Quick Answer
A strong AI business case ties a specific goal to a measured outcome and a fully-loaded cost. Most fail because they skip one of the three: no clear mandate, an over-broad "AI fixes everything" scope, or a cost estimate that ignores adaptation and error-correction. Build it in six steps: define goals, identify uses, break work into tasks, evaluate models, assess total cost, then launch and refine.
Most companies are now spending on AI. Far fewer can show what they got back.
Recent data shows a disconnect. While businesses are spending on AI, they’re not necessarily getting back what they put in.
Consider research from McKinsey, which reports that nearly 90% of businesses are regularly using AI, with 64% saying that AI enables innovation. A report from the MIT NANDA initiative, meanwhile, found that 95% of generative AI pilots fail.
With intelligent solutions now the expectation for operations, organizations need a way to build, test, and execute effective AI business cases.
Common challenges in AI business adoption
AI adoption is typically informed by small-scale use cases. For example, a business might use large language models (LLMs) to create a marketing pitch or analyze recent sales trends. Buoyed by early success, C-suite executives earmark bigger budgets for AI investment.
Despite best efforts, however, these investments don’t always deliver ROI. Here are three common challenges that can frustrate AI efforts.
Challenge #1: Everyone is doing it, so we should do it too
As noted above, the majority of businesses are experimenting with AI to help improve efficiency, track trends, and meet customer expectations. With more than half of consumers now using AI for awareness, consideration, and decision-making, companies can’t afford to ignore the impact of intelligent solutions such as chatbots and agentic AI that enable real-time decision-making.
Challenges emerge, however, when organizations take an arms-race approach. As competitors adopt AI, it’s tempting to do the same, even if there’s no solid business plan in place. While keeping up with the Joneses works initially, it leads to exactly what the data shows: Pilots that don’t live up to expectations because they didn’t have a clear mandate.
Challenge #2: Seeing AI as the silver bullet
AI solves a remarkable range of problems, but no single tool solves all of them at once. Many LLMs and GenAI models, however, are marketed as if they are the answer to every question and the solution to every problem.
The challenge here is a critical distinction. General AI, also known as artificial general intelligence (AGI), is a hypothetical system that can learn, reason, and respond like a human being to almost any situation. Generative AI, meanwhile, is a much narrower form of intelligence that is designed to solve specific problems and handle smaller-scope questions.
Although improvements in agentic AI are moving the needle closer to a generalized model, the mistake isn’t using AI. It’s treating AI as a silver bullet. These tools genuinely solve hard problems in sales, marketing, HR, finance, and IT, but they solve them one scoped problem at a time, not all at once.
Challenge #3: We’re not prepared to pay the AI piper
Most GenAI tools offer a free service tier that allows companies to see how they work and what they offer. If models deliver benefits, investment comes next.
Then, payment comes due. And for many businesses, expected and actual costs don’t align. This is often because teams only account for a portion of AI costs. The Generative AI Cost Equation developed by George Westerman of MIT helps accurately account for these costs.
Total cost = Cost of using AI + cost of adaptation + cost of detecting and fixing errors.
The cost of using AI is any direct expense, such as paying for the model itself, implementing APIs, and ensuring the LLM has access to enough computing power. The cost of adaptation refers to the time and expertise required to customize the model for specific business tasks.
Finally, the cost of detecting and fixing errors refers to the resources needed to ensure models are trained, supervised, and corrected as needed.
With the total AI cost calculated, companies should then compare it to the cost of manual processes. If costs are similar or manual operations are less expensive, AI investment may not make sense.
Report
Finance needs to prove AI’s return: CloudZero report
260 senior finance leaders (more than half CFOs) told us why the speed of seeing AI spend, not the size of it, separates who pulls ahead on AI from who gets burned.
AI business case examples
AI potential is just that: possible success under the right circumstances. Here are four business case examples of companies using AI to deliver measurable outcomes.
Customer agents: PODS
PODS used Google Gemini to create an adaptable ad campaign. Digitally displayed on the side of trucks, the campaign changed in real-time and created 6,000 unique headlines in 29 hours across 299 neighborhoods in New York City.
Employee agents: Uber
Uber used AI agents that enabled the summarization of user communications and allowed front-line staff to surface context from previous interactions.
Data agents: Stord
Fulfillment and software provider Stord used the Gemini Enterprise Agent Platform to build ML models capable of predicting estimated delivery dates and optimizing demand forecasting.
Security agents: Mitsubishi Motors
Mitsubishi Motors used AI-powered SOAR and SIEM solutions to simplify security management with automated threat detection and response.
Six steps for AI business case success
Building a business case for AI doesn’t happen by accident. Instead, it requires concerted effort from C-suite executives, IT leaders, and front-line staff.
This effort can be broken down into six discrete steps.
1. Define clear goals
First, companies need to define what they want to achieve with AI. Is the goal streamlining a specific process? Reducing data entry errors? Improving trend predictions or creating long-term sales strategies?
Clear goals are the foundation of any AI business use case.
2. Identify concrete uses
Next, teams need to identify concrete uses for AI. For example, if the goal is to streamline business security, what role does AI play? Examples include user behavior analysis to limit insider risks, or zero-day detection to help prevent network compromise.
3. Break workflows into tasks
Equipped with goals and uses, companies should break workflows down into tasks. In the case of user behavior analysis, this could include login review such as user location, time of request, and actions taken once access is granted. It could also take the form of comparison with previous data to pinpoint concerning behavior patterns.
4. Evaluate multiple models
Models such as Gemini, Claude, and ChatGPT come with benefits and drawbacks. The best model for your business depends on your goals. For example, Claude is often used for natural writing and complex coding, while Gemini infrastructure supports deep web research.
5. Assess the total cost
Using the Generative AI Cost equation mentioned above, calculate the total cost of deploying, adapting, and improving the model. But that equation tells you what AI should cost; the number that matters is what it does cost once running, attributed to the feature, team, or customer consuming it.
Aim for a line you can defend: this workflow consumed X tokens on Tuesday at a cost of $X, resolved Y tickets, and saved Z hours. Estimates get a project funded. Attribution, which is what CloudZero provides, keeps it funded by converting spend into cost per outcome.
6. Launch, test, and refine
With use cases, models, and costs identified, companies can launch AI pilot projects. Testing comes next, followed by full deployment if models work as intended.
Set the measurement before launch: the outcome the use case should move, its cost per unit of that outcome, and the threshold worth scaling. Then review on a cadence, with budget following the use cases whose cost per outcome is trending down and leaving the ones that aren’t.
Making the most of AI investments
AI investments don’t offer automatic returns. Instead, businesses need to create customized use cases that identify why AI is being used, what specific advantages AI can provide, how these benefits will be measured, and when changes are needed to ensure investments continue to drive ROI.
Making the most of AI investments. Connect every AI dollar to every operational outcome with CloudZero. Schedule your demo today.
Frequently Asked Questions about building an AI business case
