Something genuinely interesting is happening.
AI agents are starting to mediate how people find products, services, expertise. Not just answering questions — actually recommending, comparing, deciding. McKinsey projects this could reach a trillion dollars in commerce within a few years. The funnel that marketers have optimized for decades — awareness, consideration, decision — is collapsing into a single moment: the agent picks you, or it doesn’t.
This is exciting if you like watching systems change. And it raises an obvious question: what do you do about it?
The instinct is to optimize. Learn the new rules. Get your data structured so agents can read it. Build the machine-readable credentials. Race to be visible to the new intermediary before your competitors figure it out.
That instinct isn’t wrong, exactly. But it’s worth pausing to ask the product question: what’s the smallest bet that could actually work here?
I’ve been thinking about this for my own practice. Where’s the leverage point? What’s the minimum move that creates disproportionate value?
The obvious answer is: optimize everything for agents. Schema.org markup, llms.txt files, structured capability statements, the whole apparatus of making yourself machine-readable.
But when I actually look at what agents do today — they’re still mostly confabulating. They don’t have outcome data. They can’t verify credentials. They produce confident-sounding recommendations based on training data and surface patterns. The infrastructure for genuinely useful agent recommendations doesn’t exist yet.
So optimizing heavily for current agents is optimizing for a half-built system. The rules will change. The signals that matter will evolve. Heavy investment now might mean re-investment later when the real patterns emerge.
What’s the smaller bet?
Here’s what I landed on: the things that will matter to agents eventually are the same things that matter to humans now.
Clarity about what you actually offer. Genuine quality that survives scrutiny. Work that speaks for itself. Relationships that don’t depend on which intermediary is trending this year.
Agents will get better. They’ll learn to verify claims, check outcomes, distinguish substance from positioning. When that happens, the signal they’ll reward is the same signal that’s always mattered: being actually good at what you do, with evidence to show for it.
So the smallest bet isn’t “optimize for agents.” It’s “be clear and be good” — and let the clarity be machine-readable as a side effect, not as the goal.
This reframes the whole thing. Instead of racing to adapt to a new system, you’re investing in fundamentals that work regardless of which system is ascendant.
There’s a judo principle here. The instinct when facing a stronger force is to push back harder. But the art teaches something different: don’t oppose force directly. Find the point where a small shift unlocks disproportionate movement.
The agent-mediated future is a stronger force. You can push against it — optimizing frantically, chasing every new signal, rebuilding your presence for each platform shift. Or you can find the leverage point: the investment that pays off whether agents dominate or don’t, whether the rules stabilize or keep changing.
That leverage point is the work itself.
Not the positioning of the work. Not the optimization of the work for discovery. The work.
I did add some structure to my site this week — markup that clarifies who I am and what I do. But I added it because articulating that clearly is useful for humans too. The machine-readability is a byproduct.
What I didn’t do is optimize. I didn’t chase the current signals. I didn’t build for today’s agents, which are still unreliable, still forming, still likely to change.
Instead, I’m staying ready. Watching, learning, keeping options open. And putting the energy where it compounds: into work that demonstrates how I think, relationships with people who’ve experienced that thinking, and the slow accumulation of evidence that I can help with the problems I claim to understand.
The agents will catch up. When they do, they’ll find either a trail of optimization tactics or a body of genuine work.
I know which one I’d rather they find.*
* The projections are real: McKinsey estimates up to $1 trillion in US agentic commerce by 2030. The shift is happening. The question isn’t whether to respond — it’s where the leverage actually is.