AI ROI Without the Hype

• 

AI investment is rising, but the returns are not always clear. Organisations are spending on tools, platforms, and training, yet many are still working out whether that investment is producing results worth the cost.

Deloitte’s 2025 survey of 1,854 executives across Europe and the Middle East found that 85% of organisations had increased their AI investment over the previous year. Most respondents reported that satisfactory returns on a typical AI use case took two to four years.

For leaders, assessing AI ROI means understanding what has improved, how AI contributed, and whether the benefit justifies the investment.

Start With a Specific Business Problem

AI creates value when it improves something that matters to the business, such as reducing delays, improving quality, or avoiding unnecessary costs. Before choosing a tool, leaders should ask:

• What problem are we trying to solve?

• How does the process perform, and what does it cost today?

• What would meaningful improvement look like?

• How will we assess AI’s contribution?

These questions establish a baseline. For example, a team introducing AI to prepare weekly reports should first record the time spent drafting, reviewing, and correcting them. The goal is to reduce the overall workload while maintaining accuracy and usefulness.

Measure the Whole Process

Usage figures can show adoption, but they do not establish business impact. More prompts or active users do not necessarily mean better results.

Choose measures that match the problem:

• Time: Is the work completed faster, including review and corrections?

• Cost: Has the cost of completing the work decreased?

• Quality: Are outputs more accurate and useful?

• Revenue: Has AI contributed to additional sales?

• Capacity: Can the team handle more work with its existing resources?

For the reporting team, a faster draft offers limited value if checking it takes longer. Measurement should cover the whole process and account for other changes, such as a lighter workload or a simpler template. Where practical, comparing similar tasks completed with and without AI can help clarify its contribution. 

Connect Improvement to Financial Value

Once an improvement is established, leaders need to determine what it is worth.

Time saved does not automatically become a financial return. Does it reduce overtime, avoid outsourcing costs, or allow employees to complete other necessary work? If it frees up time without reducing spending, record it as additional capacity rather than a cash saving.

Similarly, fewer errors may reduce rework, but any estimate of their financial value should be supported by evidence and clear assumptions.

Compare these benefits with full investment required to make AI work, including:

• AI tools, platforms, and implementation.

• Data quality, infrastructure, and system integration. 

• Employee training, workflow changes, and change management. 

• Governance, security, and ongoing oversight. 

• Maintenance, review, and continuous improvement. 

• Wider operational changes needed to integrate AI into the business.

Assess benefits and costs over the same period, and avoid counting the same benefit twice. For example, treating saved hours as both reduced labour costs and additional capacity.

Review the Evidence Before Scaling

A focused pilot helps test whether the benefits justify further investment. The reporting team could trial AI on one recurring report, then compare results and costs with the baseline. 

Before expanding, leaders should check: 

• Are the improvements consistent? 

• Is quality maintained? 

• Do the benefits justify the full cost? 

• Would the approach remain worthwhile at a larger scale? 

Different AI investments may need different timelines. A reporting assistant may show results quickly, while an application supporting product development may take longer. Both need agreed milestones and review dates to guide decisions about whether to expand, adjust, or stop.

AI ROI becomes clearer when organisations connect a specific business problem to measurable improvement and assess that improvement against its cost. This gives leaders a practical basis for deciding where AI deserves further investment.