The hardest choices are often the ones where the options are not clearly better or worse than each other. American philosopher Ruth Chang calls these choices “on par”. Her point is not that logic goes out the window. It is that even after you have done the analysis, there can still be a judgment call about what matters most.

I think it can be a useful way to think about AI. Decisions about where and how to use it are not just about cost, speed or productivity. They are also about what kind of organisation you want to be. Analysis helps, of course. You should understand the risks, likely benefits and trade-offs. But analysis alone will not always tell you which trade-off is the right one to make.

Take customer service as an example. You can use AI to automate more interactions, reduce cost and respond faster. Those are all valid goals. But depending on how far you go, you might also make it harder for customers to get empathy, flexibility or a sensible exception when they need one. That does not mean digital is bad or that humans are always better. It just means different choices optimise for different things, and that is where values start to matter.

That is why I keep coming back to a values lens. When the trade-offs are real, values help you decide which risks are acceptable, which benefits are worth pursuing and where the line should be. They do not replace evidence or governance. They make the basis for your decision clearer.

The other thing that matters is how your organisation actually works. The same AI tool can lead to very different outcomes depending on the culture, the guardrails, the incentives and the way people make decisions day to day. So successful AI deployment is not just about picking the right technology. It is also about whether your organisation is set up to use it well.

So maybe the real question is not simply whether to invest in AI. Maybe it is whether your values are clear enough, and embedded well enough, to guide the choices that come with it. If they are, use them to shape the organisation you want. If they are not, AI will still shape it for you, just more by accident than by design.