The Framework Promise: Why CFOs Reject 60% of AI Proposals on the First Pass
There is a five-component framework that separates AI proposals that get CFO approval from those that get sent back for "more analysis." You will see the framework, the ROI math your CFO actually uses, and the three sentences that decide most enterprise AI investments before the pitch deck reaches slide three.
According to a 2026 KPMG global survey of 1,400 CFOs, 61% of AI investment proposals reviewed in the last year were rejected, deferred, or scoped down on the first review. The reason cited most often was not weak technology — it was incoherent financial framing. The proposal asked for budget without showing how the money would convert into measurable financial outcomes inside the planning horizon the CFO is accountable for.
This article gives you the framework Hong Kong enterprise leaders use to bridge that gap, anchored in real benchmarks from McKinsey, Deloitte and Boston Consulting Group, with the Hong Kong-specific compliance overlay that PCPD and HKMA expect.
What Does a CFO Actually Need to See in an AI Business Case?
A CFO reviewing an AI business case is asking four questions in sequence: What measurable outcome does this produce? Over what time horizon? At what risk-adjusted cost? Compared to which alternative use of the same capital? Anything in your deck that does not directly answer one of these four questions is decoration.
The Boston Consulting Group 2025 Build for the Future study tracked 2,750 executives across 56 countries and found that the 4% of organisations classified as "AI Future-Built" consistently structured business cases around three financial proof points the CFO recognised from existing capital allocation reviews: payback period, net present value, and a documented risk register with quantified worst-case impact.
If your AI proposal does not produce these three artefacts, your CFO does not have the inputs needed to compare your request against the dozen other capital requests sitting on the desk. The proposal is not rejected because the CFO opposes AI. It is rejected because it is incomplete.
How Do You Calculate AI ROI for an Enterprise Use Case?
AI ROI for an enterprise use case is calculated as the net financial benefit divided by total investment over a defined time horizon. The formula is unchanged from any capital project, but the inputs are unfamiliar to most operational leaders, which is where most business cases collapse.
The numerator has three components: direct cost reduction (labour hours redirected, error rates reduced, cycle time compressed), revenue uplift (conversion improvements, retention gains, higher-margin product mix), and cost-of-risk reduction (compliance failures avoided, fraud losses reduced, downtime minutes recovered). The denominator includes licence fees, integration costs, change management spend, ongoing data infrastructure, and the often-forgotten cost of internal time.
According to McKinsey's 2025 State of AI report, the median enterprise AI use case that reached production-scale deployment delivered 15–40% productivity gains in the targeted workflow. However, the same report found that organisations not measuring against a documented baseline reported "positive but unmeasurable" benefits in 47% of cases. Unmeasurable is not approvable.
Your business case must include a baseline measurement protocol: how the metric is calculated today, who owns the data, and what the sample period is. Without this, the post-deployment ROI claim has no defensible reference point and your CFO will discount the projection by the full uncertainty range.
The 5-Component AI Business Case Framework
The framework has five components arranged in the order a CFO reads them. Strategic narrative comes first to establish why this proposal exists. Quantified outcomes come second because numbers without context are noise. Investment profile comes third to show how capital is staged. Risk register comes fourth because no business case clears review without it. Governance and exit conditions close the case by showing how the organisation retains control if assumptions break.
Component 1 — Strategic narrative. Two paragraphs maximum. State the operational problem in financial terms, the strategic alternative considered, and why AI is the chosen path. Reference a peer organisation or industry benchmark to anchor the urgency.
Component 2 — Quantified outcomes. Specify the three to five metrics the project will move, the current baseline value for each, the target value at month 12, and the financial translation. Example: cycle time reduction from 6.4 days to 2.1 days equals HK$3.8 million in working capital release.
Component 3 — Investment profile. A staged budget showing pilot phase, scaled deployment, and steady-state operating cost. Include licence fees, integration time, training cost and the internal time tax. Hong Kong enterprise pilots in the 2025 Deloitte AI survey averaged HK$420,000 to HK$2.1 million for first deployment, depending on data complexity.
Component 4 — Risk register. Five to seven risks, each with likelihood, financial impact, mitigation owner and a tripwire metric that triggers escalation. Include data privacy, model accuracy, vendor lock-in, change resistance and regulatory exposure.
Component 5 — Governance and exit conditions. Specify the steering committee, the monthly metrics review cadence, the conditions under which the project pauses, and the salvage value if the project is terminated mid-cycle.
How Do You Pick the Right First Pilot to Strengthen the Business Case?
The right first pilot has three properties: it produces measurable financial outcomes within 90 days, it touches a workflow with clean baseline data, and the team owning the workflow already wants the change. Pilots that violate any one of these conditions become exhibits in the case against future AI investment.
According to RAND's 2024 study of 65 enterprise AI deployments, 80.3% of failed projects had at least one of three root causes: misalignment with business priorities, poor data infrastructure, or unrealistic expectations of model performance. All three are visible at the pilot selection stage if the selection process is rigorous.
A practical filter: rank candidate pilots by the product of expected annual value, baseline data quality score (1–5), and team adoption readiness score (1–5). The pilot with the highest product wins. This forces the conversation away from technology novelty and toward financial discipline, which is exactly the conversation the CFO wants to have.
What Risks Should the Business Case Address Before the CFO Asks?
The business case must surface five risk categories that Hong Kong enterprise CFOs have learned to ask about: data privacy and PDPO compliance, model accuracy and hallucination, vendor concentration risk, change management cost overruns, and regulatory shift exposure. Each needs a quantified worst-case and a named owner.
The Hong Kong Privacy Commissioner's 2024 AI guidance requires impact assessment, transparency disclosure to data subjects, and human review for AI-driven decisions affecting individuals. The HKMA's generative AI principles published the same year specify board-level accountability for AI risk in financial institutions. Your business case must reference both if you operate in financial services or handle personal data at scale.
Vendor concentration is the risk most often missed. According to a Deloitte 2026 survey of 850 enterprise AI buyers, 43% had no documented exit plan from their primary AI vendor. The CFO will ask about this because procurement law requires it. Your answer should specify data portability rights, model output ownership, contractual exit terms, and the cost of switching vendors at month 24.
Common Mistakes Enterprise Leaders Make Pitching AI to the CFO
The most common mistake is leading with technology capability instead of financial outcome. The CFO does not need to know the model is multimodal. The CFO needs to know that processing 12,000 supplier invoices per month at 87% straight-through rate releases 3.2 FTE of accounts payable capacity, which the workforce plan can either redeploy or convert to cost savings.
The second most common mistake is presenting AI investment as a binary go or no-go decision. CFOs prefer staged commitments. Structure the proposal as a HK$300,000 pilot with three success criteria, then a HK$1.5 million scale-up if criteria are met, then ongoing operating budget. This matches how every other capital project moves through the organisation and removes the artificial friction created by asking for the full budget upfront.
The third common mistake is missing the comparison case. CFOs allocate capital relatively, not absolutely. If the AI proposal does not show the implicit cost of doing nothing — the lost productivity, the competitive gap, the compliance exposure — the CFO has no anchor for the upside calculation. Include a 12-month and 36-month do-nothing scenario alongside the proposed investment.
How Do You Present the Business Case in a 12-Minute Board Meeting?
You present the business case in the order the CFO reads it: outcome, investment, risk, governance, ask. The whole presentation is 12 minutes maximum because that is the typical attention budget in a Hong Kong board meeting before discussion takes over. Three minutes per major section, with the financial slide visible the entire time.
Open with the strategic narrative in two sentences and one peer benchmark. Move directly to the three quantified outcomes — current baseline, target at month 12, financial translation. Show the staged investment profile with pilot, scale and steady state. Surface the top three risks with mitigation owners. Close with the specific ask: pilot budget, decision deadline, executive sponsor required.
The Harvard Business Review's 2025 study of 312 enterprise capital approval sessions found that proposals presented in this five-step structure had a 2.4x higher first-pass approval rate than proposals organised around technology architecture or vendor capability. The structure is doing the work, not the slides.
Conclusion: From Framework to Approved Investment
An AI business case the CFO approves is not a better-written deck. It is a discipline. The framework forces the proposing team to translate technology ambition into financial language, to document the baseline before the project starts, and to surface risks before the CFO has to. Organisations that adopt this discipline get more AI proposals approved, deploy faster, and produce the post-implementation evidence that earns budget for the next round.
The five-component framework is portable across industries, scalable across project sizes, and aligned with the capital allocation processes Hong Kong enterprises already use. It does not require new tooling, new vendors or new headcount. It requires the proposing team to think one level higher about why the investment exists, what it will measurably do, and how the organisation will know.
That work is unglamorous. It is also the difference between an AI strategy that compounds quarter after quarter, and one that produces a series of expensive proofs of concept nobody can defend. UD has spent 28 years guiding Hong Kong enterprises through technology cycles, helping leaders connect tools to outcomes the board can see. 懂AI，更懂你 — UD相伴，AI不冷。
Build Your AI Business Case With UD
You have the framework. The next step is identifying the right first use case for your organisation, baselining the metrics, and producing the board-ready document. UD's enterprise advisory team will walk you through every step — from AI readiness assessment, financial baseline definition and pilot scoping, to vendor evaluation and post-deployment KPI tracking. 28 years of Hong Kong enterprise experience, applied to your AI business case.
