When a board asks me what the return on AI is, I give them two tests. Once a business function has been automated, it should take fewer people to run it. And the total cost of the technology over its life should be lower than the benefits it delivers. If it doesn’t pass both, the board is paying for an experiment.
Most AI business cases never reach those tests. They are built around a tool and its efficiency gains, and they rarely ask whether the function, its data and its systems are ready. The return should be measured on the function, not the tool.
The board conversation works better when it moves on from “What is the ROI of this AI tool?” to:
“What capabilities does the organisation need so AI can create sustainable, long-term value?”
That opens up the right discussion, one centred on operating performance, future flexibility and organisational readiness.
Why the traditional business case doesn’t capture the full picture
Many business cases focus on tool-level efficiency gains. They model cost savings, faster turnaround times or automation uplift, often based on optimistic assumptions. Those gains can be real, but they don’t tell the whole story.
AI delivers meaningful value when it sits in an environment built for speed, clarity and consistency. Without those foundations:
- workflows remain slow
- data remains inconsistent
- systems remain difficult to integrate
- teams spend more time compensating for gaps than benefiting from the new capability
- the cost of implementation can exceed expectations
- the risk of low adoption increases significantly
In most cases the environment wasn’t prepared to take advantage of AI. Boards respond well when leaders set out this broader, more strategic context.
Three foundations boards should expect to see before AI scales
The most constructive board conversations focus on the operational prerequisites that allow AI to perform consistently and predictably.
1. Workflows designed for speed and clarity
Before deploying AI, high-performing organisations take time to redesign their workflows:
- simplifying steps
- clarifying ownership
- removing unnecessary approvals
- standardising handoffs
- aligning measurement
This work alone often shortens cycle times, well before AI is deployed. AI then builds on that stronger baseline.
When I helped to structure and operationalise an enterprise automation function in a multinational service business, the delivery method and the operating cadence were the decisions that mattered most. The automation depended on both.
2. Data infrastructure that supports real-time decisions
AI needs accurate, clean, consistent and well-governed data. Many organisations have data scattered across systems, defined inconsistently and not available in real time.
Building proper data infrastructure, with cloud-based platforms, clear governance and reliable access, creates immediate value:
- faster reporting
- better compliance
- less manual effort
- stronger decision-making
Once this capability is in place, AI can extend it further.
3. Systems that integrate with modern tools
Many organisations rely on applications built or customised years ago, and some of them are mission critical. These systems work well day to day but are difficult to extend or connect to modern workflows, and maintaining them consumes significant resource.
Modernising key applications, gradually or through targeted upgrades, enables the organisation to:
- improve flexibility
- simplify integrations
- keep business logic consistent
- support future innovation
This work brings immediate operational benefits, and it lets AI integrate cleanly.
How to frame the investment for the board
The board conversation shifts when leaders frame AI as enterprise infrastructure rather than as a project. Infrastructure requires:
- multi-year commitment
- ongoing budget, released in stages as capability milestones are met
- executive ownership at CFO or COO level
Projects are approved with short-term ROI expectations and are often defunded when quarterly pressures tighten. Infrastructure, like ERP, cloud platforms or core systems, is treated as a permanent capability that compounds over time. In my experience, the organisations getting value from AI fund it this way.
Applying the two tests
Funding AI as infrastructure doesn’t mean giving up on the numbers. For each function that is automated, the board should see two things:
- Fewer people running it. The function needs fewer FTEs, or the same team handles a lot more work without adding headcount.
- A total cost of ownership below the benefits. That covers licences, integration, data work, support and the cost of the change itself, as well as the subscription.
A function that passes both tests is ready to scale. If it fails, find out whether the problem is the workflow, the data or the systems before spending more.
The question that changes the conversation
When boards ask “What’s the ROI on AI?”, they are usually asking something broader:
“Are we investing in something that will deliver sustained, strategic value, or are we chasing a trend?”
The answer depends less on the AI itself and more on whether the organisation has built the operational foundations to use it well. The right board conversation looks at those foundations first.
Before the next board discussion on AI, it is worth asking three things: which workflows are ready, which data the board would trust for a decision, and which systems would need to change. Then agree how each automated function will be measured against the two tests. The answers make a better starting point than a tool-by-tool business case.
This is the question our AI and Automation diagnostic answers in two to four weeks. If your board is asking about the return on AI, book an introductory call.
First published in my LinkedIn newsletter, Leadership & Transformation.