ticket – £23,000 per month of avoidable cost.
– Our sales team of 12 representatives is currently closing 22% of qualified leads. Industry benchmark is 28%. Each percentage point of close rate improvement at our average deal size of £18,000 represents approximately £190,000 in additional annual revenue.
– Our finance team spends approximately 14 days per quarter on manual reconciliation. At fully loaded cost of £65 per hour for finance staff, that is approximately £29,000 per quarter – £116,000 per year – in labour time on a process that could be substantially automated.
Specificity at this stage is what makes the rest of the business case defensible. If you know the cost of the problem, you can calculate what it’s worth to solve it.
Step 2: Identify the AI Solution with Evidence
Once the problem is defined in financial terms, identify the AI approach that addresses it and assemble external evidence that it works.
For each proposed solution, document:
What the AI actually does (in plain language, not technical terms)
What similar organisations have achieved with it – and from what sources
What the implementation requirements are: data, integration, change management
What the limitations and risks are
On the evidence point, use primary-source data where possible. Gartner’s April 2026 research, surveying 314 organisations on data from September and October 2025, found that CFOs implementing strategic AI programmes are on track to add 10 margin points by 2029. Gartner’s February 2026 survey of more than 300 CFOs found that 60% are increasing AI investment by 10% or more and 75% are raising their technology budgets – not because they are speculating on future returns, but because they are seeing early results.
McKinsey’s research on enterprise AI deployments reports a median ROI of 210% over three years, with an average payback period of approximately 16 months. These are the benchmarks that a well-built AI business case should aspire to – and explain to leadership why it is credible to project them.
Step 3: Build the ROI Model
The ROI model should have three components:
Investment: Include all costs, not just licence fees. Integration work, data preparation, staff training, project management time, and ongoing maintenance all belong in the investment figure. Underestimating costs is the most common reason AI ROI projections are later challenged.
Returns: Map returns to the specific metrics you defined in Step 1. If the problem was support ticket cost, the return is measured in cost per ticket. If the problem was close rate, the return is measured in additional revenue from improved close rate. Do not mix financial and non-financial returns – keep the ROI model purely financial.
Timeline: Be honest about when returns appear. McKinsey’s 16-month average payback means most AI investments do not break even in the first year. A leadership team that expects positive ROI in month six and sees a flat month nine will question the entire programme. Set realistic expectations upfront.
Step 4: Manage the Risks
Every AI business case needs a risk section. The risks leadership will ask about are:
Data quality: Does the organisation have the data the AI needs to work? AI quality is a direct function of data quality. If your CRM is incomplete or your transaction records are inconsistent, the AI model will be unreliable.
Change management: Will staff adopt the AI? The McKinsey research that shows high performers are nearly 3 times more likely to redesign workflows before deploying AI is significant precisely because workflow redesign is a change management challenge, not a technology challenge.
Vendor lock-in: What happens if the chosen AI vendor raises prices or changes their terms? A business case should include a mitigation for this risk – contractual protections, data portability requirements, or a strategy for building on open-source models where appropriate.
Step 5: Present It Properly
A well-constructed AI business case presents in five minutes. If your leadership team needs longer than that to understand what you’re proposing, why it works, what it costs, and what it returns, the case is not yet ready.
The presentation structure that works:
1. The problem in financial terms (one slide, one number)
2. The proposed solution with external evidence (one slide)
3. The investment required (one slide – complete, not just licences)
4. The projected ROI with a clear payback timeline (one slide – conservative scenario and base case)
5. The risks and mitigations (one slide)
6. The approval request: what you need, by when, to begin
McKinsey’s analysis of AI high performers finds that these organisations are 5 times more likely to make a significant financial bet on AI than the average organisation. The reason is not that they are more risk-tolerant – it is that they build the business case rigorously enough that the investment decision becomes straightforward.
Frequently Asked Questions
How detailed does an AI business case need to be for a small business?
For an SMB with AI investment below £100,000, a four-to-six-page document covering the five elements above is typically sufficient. The detail should be proportionate to the investment: a £15,000 deployment needs a clear ROI model but not a 40-page strategy document. A £200,000 multi-year programme warrants more rigour.
What if I don’t have the internal data to build a credible ROI model?
Start with industry benchmarks and work backwards. McKinsey’s 20–30% cost savings in automated functions gives you a base case for a cost-reduction project. Gartner’s finding that organisations with AI next best actions are 2.6 times more likely to achieve commercial growth gives you a framework for a revenue case. Use these as your conservative scenario, then build your specific numbers around your business context.
What makes the difference between AI projects that succeed and those that don’t?
McKinsey’s research identifies two consistent differentiators: organisations that redesign their workflows around AI before deploying it, and organisations that measure AI impact on business outcomes – not just technology adoption rates – from the first week of operation. Both of these are discipline choices, not technology choices. The technology itself rarely determines whether an AI project succeeds.
Ready to Build Your AI Strategy?
DoSystems helps SMBs design and implement AI systems that connect directly to business outcomes – cost savings, revenue growth, and competitive advantage. Contact us at DoSystemsInc.com to start with a free AI readiness assessment.
Comments are closed