Tags
Share
“The Answer … is … Forty-two.”
In The Hitchhiker’s Guide to the Galaxy, Deep Thought, the most powerful computer ever built, takes 7.5 million years to calculate the Ultimate Answer to Life, the Universe, and Everything. When the answer finally arrives, to the sages' disappointment, it is “42.” Only then does everyone realize that the investment might have benefited from a more grounded and outcome-focused question in the first place.
Some enterprise AI business cases have the same problem. They do not take quite as long, thankfully, but the investment often begins with ‘AI for AI’s sake’. Organizations commit to platforms, pilots, and models before clearly defining the tangible business outcomes they want to achieve, or how those outcomes will translate into real value.
The board does not fund AI because it is impressive. It wants a credible change in business performance. It's vital to get your questions right if you want your supercomputer to yield sensible, actionable results.
McKinsey’s 2025 global survey found that 88% of organizations use AI regularly in at least one business function, but only 39% report any enterprise-level EBIT impact. Adoption is spreading faster than value.
A strong enterprise AI business case closes that gap. It begins with the right question, establishes a measurable baseline, models the full investment, tests the return under different scenarios, and makes the risks visible. It translates technical possibility into the language of capital allocation.
Why AI Business Cases Fail to Get Approved
A compelling demonstration of AI engineering can impress and win attention. It cannot, by itself, win an investment decision.
Many proposals begin with what a model can generate, automate, predict, or optimize. The board starts elsewhere: Which business problem is being solved? Why does it matter now? What is the total commitment? How will the organization recognize value when it appears?
Weak cases describe benefits without establishing what the current process costs. They turn hours saved into cash returns without explaining how that capacity will be used. They include model and platform costs but omit data preparation, integration, governance, adoption, and ongoing operation. They present one confident ROI figure even though adoption, volume, accuracy, and implementation time remain uncertain.
A benefit without a baseline is not a return. It is an aspiration.
BCG’s 2025 research found that only 5% of companies were generating AI value at scale, while 60% reported little or no material value despite substantial investment. The lesson is not that AI lacks value. It is that investment alone does not create it.
This is also why AI programs fail to deliver ROI even when pilots work. A pilot may prove the technology can perform a task. The business case must show that the organization can integrate that capability into a workflow, support it, govern it, drive adoption, and convert the result into economic value.
The Five Components of a Credible AI Business Case
A useful AI business case template should create an AI investment decision framework: a transparent model of the problem, intervention, assumptions, risks, and evidence required to continue funding.
1. Establish the Current-State Baseline
Before estimating AI value, quantify what happens today.
Depending on the use case, that might include cost per transaction, processing time, error and rework rates, employee hours, customer abandonment, conversion, risk losses, vendor spend, or revenue delayed.
The baseline is the denominator of every ROI claim. Without it, a 20% improvement has no reliable economic meaning.
Teams must also distinguish activity from value. If AI saves 20,000 employee-hours, what happens to those employees? Do they reduce overtime, increase throughput, improve service, shorten time to market, or release people for higher-value work?
Time saved is not automatically money saved. The business case has to explain the conversion mechanism.
2. Prioritize Use Cases by Value, Feasibility, and Risk
An organization may identify dozens of plausible AI opportunities. They do not all belong in the first investment case.
Assess each use case against three tests:
- Value: What measurable outcome could change, and how material would it be?
- Feasibility: Are the data, technology, workflow, integration, and skills sufficiently ready?
- Risk: What regulatory, operational, security, reputational, or adoption risks accompany it?
A high-value use case with poor data readiness may need foundation work first. A modest automation opportunity may deliver quick payback but little strategic advantage. A more ambitious product or decision-support use case may justify investment if it strengthens revenue, differentiation, or retention.
The goal is not to select the most futuristic idea. It is to identify the best risk-adjusted route to value.
3. Model the Full Investment
AI costs extend far beyond the model. A credible investment model should include:
- data engineering and quality improvement
- model, platform, license, and vendor costs
- cloud, compute, inference, and token consumption
- application and workflow integration
- security, privacy, validation, and governance
- testing, human oversight, training, and adoption
- monitoring, maintenance, and incident response
- internal leadership and subject-matter time
This is where Data Engineering & Analytics become part of the financial argument, not technical footnotes. Production value depends on the system around the model.
Deloitte notes that AI spending is volatile and nonlinear: workload type, model complexity, infrastructure intensity, and token use can all affect cost. A small proof of concept may have very different unit economics from an enterprise deployment handling millions of interactions.
A serious AI ROI model therefore needs assumptions about volume, model choice, utilization, and future consumption—not only an initial implementation estimate.
4. Project ROI Through Scenarios, Not Certainty
The strongest business cases do not hide uncertainty. They price it.
Build at least three scenarios:
- Conservative: slower adoption, lower performance, longer implementation, and higher cost
- Base: the most defensible assumptions supported by current evidence
- Upside: stronger adoption or performance, without treating best-case outcomes as inevitable
Show the assumptions behind each scenario: adoption, transaction volume, model accuracy, utilization, time to deployment, benefit persistence, and recurring cost.
Separate direct cost reduction, revenue or margin improvement, avoided loss, faster decisions, improved experience, and strategic value. Not every benefit converts neatly into cash. That is acceptable. The problem begins when soft benefits are presented as guaranteed financial returns.
Here, the Hitchhiker’s Guide’s best-known advice is unexpectedly useful: “Don’t Panic.” Uncertainty does not make the case impossible. It means using ranges, assumptions, and decision gates instead of pretending to know one perfect answer.
5. Make Risk and Mitigation Part of the Financial Case
Risk should not sit in a final paragraph labeled “considerations.” It changes the expected return.
For each material risk, show what could prevent value, its likely impact, the mitigation and its cost, the owner, and the evidence that would trigger continuation, redesign, pause, or termination.
This supports stage-gated investment:
- Validate the baseline and data.
- Prove the workflow improvement.
- Demonstrate adoption and operational stability.
- Approve scaled deployment once the economics are clearer.
Funding the next evidence point does not weaken the ambition. It makes the investment governable.
How to Quantify the Cost of Low AI Maturity
An AI business case should estimate not only the return from moving forward but also the cost of staying where the organization is.
Low AI maturity often shows up as repeated pilots that never reach production, duplicated tools, underused platforms, manual workarounds, late governance remediation, delayed benefits, or productivity gains that never turn into operating results.
These costs may be spread across technology budgets, employee time, cloud consumption, rework, risk functions, and delayed revenue. Estimate current pilot spend, duplicated licensing, manual hours, delay cost per quarter, remediation, and value deferred because a use case cannot reach production.
This is where an AI readiness assessment strengthens the business case. It reveals whether the proposed investment addresses the real blocker—or adds another tool on top of unresolved data, governance, talent, or operating-model gaps.
AI Investment
Building an AI investment case without a reliable baseline?
Exadel’s Full AI Readiness Assessment turns AI readiness insights into a CFO- and board-ready investment case.
Presenting to the Board: What CFOs and Boards Actually Want to See
A board presentation is not a compressed technology strategy. It is a decision document.
The board does not need more certainty than the evidence supports. It needs clarity about the uncertainty.
A credible proposal should answer:
- What decision are you asking us to make?
- Which business outcome will change?
- What is the complete funding requirement?
- Which assumptions drive the return?
- What is the downside case?
- What evidence will we see, and when?
- Who owns value realization?
- When would we stop, redesign, or expand?
The emphasis should be on risk-adjusted return, not novelty.
Gartner reported in May 2026 that 63% of finance organizations said AI implementation had been slower than expected in 2025. A CFO will therefore test timelines, dependencies, adoption assumptions, and operating costs—not simply accept a projected productivity gain.
Apply the CFO test:
- Is the baseline verified?
- Is the total cost complete?
- Is value realization owned?
- Is the return presented as a range?
- Is risk funded rather than merely acknowledged?
- Are there evidence gates before the next release of capital?
The board is not buying an AI model. It is buying a change in business performance.
Common Objections—and How to Answer Them
“We already tried AI and it didn’t work.”
Diagnose the experience. Was the use case poorly selected? Was the data weak? Did the pilot lack a production path? Did adoption fail? Was value never measured against a baseline?
One failed pilot is not a verdict on AI. It is evidence about the conditions under which that initiative failed.
“Our data isn’t ready.”
Define the specific gap, determine whether it blocks the chosen use case, and include remediation in the investment model. “Data is not ready” is too broad. A named quality problem, timeline, and remediation cost are actionable.
“We don’t have the talent.”
Present the options: build, hire, buy, partner, or use a hybrid model. Include long-term ownership, not just initial delivery. Someone must operate, monitor, improve, and govern the capability after launch.
“The ROI is too uncertain.”
Reduce the size of the first decision. Use scenarios, stage gates, and defined evidence thresholds. Fund the next proof point rather than asking the board to accept the whole forecast at once.
“The risks are too high.”
Show the use-case classification, required controls, human oversight, monitoring, and exit plan. Risk avoidance is not the same as risk management.
For PE-backed organizations, the case may also need to show how AI contributes to the hold-period value creation plan. Our article on AI maturity assessments for portfolio value creation explores that perspective.
The Role of an AI Readiness Assessment in Building a Credible Business Case
A business case is only as credible as the baseline and assumptions underneath it.
An AI readiness assessment validates the factors that determine whether value can be delivered: the baseline, use-case feasibility, data and technical dependencies, production requirements, governance costs, talent gaps, ROI assumptions, and investment sequence.
An assessment does not replace the business case. It makes the business case harder to fool yourself with.
Different teams hold different parts of the truth. The business sees the opportunity. Technology sees the architecture. Data teams see quality and access constraints. Risk teams see exposure. Finance sees the assumptions and capital commitment. A structured assessment brings those perspectives into one decision model.
Exadel’s Full AI Readiness Assessment evaluates strategy, data, technology, talent, governance, and ROI, then translates the findings into an Investment and Business Case Model and Executive Strategy Presentation.
The result is not a promise that one ROI number will come true. It is a defensible view of what to fund, what to fix first, which outcomes are realistic, and what evidence should determine the next decision.
Ask the Right Question Before You Fund the Answer
A strong enterprise AI business case does not eliminate uncertainty. It makes uncertainty visible, measurable, and governable.
It tells the board why action matters, why this use case comes first, what the complete commitment is, what could derail it, and how the organization will know whether to continue.
Deep Thought eventually produced a perfectly clear answer. The problem was that nobody had defined the question well enough to make the answer useful.
Your CEO does not need “42” in their annual report.
They need the right question, a credible range of answers, and the evidence to decide what to do next that will produce the greatest value.
Explore Exadel’s AI Readiness Assessment
Build the objective baseline, roadmap, and business case behind your next AI investment decision
.png)
