Mind the Value Measurement Gap: Measure the Business Impact of AI Investments

As organizations mature in their execution of data and AI initiatives, a burning question remains: How do we measure the effectiveness of our teams and our impact on the business? This isn’t the perennial “What’s my data worth?” dilemma often answered theoretically. Today’s challenge is concrete: to define and track the metrics used to justify continued investment in data and AI innovation. After years of experimentation, this is the year the rubber truly hits the road.

A recent Snowflake sponsored study of 3,324 organizations found that 57% of respondents had implemented generative AI. Of early adopters, a whopping 92% reported a positive ROI. However, only 64% had actually measured it. Mind that gap. Fortunately, many Snowflake customers have. 

Back in December 2024 we recognized value-driven AI and data leaders — those who were building practices to monitor and measure the value that data and AI initiatives have delivered to their organizations. We then sat down with many of them to hear their experiences. What we came away with was the criticality of knowing the value of an AI initiative, the perceived risks of not delivering value and the knowledge that, while challenging, value measurement is doable. 

Lessons learned and best practices

As the CDO of a large U.S.-based food distributor noted, “When you execute an AI initiative, you are investing your career in it. It’s where careers are made.” Most executives don’t want to take that risk without the confidence that the project will be a success, and that means ensuring that their AI initiatives are going to deliver a return on the investment. They have to measure it. For some leaders, it’s not always straightforward. Value can be elusive when measuring improved decision-making or productivity gains. Others say, “You’ve got this.” Technology leaders know how to do it; it’s not different from any other tech initiative.

What they also tell us is that value doesn’t happen overnight. Early experimental stages usually don’t require a complete business case and might not be expected to show returns. Building a platform does require funding. However, low-hanging fruit — even if it’s infrastructure consolidation or modernization — can justify these foundational investments. Start with what you’re currently doing with data and do it better. Then the platform buildout can happen with incremental initiatives, adding capabilities as needed. For most teams, the need for value measurement comes when an AI initiative goes into production at scale.  

Another valuable lesson is to know when to stop, cut losses and reallocate resources. While much has been said about predicting and assessing the value of an AI initiative, some data executives struggle with the decision to stop.

Finally, don’t get discouraged. Think of value measurement as a muscle. It gets stronger as it’s used. Value measurement seems daunting at first, and potentially painful. However, with practice, it gets easier over time.

Six steps to value measurement

In Snowflake’s CDO’s Guide: Measuring AI’s Business Value, AI and data leaders share these lessons learned and best practices. Here are six steps to help get you started:

1. Identify the impact owners. Consider both where the costs are incurred and the value delivered. This is your value measurement team.

2. Agree on the business metrics. It’s not about a dashboard but rather about the actual challenge or opportunity. Think marketing conversions, sales orders, product defects or truck shorts. These are your measures of success.

3. Benchmark current values to predict impact. What is the current state of the metric you’ve chosen? This will be your baseline. How do you expect it to change? That will be your goal post. 

4. Monitor, measure and report outcomes. Regular communication to senior leadership builds the case for continued investment; communication across the company generates future initiatives and builds grassroots support and FOMO.

5. Ask for help from your data and AI platform provider. An AI-ready platform should actively enable your AI strategy. The Snowflake value engineering team offers a business value assessment and business value realization to help you ensure a strategic alignment between your key business objectives and the capabilities of Snowflake’s AI Data Cloud. The assessment is a collaborative exercise conducted prior to investment, establishing a baseline of current state and end goals; the realization is conducted periodically during the lifecycle of an initiative to ensure business value delivery. The process includes both qualitative and quantitative analysis, with employee surveys, industry benchmarks and business value realized used to measure both costs and benefits. The exercise also includes a FinOps review to assess cost visibility, control and optimization.

6. Don’t forget your bottom line. Work collaboratively to ensure alignment between investment and strategic outcomes. In other words, costs incurred must deliver measurable net business value.

Measuring AI’s business value is key to justifying continued investment, prioritizing the applications of it and balancing its future benefits against potential risks. Measurement is challenging: Benefits can be indirect (for example, better decision-making or increased productivity); effects can be distributed and varied; and value can accrue over time. But it is doable.

For more best practices on how to measure and monitor AI outcomes, see our recently published CDO’s Guide: Measuring AI’s Business Value.