Most AI projects inside infrastructure firms do not die from bad technology. They die in the budget meeting, when a champion who is excited about a pilot cannot answer a straightforward question from the finance director: what does this actually return, and by when. Global research backs this up. In its November 2025 Global Survey, McKinsey found that although 88 percent of organizations report regular AI use, only 39 percent see any EBIT impact at the enterprise level, and nearly two thirds have not advanced beyond experimentation or early pilots, according to McKinsey's State of AI report. A good idea without a defensible business case rarely survives contact with a capital allocation committee.
The UAE context makes this both easier and harder. Easier, because the macro tailwind is real. Artificial intelligence is projected to add more than USD 96 billion to the UAE's GDP by 2031, with AI expected to account for 13.6 percent of national GDP, the highest share among GCC countries, according to Emirates NBD Research's June 2025 analysis. Harder, because that same optimism means capital committees have already heard several AI pitches this year, and more than one of them likely stalled after approval. Firms that have already worked through an AI implementation roadmap or completed a readiness assessment still need one more document before funding follows: a business case built to survive real scrutiny, not just an enthusiastic pilot demo.
Why a Formal Business Case Matters More in Infrastructure
A marketing team piloting a chatbot can absorb a failed experiment quietly and move on. An infrastructure firm cannot do the same with a predictive maintenance system tied to a decade-long asset lifecycle, or an AI scheduling tool integrated into a live construction program. Capital committees in this sector are used to evaluating investments against depreciation schedules, safety cases, and multi-year operating budgets. An AI proposal that shows up without that same rigor gets treated as a discretionary line item, and discretionary line items are the first things cut when a budget tightens.
Analysts have grown skeptical of AI proposals for good reason. Gartner predicts that more than 40 percent of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls, as Forbes reported in July 2026. That statistic should not discourage infrastructure firms from pursuing AI. It should discourage them from pitching it the way most failed projects were pitched: with a vague benefit statement, a vendor's marketing deck standing in for real analysis, and no named owner accountable for results.
Start With the Real Cost Baseline, Not Just Licensing
The single most common flaw in an AI business case is treating the vendor's license fee as the total cost. It rarely is. A realistic cost baseline for an infrastructure AI deployment includes data preparation and cleansing, integration with existing ERP, SCADA, or asset management systems, staff time for training and retraining, ongoing model monitoring, and the compliance overhead of meeting UAE data residency and sector-specific rules. Firms that have already gone through a structured provider evaluation will recognize this list. It is the same three-year total cost of ownership model that should inform vendor selection, and it belongs in the business case just as much as in the procurement scorecard.
Compliance costs deserve their own line item rather than a footnote. Depending on the sector and the data involved, a project may need to budget for legal review, data processing agreements, and in some cases infrastructure changes to keep sensitive data onshore. Our guide to AI compliance obligations for UAE infrastructure firms sets out which rules actually apply and which widely repeated compliance claims do not hold up, which is useful reading before finance asks what happens if a regulator changes the rules mid-project.
Build the Benefit Side With Numbers a Finance Director Will Trust
Once the cost side is honest, the benefit side needs the same treatment. This is where UAE-specific data helps, because the regional picture is more encouraging than the global one. The KPMG UAE Tech Report 2026 found that 97 percent of UAE organizations have embedded AI agents into their workflows, well above the 87 percent global average, and that 60 percent of UAE organizations expect AI deployed at enterprise scale to deliver measurable returns within twelve months, according to reporting on the KPMG findings. Notably, zero respondents in that survey reported negative value realization from their technology investments, and half of UAE organizations now invest between USD 50 million and USD 99.9 million annually in digital technology, a scale that puts serious weight behind AI as a board-level priority rather than an experiment.
That regional confidence is a useful data point to cite, but it is not a substitute for a firm's own numbers. A business case should translate expected benefits into the same units the finance team already tracks: hours of downtime avoided per quarter, crew hours saved per project phase, or a percentage reduction in rework on a specific asset class. Where a firm does not yet have its own baseline, a pilot's job is to generate one, not to prove the technology works in the abstract. The PwC AI Jobs Barometer for the UAE shows the country's AI talent market accelerating quickly, which is a reminder that workforce capacity to actually realize projected benefits is itself a cost and a constraint that belongs in the model, not an assumption that skilled staff will simply appear.
Three ROI Models Infrastructure Firms Actually Use
Different types of AI investment call for different ways of quantifying return. Trying to force every project into a single ROI formula is a common reason business cases feel unconvincing to people who evaluate capital projects for a living.
- Cost avoidance model: used for predictive maintenance and safety monitoring, where the benefit is downtime, failures, or incidents that did not happen. Requires a credible baseline failure rate and a documented cost per incident to be persuasive.
- Productivity and throughput model: used for scheduling, document processing, and design optimization tools, where the benefit is measured in hours saved, units processed per crew, or project phases completed on time. This model is the easiest to validate with a short pilot.
- Risk reduction and compliance model: used for AI tied to regulatory reporting, safety case management, or data governance, where the benefit is avoided penalties, reduced insurance premiums, or a lower probability of a costly compliance failure. This model is the hardest to quantify precisely, so it should be presented as a range with clearly stated assumptions rather than a single number.
A Practical Five-Step Business Case Template
Step 1: Establish the Current-State Baseline
Before proposing anything, document what the current process actually costs today: labor hours, error rates, downtime, or compliance exposure, measured over at least one full operating cycle. Without this, any claimed improvement is unverifiable.
Step 2: Scope a Pilot With Predefined Success Metrics
Agree on the specific metrics that will define success before the pilot starts, not after the results come in. A pilot without predefined thresholds tends to be interpreted generously by whoever is presenting it, which undermines credibility with a skeptical finance committee.
Step 3: Model Three Scenarios, Not One Number
Present a conservative, base, and upside case for both cost and benefit. A single number invites a committee to challenge it as either too optimistic or arbitrarily chosen. A range, with the assumptions behind each scenario stated plainly, signals that the analysis has actually grappled with uncertainty.
Step 4: Price In Governance and Change Management
Include the cost of the people side of the rollout: training time, an internal champion's partial time allocation, and a documented change management plan. Projects that treat adoption as free almost always underdeliver against the business case, because the technology working in a lab is a different problem from staff actually using it in daily operations.
Step 5: Ask for Phased Funding, Not One Large Number
Structure the funding request around gated milestones tied to the pilot's predefined metrics, rather than a single upfront commitment for full-scale rollout. This gives the capital committee a lower-risk decision to approve now, with a clear, evidence-based checkpoint before the next tranche of investment.
Common Mistakes That Sink an AI Business Case
- Presenting a single ROI figure with no sensitivity analysis or downside scenario
- Using the vendor's marketing statistics instead of a firm's own pilot data or a documented baseline
- Leaving out integration, retraining, and compliance costs that surface only after signature
- Failing to name a specific business owner accountable for realizing the projected benefit
- Asking for full-scale budget before a scoped pilot has produced any real evidence
Conclusion
An AI business case is not a formality to clear before the real work starts. It is the document that determines whether a project gets the sustained funding, staff time, and organizational attention it needs to actually deliver, or whether it becomes another pilot that quietly disappears from next year's budget. UAE infrastructure firms have a genuine tailwind that firms in many other markets do not: strong national investment, a regional AI talent pipeline, and boardrooms that are, on balance, more receptive to AI proposals than the global average. That tailwind is an argument for building the case properly, with an honest cost baseline, benefits stated in the finance team's own units, and a phased funding request tied to evidence, not for skipping the exercise because the macro story already sounds convincing enough on its own.