How you should calculate the ROI in AI
The problem isn’t asking about ROI. Taking too rigid an approach could cost you the real prize: broad, compounding businesses transformation.
There’s a growing impatience in boardrooms. As artificial intelligence investments climb, so does the pressure to answer one blunt question: What’s the return on investment?
It’s a fair question — on the surface. After all, if traditional technology investments had to clear a hurdle rate, why should AI be any different? Yet treating AI, particularly generative and agentic AI, like any other IT rollout risks misunderstanding its nature and stifling its potential before it takes root.
The problem isn’t asking about ROI. It’s asking too early, too narrowly, and too literally. Taking too rigid an approach could cost you the real prize: broad, compounding businesses transformation. Here are a few considerations CFOs and other leaders should keep in mind when assessing the ROI of AI technologies.
Why the old ROI playbook doesn’t fit
Traditional AI — essentially predictive analytics and automation — lent itself to clean ROI maths. Predictive maintenance, for example, can cut downtime. Invoice automation can trim headcount and deliver savings measurable within a budget cycle.
Generative AI and agentic AI, however, are a different game. Their returns aren’t always tied to a single cost saving or revenue stream. They create new work products, redesign roles, and increasingly make autonomous decisions.
Some AI advisers and vendors may point to the ROI they’ve delivered elsewhere. This can be helpful but it can also be a false promise. Even when comparing identical initiatives, the economics rarely translate. Technology is evolving rapidly: doing the same project a year later may cost half as much.
Although it’s hard to quantify the ROI of AI investment directly on a spreadsheet, our global research portrays a positive outlook: 75 per cent of organisations who have deployed generative AI state the ROI is meeting or exceeding expectations.
What distinguishes these successful organisations from others comes down to factors like data quality, workforce readiness, leadership alignment, and how deeply the AI is integrated into your core processes. AI doesn’t create value in isolation; it amplifies your existing strengths and exposes your weaknesses.
Are AI agents software or employees?
A new complexity is emerging: how do we treat AI agents in our financial models? Are they priced like software licences, or equivalent to employee salaries?
Some vendors are adopting labour-replacement pricing models, charging a fraction of a human’s cost for similar work. Others use per-outcome pricing or consumption-based billing models.
At first glance, comparing AI agent costs to human salaries feels logical, but it’s fraught with issues. Cheaper, faster, open-source alternatives are already emerging — and they will only accelerate. If you peg an AI agent’s cost model tightly to today’s human equivalent, you risk overpaying while nimbler competitors achieve similar outcomes for a fraction of the cost.
Timing is everything
Expecting full financial ROI within 12 months of deploying AI is like planting an orchard and demanding fruit by next quarter. Early AI initiatives require groundwork — training data, employee adoption and iteration cycles — before they bear results.
Leading organisations focus on leading indicators early: model accuracy, process speeds, adoption rates. They measure short-term progress towards long-term impact.
Ironically, the greatest risk today isn’t investing in AI without perfect ROI certainty — it’s not investing at all, while competitors build capabilities that won’t just cut costs, but reshape industries.
Broader than a use case
The smartest companies aren’t just deploying AI to automate a task — they’re wiring it into how the business operates. Predictive maintenance in manufacturing, for example, doesn’t just reduce downtime; it enables agile, just-in-time production and opens new service revenue streams.
ROI, in these cases, isn’t the result of a single project — it’s the compound effect of an
AI-enabled business model. Successful organisations are embedding AI into the decision-making DNA of their businesses, across end-to-end processes. If you’re measuring project-level ROI alone, you’re missing the point.
A new discipline for executives and boards
None of this argues for blind faith in AI. Every AI initiative should start with a business case, a working hypothesis for value, and clear milestones. But demanding definitive ROI too soon, or measuring the wrong things, could suffocate investments that matter most.
Boards should insist on ROI frameworks that track tangible gains where possible, while also capturing strategic enablers like speed, resilience and competitive differentiation.
And while it’s tempting to wait, hoping for lower costs or clearer case studies, delay comes at a price. Capability, context and culture can’t be fast-tracked.
If you’re not building foundational muscle now — through data infrastructure, pilot initiatives, and organisational learning — you’ll find yourself behind when the tech is cheaper, and your competitors are already scaling.
ROI must guide your AI journey, not gate it.
Dr Kellie Nuttall is AI Institute Leader at Deloitte Australia.
-
Disclaimer
This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.
Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.
About Deloitte
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. Please see www.deloitte.com/au to learn more.
Copyright © 2025 Deloitte Development LLC. All rights reserved.
-
