Your business will need an agent, just as it once needed a website

In 1998 plenty of companies decided a website was a fad. They were right for a couple of years. Then they stopped being right, and the ones who arrived late paid a price that was in nobody’s forecast: they didn’t lose a campaign, they lost the channel.

We’re at the same point with a different technology. With one important difference: this time it isn’t about having a presence. It’s about having an interlocutor.

The storefront stops being looked at

A storefront is passive. It shows what’s there and waits for someone to walk in. That has worked for thirty years because there was always a person on the other side with time to look.

Less and less. People have started asking an assistant instead of working through ten links, and they take the answer. They don’t compare — they delegate. Your website can be flawless and still not appear there. Not because it’s badly built, but because it was made to be looked at, not to be interpreted.

The more advanced companies have already reacted by adding a chatbot. But a chatbot is still a storefront, just with a shop assistant behind the glass: it answers if you ask, it’s available, and it does nothing else. It’s reactive, and we’ve spent years confusing reactive with intelligent.

From answering to acting

What’s coming is something else. An agent representing your company doesn’t wait: it has an objective and acts to meet it.

It completes a task end to end instead of explaining how to do it. It confirms an appointment. It brings up what you offer when it’s relevant. It puts forward a proposal. It flags something before anyone asks.

The difference from a chatbot isn’t technology. It’s initiative: one waits to be spoken to, the other has something to achieve.

And here’s what almost nobody is watching

Everything above still assumes there’s a person on the other end. That’s where most analyses stop, and where it starts to matter.

The scenario being built is different: agents talking to agents.

Your customer won’t fill in your form. Their agent will talk to yours. It’ll ask about availability, terms and lead times. It’ll compare against three other suppliers. It’ll close if it fits and drop you if your agent can’t answer. All without a single person having seen your website.

And this isn’t conference speculation. There are already open, named standards for exactly this: for an agent to discover what a system can do, and for agents from different organisations to find each other, negotiate capabilities and delegate work between them. They’re published and the major providers are adopting them. The infrastructure isn’t up for debate — it’s being rolled out.

Once that becomes widespread, the question for your business won’t be whether you have a website. It’ll be whether you have something to answer with when the one asking isn’t a person.

Whoever doesn’t will still exist. They just won’t be in the conversation where the decision gets made.

You’ll be delegating too

Worth looking at the other side, because it arrives sooner than you’d think.

You’ll delegate as well. An agent handling your renewals, comparing options before you have to decide, cancelling what you don’t use, dealing with your suppliers’ agents. Not a tool you open: something working while you’re not there.

In that world, the customer your business serves often won’t be a person, but a system acting on their behalf, with explicit criteria and no patience. That one isn’t won over by a good product page.

What starting now actually means

You don’t need to reinvent the company. You need to start accumulating judgement, which is the one thing you can’t buy later.

Build something small and real, in production. A bounded process that hurts a little and isn’t critical. Internal pilots teach you nothing, because nobody behaves like a customer: the learning is in what people ask once they’re off your script.

Require it to execute, not to talk. If the first version only answers questions, you’ve built a chatbot. The bar is completing a whole task.

Look at agent-to-agent protocols now, not when they’re mandatory. Not to implement them tomorrow, but to avoid making decisions today that you’ll have to undo later.

None of this is about picking the right use case. It’s about having mileage when it stops being optional. The companies with solid agents a couple of years from now won’t be the ones that decided then — they’ll be the ones who spent a while getting small things wrong.

That it already works isn’t a promise

This isn’t a five-year horizon. There are small businesses — clinics, local shops — operating this way today: the customer resolves their question and books by talking, the appointment lands in the business’s real system, and the owner finds out once it’s done. No person in the middle, no app to install.

Several of those are open and accessible in Projects, in case you’d rather see it running than have me describe it.

If the storefront stops being looked at and starts being questioned, having a storefront stops being enough.