By Donald D. Hook — Former CTO & CIO, Full On Consulting | April 2026 | 12 min read
My feed is full of AI business case frameworks right now. Five steps to AI ROI. Seven steps to your AI roadmap. Proven approaches from advisors with years of experience. Some of them are not wrong. But most of them are missing the steps that actually determine whether your AI initiative succeeds or fails.
After 20+ years leading enterprise technology transformations as a CTO and CIO — including AI and data initiatives across industries — I can tell you that the gap between a business case that looks good on a slide and one that actually delivers comes down to a handful of things that most frameworks skip entirely.
Here is what is wrong with most AI business case approaches — and what a real one requires.
6 Critical Flaws in Most AI Business Case Frameworks
❌ Starting with the vision instead of the problem
The most common first step in AI business case frameworks is 'paint the vision' — with examples like 'build an AI platform' or 'make faster decisions.' That is not a business problem. It is a technology preference dressed up as strategy. A vision without a grounded, specific problem produces AI initiatives that technically deliver but never change anything.
✓ What to do instead:
Start with a named, measurable business problem. Not 'improve efficiency' — but 'our invoice processing takes 12 days, costs $47 per invoice, and has an 8% error rate. Our top competitor processes in 2 days at $11 per invoice.'
❌ Projecting ROI before assessing data readiness
Most AI business case frameworks go directly from 'define the opportunity' to 'calculate the ROI' — without ever asking whether the data required to power the proposed AI solution actually exists, is accessible, and is clean. This is how organizations end up with approved business cases they cannot execute.
✓ What to do instead:
Data readiness assessment belongs between problem definition and ROI projection. The cost of data remediation — often 2–3x the cost of the AI work itself — belongs in the business case.
❌ Treating AI as a single thing
The most damaging assumption in most AI frameworks: that 'AI' is one category of solution. Predictive AI, Generative AI, Agentic AI, Computer Vision, and NLP have completely different cost structures, data requirements, implementation timelines, and risk profiles. A business case that says 'we will use AI to solve this' without specifying which type is not a business case — it is a hypothesis.
✓ What to do instead:
Understand the AI landscape before designing the solution. Match the right AI type to the problem based on fit — not on what is trending. Not every problem needs a large language model.
❌ No change management plan
AI ROI is not realized when the model goes live. It is realized when people change how they work. A demand forecasting model that planners ignore. A document processing tool that agents work around. A customer service AI that the team doesn't trust. These are not technology failures — they are change management failures. And they happen constantly, because most AI business cases treat adoption as a given.
✓ What to do instead:
Change management must be a line item in the business case — with a plan, a budget, and an adoption measurement framework. ROI projections that assume 100% adoption without a plan to achieve it are not credible.
❌ Governance and risk are missing entirely
In regulated industries — healthcare, financial services, insurance, government — the AI business case must address governance, compliance, and risk management. In every industry, boards and legal teams are increasingly asking hard questions about AI outputs, data usage, and liability. A business case that ignores these is not board-ready, and it will get sent back.
✓ What to do instead:
AI governance — policy, oversight, risk management, data classification, vendor due diligence — belongs in the business case scope and budget. Treating it as an afterthought is how organizations end up in regulatory trouble after the fact.
❌ No build vs. buy vs. partner decision
Most frameworks skip from 'define the ROI' to 'scope the first project' without addressing the foundational question: are you building this AI capability internally, buying a vendor solution, or partnering with a specialist? This decision changes the cost estimate, timeline, talent requirements, and risk profile by an order of magnitude.
✓ What to do instead:
The build/buy/partner decision must be made — and documented — before the business case numbers mean anything. Each path has different implications for total cost, speed to value, and long-term ownership.
The Right Sequence for an AI Business Case
The organizations getting real, sustained value from AI are not the ones who moved fastest. They are the ones who built the business case in the right order — starting with the problem, not the technology.
Define the specific business problem
Named, measurable, and understood deeply before any technology discussion. Not 'improve efficiency' — but a specific process, a specific cost, a specific gap with a specific number attached.
Understand the AI landscape
What are the different types of AI — Predictive, Generative, Agentic, Computer Vision, NLP — and what can each actually do? What does each cost, what data does it require, and what are the implementation risks? You cannot match the right tool to the problem without this foundation.
Match the right AI type to the problem
Based on fit — not trend. The most expensive, most complex AI is not always the right one. A simple predictive model may solve your demand forecasting problem better, faster, and at one-tenth the cost of a generative AI solution.
Assess data readiness
For the specific AI type you have selected, does the required data exist? Is it accessible? Is it clean enough? If not, what does remediation cost and how long does it take? This belongs in the business case — not as a discovery item after approval.
Make the build vs. buy vs. partner decision
Before the financial model means anything, you need to know which path you are on. Build, buy, and partner have completely different cost structures, timelines, and risk profiles.
Build the full business case
Now — and only now — build the ROI model. Include implementation costs, data readiness costs, change management, governance, and ongoing operations. Use conservative estimates. An ROI projection that requires 100% adoption or perfect data is not a business case — it is a best-case scenario.
Define the pilot
Identify the highest-confidence use case to validate the approach before full investment. The goal of the pilot is to prove the hypothesis with real data — not to 'sell' the initiative internally.
Why Matching the Right AI Type to the Problem Matters
This is where an experienced AI advisor earns their value. Most business leaders cannot evaluate AI tool capabilities themselves — and most of the vendors they talk to have a strong incentive to recommend their platform, not the right platform.
The default in most organizations right now is to reach for a large language model or a generative AI solution for every problem — because it is the AI they have heard of. The result is organizations spending $500K building a generative AI solution to a problem that a $50K predictive model would have solved better.
| AI Type | Best For | Key Data Requirement | Common Mistake |
|---|---|---|---|
| Predictive AI | Forecasting, anomaly detection, risk scoring | Clean historical data at scale | Using GenAI when prediction is the actual need |
| Generative AI | Content creation, code generation, summarization | Curated knowledge bases, prompt engineering | Deploying without data classification or governance |
| Agentic AI | Multi-step autonomous task completion | Reliable APIs, structured workflows | Deploying before process design is complete |
| Computer Vision | Image/video analysis, quality inspection | Labeled image datasets | Underestimating labeling cost and time |
| NLP | Document processing, chatbots, search | Domain-specific text corpora | Treating off-the-shelf models as production-ready |
The Meta-Problem With Most AI Frameworks
Many AI business case frameworks correctly diagnose that organizations "start with the AI." But then their Step 1 is "paint the vision" — with examples like "build an AI platform" or "achieve faster decisions." That IS starting with the AI, just with a better slide title. Rebranding the same mistake as a methodology does not fix it.
The organizations that succeed with AI start with a specific problem, find the right tool for that specific problem, validate that their data can support that tool, and build a business case that a CFO can stress-test. Everything else is a framework looking for a problem to justify itself.
Not Sure If Your AI Business Case Is on Solid Ground?
Full On Consulting helps enterprise leaders build AI business cases that start with the right problem, match the right tool, and produce ROI projections that survive CFO and board scrutiny. We have led AI initiatives from concept to production — and we will tell you honestly if your current approach has gaps.
AI Strategy & Readiness ServicesSchedule a Strategy CallFrequently Asked Questions
Why do AI business cases fail?
Most AI business cases fail because they start with technology rather than the business problem. Common failure points include: building the business case before assessing data readiness; treating all AI types as interchangeable when they have completely different cost, data, and implementation profiles; skipping change management; and omitting governance and risk from the plan. The result is an ROI projection built on assumptions rather than validated requirements.
What should come first in an AI business case — the vision or the problem?
The business problem must come first. A vision statement like 'build an AI platform' or 'make faster decisions' is not a business problem — it is a technology preference. The business problem should be specific, measurable, and named: 'Our claims processing takes 14 days and costs $X per claim. Our top competitors do it in 3 days.' Only after the problem is defined precisely can you identify the right AI approach, assess data readiness, and build a credible ROI estimate.
What are the different types of AI and why does it matter for the business case?
The five primary enterprise AI types are: Predictive AI (forecasting, pattern recognition), Generative AI (content, code, synthetic data), Agentic AI (autonomous task completion), Computer Vision (image and video analysis), and Natural Language Processing (document understanding, chatbots). Each has a completely different cost structure, data requirement, implementation complexity, and risk profile. Using a Generative AI solution where a simple predictive model would work costs 5-10x more and introduces unnecessary complexity. Matching the right AI type to the problem is one of the highest-leverage decisions in the entire business case.
How does data readiness affect an AI business case?
Data readiness is the most common reason AI business cases collapse in execution. Every AI type requires specific data — the right volume, the right quality, the right accessibility, and the right governance. A Predictive AI model for demand forecasting requires years of clean historical transaction data. A Generative AI solution for customer service requires structured knowledge bases. Before projecting ROI, you must assess whether the data required for your specific AI approach exists, is accessible, and is clean. If it is not, data remediation costs belong in the business case — and they are frequently 2-3x the cost of the AI work itself.
Why is change management required in an AI business case?
AI ROI is not realized when the technology goes live — it is realized when people change how they work. An AI model that produces better demand forecasts only delivers value if planners actually use those forecasts instead of their spreadsheets. An AI-powered customer service tool only reduces cost if agents change their workflow to use it. Without a change management plan — stakeholder engagement, communication, training, adoption measurement — the technology can be technically successful while the business outcome never materializes. Change management costs belong in the business case, not as an afterthought.
What is the right sequence for building an AI business case?
The right sequence is: (1) Define the specific business problem — named, measurable, and understood before any technology discussion; (2) Understand the AI landscape — what each type can and cannot do, what it costs, and what data it requires; (3) Match the right AI type to the problem — based on fit, not trend; (4) Assess data readiness for that specific AI type; (5) Build the full business case including implementation costs, change management, governance, and a realistic ROI model; (6) Define the pilot — the highest-confidence use case that validates the approach before full investment.