Dario Amodei wrote something that I haven't been able to put down. The key challenge in an AI economy, he argued in his essay "Policy on the AI Exponential", won't be incentivising growth — it'll be finding a way for everyone to share in the benefits. He's writing about national policy. I keep thinking about it at the scale of a 40-person business.
Here's why: growth isn't the constraint anymore for most of the organisations I work with. The constraint is deciding who benefits from what you've already got — or what you're about to have, once AI makes certain kinds of output dramatically cheaper. That's a distribution question, not a growth question. And if you're running a small-to-mid-sized organisation in Canada right now, you're already making that distribution decision, whether you've named it that way or not.
The Same Question, Smaller
Amodei's frame is built for sovereign wealth funds and labour market policy. He's worried — rightly — about a world where productivity compounds but the benefit concentrates. Where GDP climbs and most people's lives don't. That's the macro version. But the same dynamic is sitting on your desk in miniature.
If you've automated some part of your workflow — customer support responses, report generation, meeting summaries, first-draft anything — you've made output cheaper. The work that used to take your team two hours now takes twenty minutes. You've just created capacity. The question is: who gets it?
Does the team member whose work got faster get their hours back? Do they take on new projects, or do you hire fewer people next quarter? Does the time savings flow to the customer as faster turnaround, or does it stay inside your walls as margin? Do you reinvest it, or does the owner pocket it?
These aren't rhetorical questions. You're answering them every week, probably by default. Most leaders I talk to haven't stopped to name the distribution decision explicitly — they're just moving fast and making the call that feels obvious in the moment. But the call you make about who benefits from AI-generated capacity inside your organisation is the same distribution question Amodei is naming at the policy level. It's just happening in your 40-person business instead of at the scale of a G7 economy.
What This Actually Looks Like
Let me be specific. If you've automated your customer support triage and your support team now handles twice the volume in the same shift — you've made a distribution choice. You kept headcount flat and captured the efficiency gain as throughput. That might be the right call. It might not be. But it's a call, and it has a winner and a non-winner baked into it.
If you're using AI to draft proposals and your business development team is suddenly clearing three RFPs a week instead of one — same question. Do they take on more prospects, work fewer hours, or does the pipeline just move faster without changing anyone's Wednesday? The answer depends on what you decide benefits from the new capacity: growth, margin, or the people doing the work.
I'm not arguing for a specific answer here. I don't think there's one right way to distribute AI-generated capacity inside every organisation. But I do think most leaders aren't naming the question explicitly, and that means the answer is getting made by inertia. The default is almost always: capture the efficiency, keep the structure the same, move on. That might work. It also might be the exact dynamic Amodei is warning about, just at a smaller scale and happening faster than anyone expects.
Here's what I know from watching this play out: the organisations that are thinking about this explicitly — who's doing different work now, who's benefiting from the speed, what trade-offs we're making — are the ones building trust with their teams. The organisations that aren't naming it are the ones where people feel the ground shifting but can't quite articulate why. That's not a good place to be six months into an adoption curve that's only going to accelerate.
Why the $350M Matters (Even If You'll Never Touch It)
This is why Anthropic's $350M Economic Futures commitment — announced the same week as Amodei's essay — matters, even though you'll never see a dollar of it. Half of that ($200M) is going to empirical research on how AI affects labour markets, what interventions work, what policy structures hold up under real-world conditions. The other half is funding fellowships for early-career Americans to work on the distribution question from inside government, nonprofits, and academia.
What that tells me: the people building the most capable AI systems in the world think the distribution question is an open research problem, not a solved one. They're putting serious money behind figuring it out at the policy level. And they're doing it because they know what happens if growth compounds and benefit doesn't follow — you get a backlash, you get regulation written in anger, you get a social contract that breaks.
The same logic holds inside your business. If the people doing the work don't see a benefit from the tools making their work faster, you're building the conditions for exactly the kind of resentment and mistrust that makes good people leave. You don't need a $350M research fund to know that. But it helps to see that the smartest people in the room are treating this as the hard part, not the easy part.
The Question Worth Asking
I don't have a framework for you. I don't think there is one yet — at least not one that works across every context. But I do think the question is worth naming explicitly: when AI makes output cheaper inside your organisation, who benefits? Your team? Your customers? Your margin? All three, in some balance you've worked out intentionally?
If you haven't named the question yet, you're still answering it. You're just answering it by default.
And default answers tend to favour whoever holds the most power in the room, which is rarely the team doing the work.
Growth is no longer the hard part. It hasn't been for a while. The hard part is deciding who gets to share in what growth produces — at the scale of a country, yes, but also at the scale of a 40-person business making a decision about who benefits from a faster workflow on a Tuesday in July. That's the question I keep coming back to. If it's yours too, I'd like to keep thinking about it with you.
This is the kind of thing I think out loud about every month in Aurora Brief — not answers, but the questions worth asking and what I'm learning as I try to figure them out. If that sounds like your kind of conversation, come read.