How to measure the real ROI of AI

AI business cases often fail in opposite ways. Some promise dramatic savings without a baseline. Others count only software cost and ignore the value of faster, better and more scalable work.

Measure the whole workflow

The Productivity Commission notes that AI productivity gains require complementary investments in management, skills and business processes and may take time to become visible. A credible business case therefore includes implementation, integration, training, governance and ongoing operating costs - not only the model subscription.

Use six value lenses

• Capacity: validated staff time released and how it will be redeployed.
• Speed: elapsed time from trigger to customer or business outcome.
• Quality: error, rework, consistency and acceptance rates.
• Growth: qualified conversion, service capacity, retention or new revenue enabled.
• Experience: client effort, visibility, response time and staff usability.
• Risk: avoided incidents, stronger controls, better evidence and reduced key-person dependency.

Build the baseline before deployment

Record the current frequency, elapsed time, active staff minutes, handoffs, rework, error rate and cost. Separate measured facts from assumptions. Then define the post-launch window and data owner. Without that discipline, a faster-feeling process can be mistaken for a profitable one.

Do not count released time as cash automatically

Saving ten hours does not create a cash benefit unless overtime, contractor spend or headcount need changes. It may still create valuable capacity, but the business must show how that capacity is redeployed into more client work, faster sales, quality, innovation or reduced workload.

Use an evidence ladder

Label benefits as baseline, target, observed, validated or financially realised. Review at 30, 60 and 90 days. This avoids inflated case studies and creates better decisions about whether to expand, redesign or stop the use case.