In 2025, 85% of organizations increased AI investment, and 91% plan to do the same this year, according to Deloitte.
Despite continued spending, however, ROI lags behind, with just 6% seeing payback within one year. While AI use cases tend to have a longer payback period, often in the 2-4 year range, companies can’t afford to keep spending money without some measure of its practical impact both immediately and over time.
Here’s what companies need to know about AI ROI, why it matters, and how to measure it effectively.
What is AI ROI?
AI ROI (return on investment) is a measure of the financial and operational value that an organization realizes from its AI investments.
This value is measured relative to the total cost of these investments. This cost includes infrastructure, model pricing, token costs, data preparation, model review and adjustment, and any in-house and outsourced talent necessary.
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.
Laying the groundwork for AI ROI
Deploying AI doesn’t guarantee ROI. While service agents and inference engines offer the potential for faster and more accurate outputs, ROI is largely determined by how companies use AI, not the model itself.
Consider a business that deploys an API-based model using ChatGPT or Google Gemini. Every input and output costs tokens, the cost of which can quickly add up. If teams spend time asking similar questions over and over, or create prompts that are too vague or too specific to deliver actionable answers, ROI suffers.
No matter how advanced the model, the human element of the equation determines ROI. Used effectively, token-based options allow companies to enable predictive and proactive operations across finance, HR, sales, marketing, and IT functions. Deployed simply to keep pace with the competition and without the necessary training and skills, AI can prove costly and complex.
ChatGPT creators OpenAI recommend asking four key questions to establish AI ROI table stakes.
1. How much useful work gets done? First, companies need to define useful work and what “done” means in the context of their business. This allows the mapping of token usage and costs to specific outcomes, such as addressing customer issues or improving business decisions.
2. What does a successful task cost? Next is cost. This cost includes the model itself, whether per-seat or per-token, any costs associated with adapting or integrating the model, and the time and effort required to review and adjust the model over time.
3. How often does AI find the correct answer? Correctness also impacts overall efficacy. For example, a lower-cost model that only returns correct answers in 40% of cases may not be as cost-effective as one that costs more but is 85% accurate.
4. Is cost efficiency increasing or decreasing? Over time, does the same workflow cost more, less, or the same? If costs are rising but workflows are no more complex, organizations may need to reevaluate their AI budgets and token limits.
Three approaches to AI ROI measurement
Once companies understand what their AI is doing and where deployments may need adjustment, they can begin measuring ROI in earnest. Research from MIT Sloan suggests three possible AI ROI approaches, each of which comes with key metrics and potential pitfalls.
The function-focused approach
This approach focuses on a single function or a small number of functions or processes.
Key metrics:
- Response times
- Error rates
Possible pitfalls:
- Siloed metrics
- Limited transparency across the organization
The coordinated approach
The approach takes a balanced approach to broadly applicable AI tools.
Key metrics:
- Average cost of AI compared to the average cost per worker for the same function
- Sales conversion rates
Possible pitfalls:
- Limited comparability
- Fragmented insights
The enterprise portfolio approach
This approach focused on enterprise-wide implementation and governance of AI.
Key metrics:
- Net present value (NPV) and internal rate of return (IRR)
- Employee net promoter score (eNPS)
Possible pitfalls:
- Process and operational complexity
- Lack of exploratory initiatives
What are some examples of AI ROI?
It’s one thing to understand the concept of AI ROI. It’s another to see it in practice. Here are two examples, one hypothetical and one from a recent study.
In our hypothetical example, an e-commerce company aims to increase sales volume and reduce customer churn. Data analysis shows that over the past six months, sales numbers have largely flatlined, and the total number of returning customers has dropped.
To address this issue, the business deploys an API-based AI solution and connects it to sales and marketing software. Using a combination of historical data and first-party data provided by consumers, the company creates a handful of AI prompts to help identify the source of churn and suggest ways to improve key metrics. The AI agent produces a new set of marketing materials and automates the post-purchase customer follow-up process.
Metrics are tracked each month and reveal a steady increase across both total sales volumes and repeat customer business, the value of which outpaces the cost of the initial and ongoing AI expense. As a result, AI ROI is positive.
Our real-life ROI study comes from Mass General Brigham (MGB) hospital in Somerville, Massachusetts, which deployed AI-powered ambient documentation technology (ADT) to help reduce clinician burnout. ADT solutions listen to clinician-patient conversations and then draft structured notes using generative AI.
Doing so reduces the amount of time clinicians spend making notes during appointments and the amount of time they spend reviewing the accuracy of these notes post-visit. In the MGB study, the use of ADT was associated with a 21.2% absolute reduction in burnout prevalence.
This is the other side of the AI ROI coin. While return on investment can be measured in terms of pure revenue generated or money saved, it can also be measured by the impact it has on people. In the case of MGB, AI-assisted notetaking reduced clinician burnout, which improved their efficiency and the efficacy of the hospital as a whole.
Why does AI ROI matter for organizations?
AI ROI isn’t just about monetary return. It’s about the larger impact of AI on organizations, from improved customer engagement to reduced employee burnout, and yes, revenue.
Companies also need to consider the push-pull nature of increasing AI usage with rising AI costs. Even as models become more sophisticated, tokens and operations become more expensive. This makes ROI both a critical metric and a moving target. If businesses don’t regularly measure, evaluate, and adjust AI strategies, they could face rising budgets and falling impacts, in turn making it harder to justify ongoing AI investments.
CloudZero, the AI ROI company, delivers the platform that closes the gap. Finance and engineering teams get a shared, real-time view of AI costs by model, feature, and token, and CloudZero solution maps this spend to the business metrics and outcomes that justify continued investment.
See your AI ROI in real-time with CloudZero. Schedule your demo today.
Frequently asked questions about AI ROI
