Measure the ROI of AI Investments with Confidence

Most leaders ask a simple question. Is our AI investment paying off? However, the 2025 AI Deployment & Governance Survey Report by the Governance Institute of Australia reports that 93.7% of respondents have not been able to measure return on investments (RoI) in artificial intelligence (AI). That gap makes capital decisions harder and leaves value on the table.

It doesn’t have to be this way. This article is trying to make this easier for leaders and investment teams to invest in AI with confidence by providing a high-level framework for how RoI of AI investments can be defined and measured.

Measuring RoI on emerging technology investments is always difficult and AI is no different. However, in my experience, as a first step to overcome this challenge, it helps to contextually define the type of investment. A common terminology and understanding of investments make it easier for decision makers to obtain confidence in what they invest in, and equally importantly, what they don’t invest in. It is also a very useful approach to consistently measure and monitor a portfolio of AI investments – something that most companies are dealing with as AI impacts multiple parts of its organisation, customers and suppliers. Table 1 below provides a classification method of AI investment classes, including what each AI capability focuses on, and examples use cases:

Table 1: AI Investment Classes

Tying AI investments to outcomes is obviously critical in measuring the RoI. This is often easier said than done and therefore also a point that can be tricky, as an AI investment can have many direct and indirect impacts. To help make this easier, Table 2 below provide examples of outcomes and metrics that can be used to measure RoI:

Table 2: AI Investment Classes, Outcomes and KPIs

How to apply in three steps

• Tag each initiative with one primary capability and up to two secondary capabilities.
• Select one outcome only (keep it simple). Revenue, Cost, Service quality, or Resilience.
• Pick one KPI from Table 2 and connect it to dollars with a simple formula.

From capacity to cash

Many AI use cases free up time rather than reduce spend. Track both capacity value and cashable savings. Here is a simple approach on how to deal with this:

• Capacity value = (hours freed) x (loaded rate) x (adoption) x (success).
• Cashable savings = (capacity value) x (cash realisation percent), where the latter is measured once actions reduce spend such as contract cuts, roster changes, or backfill avoidance.

Use Case Example

Imagine a Procurement team seeking to invest in a GenAI solution (the Generative AI Investment Class) to rapidly complete contract summaries. Applying the above framework would look as follows:

• Outcome: Cost reduction or productivity (obviously both are possible).
• KPI: Minutes per contract, where lower is better.
• Capacity value – where Procurement team members “get time back” to use on more value adding activities, and cashable savings – for example if the number of resources can be reduced due to higher productivity. Both are calculated as set out above (Tip: This is best done with Finance once review spend is reduced.)

Go forth and invest with confidence

Hopefully this article has been of value to you, and it adds to your capability to develop an AI Investment Case or make AI investment decisions with more confidence. As always, I would love to hear your thoughts on this topic as this is a major inhibitor for many organisations – not just in Australia.

If you want support to measure ROI across your AI investment portfolio, then I can help you apply the framework to your AI investments. Feel free to book a thirty-minute session with me by sending a LinkedIn DM or connect with us, and we can discuss your context and the fastest path to a high-impact AI Investment.

Tom Dissing is the founder and Managing Director of Technology Connect. He helps boards and executives drive growth and avoid disruption through artificial intelligence (AI), innovation and venture building

Copyright © 2025 Technology Connect. All rights reserved.