Capability: Treat AI spend like a resource you manage

The second part is capability. For a CFO, it lands as a money question. Microsoft's Satya Nadella has a useful name for it — token capital — the AI capability a company builds and owns, set against the human capital already doing the work. Like any resource on your balance sheet, it should be planned, measured, and allocated on purpose, not handed to whoever files the most enthusiastic request.

Nadella was describing the frontier of the economy. The same logic lands harder one level down, on the CFO whose learning loop still runs through an ERP designed to record the past rather than act on it.

The aim is balance between token capital and the human capital already doing the work, so the two reinforce each other on a given process instead of doubling up. A CFO is already fluent in this. Capital allocation under constraint is the job. AI just adds a line item that behaves a little differently and moves faster than most.

Framed this way, capability stops being an IT abstraction and becomes something you can govern with controls you already trust.

The move that separates real ROI: Redesign, not automation

Here is the point most likely to change what your company does on Monday. There are two ways to put AI to work on a process, and they pay out very differently.

The first is to automate a task inside a process you already have. This works. You will see efficiency gains, and they are real. But you are pouring new capability into a container that was shaped around human limits: the handoffs, the approvals, the workarounds that exist because people needed them.

The second is to rebuild the process itself around what AI now makes possible, stripping out the constraints that were baked in back when only people did the work. That is harder. It is also where the higher returns live. The companies pulling meaningful ROI out of AI are not the ones that bolted it onto yesterday's workflow. They are the ones that asked what the workflow should look like now and rebuilt accordingly.

This is the line between an activity and a foundation. Automation gives you a faster version of what you already had. Redesign gives you something the old operating model could not produce at all.

One more question the board will eventually ask

AI also shifts the security and governance picture. The risks are not the same ones you have managed before, and not every control you already own transfers cleanly to them. It is a real topic and worth its own conversation, but it sits alongside the foundation rather than at the center of this answer.

So what do you actually say?

When the board asks what you're doing with AI, skip the pilot count. Describe the foundation instead. Tell them what the company can now draw on that it couldn't before. Tell them how you're allocating AI spend as deliberately as any other capital. Tell them which processes you're rebuilding rather than merely speeding up, and how fast you intend to keep building.

hat is an answer a board can trust, because it describes something that lasts. The pilots will come and go. The foundation is what decides whether any of them were worth running. The CFOs who can describe it are the ones who make the AI question look easy — not because they ran better pilots, but because they built the thing underneath.