James, AI, and the Case for Works

The Bible’s Letter of James was written to people who had plenty of belief and not much to show for it. They could recite the creed. They could pass any test of doctrine you put in front of them. What they couldn’t always do was point to something in their lives that the belief had visibly changed. James had a blunt response to this. “Show me your faith without your works,” he wrote, “and I will show you my faith by my works” (James 2:18).

It’s a strange and useful sentence, because it refuses the thing we most want to do, which is to be judged by what we hold in our heads.

I keep thinking about that sentence when I write — about AI, or about anything else. Am I writing out of something I’ve actually done? Or is this just my attempt to be poetic?

Most of the public conversation about AI right now appears to be a (maybe one-way) conversation about what it is. Is it conscious, or only convincing? A mind, or a mechanism? Something that automates and diminishes human participation in work, or something that leads to new ways to work, unfettering imagination and democratized creation? The first chapter of something, or the last chapter of something? These are real questions, and I don’t want to wave them away. But notice what almost none of the conversation is about: what the person doing the writing has actually done with the thing. I don’t mean that as a complaint or criticism, only an observation. It’s more a question I have to put to myself: the people doing the most with these tools are often the ones saying the least about them, and I’d rather not mistake my own supply of words for a contribution.

We’ve built an enormous philosophy on top of a tool most lawmakers and philosophers, and maybe even educators in our universities, have barely used in practice.

There’s a difference between an opinion about AI and a report or a demonstration. An opinion is a position you can hold without moving. A report and a demonstration are accounts of what happened when you moved — what you tried, what worked, where it failed you, what you changed about how you work because of it. James’s point isn’t that belief is worthless. It’s that belief which has produced nothing is hard to tell apart from belief that was never really there. The works are how faith becomes legible. They are also, often, how it becomes real.

This matters most exactly where the conversation is loudest, which is the question of whether the machine is, in some sense, someone.

Maybe this is something we should say more often, and then show we mean it in the work we do. The machine is not sentient. It doesn’t want anything. It doesn’t know that you or I exist. But that turns out to be the less interesting half of the sentence. The more interesting half is that we are sentient — and we are entirely capable of treating a thing as a someone long before it is one, or when it never will be.

That’s the real danger, and it’s a danger about us, not about it. Not that the tool wakes up. That we fall asleep. That we let a system with no stake in the outcome stand in for our judgment, absorb our moral weight, and receive the kind of trust that should be reserved for things that can be held responsible. A hammer can’t betray you. But a hammer also never tempted anyone to stop deciding where to put the nail.

I’ve come to think I can’t reason my way to the line between using a tool and deferring to a counterfeit person. I can only find it by working. The person who has actually built something with these systems knows, in their hands, where the tool ends — knows what it’s uncannily good at and what it quietly gets wrong, knows which judgments it can carry and which ones it must never be allowed to carry. That knowledge isn’t a philosophical position. It’s a callus. It comes from use. We may feel a certain mutual symbiosis. You could argue that suggests sentience. But maybe it simply acknowledges there is a way to capture what we not long ago labeled data exhaust that can benefit in a continuous improvement feedback loop. Maybe it points to yet another example of how technology can become useful to the degree it extends and amplifies its users, and how users can personalize that technology through use.

Which means the most useful thing almost any of us can add to this moment isn’t another position. It’s a report. It’s a demonstration of what was made. Here’s where X or Y helped. Here’s where I caught it being confidently wrong. Here’s the judgment I kept for myself by building a human-in-the-loop action into a workflow, and why. That kind of contribution compounds — the next person can build on it. Crowdsourcing in-situ learning. Positions don’t compound. They only accumulate.

So I’d offer James’s sentence, lightly adapted, as the test for this whole moment. Show me your view of AI without your works — and I’ll show you mine by my works. Not because the doing settles the philosophy; it doesn’t. But a view that has been through the work is a different kind of thing than a view that hasn’t. It has been somewhere. It has cost something. It has been corrected by contact with how the tool actually behaves.

I think we need to share more of what we’re actually doing. Take a concrete question: does good data governance have to come before good AI, or do the two get built together as you go? That’s worth knowing. My guess is that the answer varies — sometimes one comes first, sometimes they arrive together. Or take a larger question: is a model like Claude or GPT or Gemini a good personal counsellor, or is it better understood as a remarkable hyper-spatial statistical engine, one that amplifies us and helps us keep learning in a world where there’s a great deal left to learn? Those aren’t questions to settle by hypothesis. They’re questions to prove out.

Which leaves a smaller, more practical set of questions to which I don’t have tidy answers: How do we do this in the cross-functional collective of LinkedIn, of all places — or wherever else the conversation actually happens? How do we communicate without all the sigmas and other math symbols that may be necessary in certain places, but only serve to obfuscate understanding needed to unlock potential in so many others?

The conversation we need isn’t quieter. It’s just more of it carried out in the open, in the form of what people have made and what they learned making it. Faith, James insisted, is completed by works. So, I think, is understanding.