Small Teams Have the Most to Gain
There's a popular story that AI is a big-company game. You need a platform team, a fleet of agents, a budget line, a head of AI. That's backwards. The organizations with the most to gain from this are the small ones, and I'm watching it happen up close.
When you're small, every person is already doing three jobs. The constraint was never ambition, it was hours. AI doesn't make a small team act big. It makes a small team stop drowning.
Leverage, not headcount
The old reflex, when there's too much work, is to hire. The new question is better: what would this look like if an agent were already on the team?
We ask that before we ask for more people. Most of the time the answer isn't "replace someone." It's "stop making a person do the soul-crushing part." The report nobody wanted to assemble. The data nobody wanted to reconcile. The first draft nobody wanted to start from a blank page. That's the work machines were made to munch through, and handing it over frees our small group to spend its scarce attention on the things only we can do.
Leverage is the whole point of being small. AI is the biggest lever we've been handed in a decade.
What we actually automate
Not the interesting decisions. The repetitive ones. The glue work between tools. The summarizing, the routing, the first pass at a thing that a human then sharpens. We give an agent a real account, real access, and a clearly-scoped job, then we review the output like we'd review anyone's.
The trick isn't a clever prompt. It's choosing tasks where being roughly right fast beats being perfectly right slow, and keeping a human on the parts where wrong is expensive.
On the "slop" panic
I hear the worry that this floods the world with low-quality output. It's a fair worry. But let's be honest about the baseline. Plenty of human-made work was already sludge: the box-checking, the filler, the report written so a process could say it was written. If a machine absorbs that drudgery and a person gets their afternoon back to do something that matters, that's a trade I'll take.
Slop isn't a property of who made it. It's a property of whether anyone cared. We care. That doesn't change because the first draft came from a model.
The standard doesn't move
This is the part I keep coming back to with my team. The tools got radically better at producing things. They did not get better at deciding what's worth producing, or at noticing when something is quietly, confidently wrong. That's still on us, and in a small shop there's nowhere to hide a bad call.
So we go all in on the leverage and we hold the line on the judgment. Use the machine for everything it's good at. Own every outcome as if you'd typed it yourself. Because as far as our customers are concerned, you did.
If you're running something small and you're waiting for permission to start, this is it. You don't need a strategy deck. You need one boring, annoying task and the willingness to hand it off today.