There is a four-part business case that separates the AI investments a CFO signs off on from the proposals that quietly die in the finance committee. It is not the deck with the biggest promises. It is the one that survives scrutiny. Here it is.
What is an AI business case, and why does it matter now?
An AI business case is a structured financial argument that connects a specific AI investment to measurable business outcomes, quantified costs, and a defined payback period. It matters now because Gartner forecasts worldwide AI spending will reach US$2.59 trillion in 2026, up 47% year on year, and boards are demanding proof before releasing budget.
The pressure is real. According to Bain, 42% of CFOs plan to increase AI spending by at least 30% over the next two years.
That money will not flow without a case that a finance leader can defend to the board.
Why do CFOs reject most AI investment proposals in 2026?
CFOs reject most AI proposals because the numbers do not hold up. According to MIT's 2025 GenAI Divide report, based on 150 executive interviews and 300 public deployments, 95% of generative AI pilots fail to deliver measurable profit-and-loss impact. A CFO who has seen that statistic treats every new proposal as guilty until proven otherwise.
The credibility gap is the core problem. Deloitte's 2026 State of AI research found that while 66% of organisations report productivity gains, only 20% are growing revenue through AI, and fewer than a third can measure ROI with confidence.
When a proposal claims transformation but offers no measurement plan, the CFO reads it as optimism, not analysis.
What are the four components of an AI business case a CFO will approve?
A CFO-ready AI business case has four components: a defined problem with a baseline number, a quantified benefit tied to that baseline, a full cost model including production and change costs, and a measurement plan with named KPIs. Each component answers a question the CFO will ask before signing.
The four components in practice:
--- The baseline. State today's cost or performance in one number. "Our claims team processes 1,200 cases a month at HK$85 per case."
--- The quantified benefit. Tie the AI outcome directly to that baseline. "Automating triage cuts cost per case to HK$52, saving HK$475,000 a year."
--- The full cost model. Include licences, integration, infrastructure, and change management, not just the software fee.
--- The measurement plan. Name the KPIs, the reporting cadence, and the point at which the project is judged a success or stopped.
How do you quantify AI ROI without overpromising?
You quantify AI ROI by anchoring every benefit to a measured baseline and using conservative ranges rather than single-point promises. A defensible case says "we expect a 30% to 45% reduction in handling time" with the method shown, not "AI will transform our operations." Ranges signal rigour; absolutes signal hype.
Directional honesty builds trust with finance. McKinsey estimates generative AI could add US$2.6 trillion to US$4.4 trillion in value globally, but that figure describes a market, not your P&L.
Translate it down to your own workflow, your own volumes, and your own labour cost. A CFO trusts a small, verifiable number over a large, borrowed one.
How much do enterprise AI projects actually cost to run at scale?
Enterprise AI projects typically cost far more at production scale than in the pilot. MIT's 2025 research found that generative AI deployments commonly run 3 to 5 times initial projections once they reach production volume, driven by inference costs, integration, and support. A business case that ignores this gap loses credibility the moment costs arrive.
The hidden cost drivers deserve explicit lines in the model:
--- Inference at volume. A pilot serving 50 users costs a fraction of one serving 5,000.
--- Integration with legacy systems. Connecting AI to existing ERP or CRM platforms is often the largest single line.
--- Change and training. Adoption work is a cost, not an afterthought.
Naming these upfront is what separates a serious case from a spreadsheet the CFO will not trust.
Which AI use cases deliver the highest ROI by industry?
The highest-ROI AI use cases are concentrated in back-office automation and predictive maintenance. According to Deloitte and McKinsey analysis, financial services back-office automation delivers 3x to 7x returns, the highest of any sector, while manufacturing predictive maintenance delivers 1.5x to 5x returns with a 6 to 18-month payback.
The pattern is instructive for a business case. High-ROI use cases share three traits: a high-volume repetitive task, a clear cost baseline, and a measurable output.
For a Hong Kong logistics firm, that might be automated customs-document processing. For a professional-services group, it might be first-draft report generation. Start where the baseline is already measured, because that is where the CFO can verify your maths.
What should a Hong Kong enterprise measure to prove AI value to the board?
A Hong Kong enterprise should measure a small set of KPIs that map directly to the original baseline: cost per transaction, cycle time, error rate, and adoption rate. According to the Standard Chartered Hong Kong SME index, only 23% of local enterprises have AI deployments with measurable financial impact, so credible measurement is itself a competitive advantage.
The board cares about a specific story: money in, money out, and how fast. Report the baseline, the current figure, and the trend line, refreshed monthly.
McKinsey's early-2026 research shows nearly 70% of Hong Kong white-collar workers now use AI while only 14% of executives report frequent use. That gap means the leaders who can actually quantify impact will stand out sharply in their own boardrooms.
What common mistakes sink an AI business case?
The mistakes that sink an AI business case are predictable: no baseline, benefits stated as absolutes, cost models that stop at the licence fee, and no plan to measure success. Each one gives the CFO a reason to defer. Deloitte found fewer than a third of organisations can measure ROI with confidence, which is exactly the gap a strong case must close.
Three traps appear most often:
--- Technology-first framing. Leading with the model instead of the business problem.
--- Ignoring change cost. Assuming adoption is free, when S&P Global found 42% of companies abandoned most AI initiatives in 2025.
--- No stop condition. A case with no point of failure looks like a blank cheque.
Turning the framework into an approved budget
A CFO does not fund excitement. A CFO funds a clear problem, a conservative benefit, an honest cost, and a way to know if it worked. Build the case in that order and you move from "interesting idea" to "approved line item."
The organisations pulling ahead in 2026 are not the ones spending the most. They are the ones who can prove what they spent actually returned. We understand AI. We understand you. With UD by your side, AI never feels cold.
Build your AI business case with a partner who has done it before
Now that you have the framework, the next step is turning it into numbers your CFO will approve. We'll walk you through every step, from an AI readiness assessment to ROI modelling, cost planning, and board-ready measurement, drawing on 28 years of enterprise experience in Hong Kong.
