I've been building software long enough to have sat through this announcement a few times. The web was going to change everything. Then mobile. Then the cloud. Each one arrived with the same certainty — that the ground had moved and nothing would work the way it used to.
Some of that turned out to be true. A lot of it didn't. And the part that was true was rarely the part being shouted about at the time. That gap between the noise and what actually mattered is the useful thing to carry into the current wave, so it's worth being specific about what it looked like.
The claims that aged well, and the ones that didn't
When the web arrived, the loud claim was that every business needed a website with an animated intro and a "digital strategy". Most of those sites did very little. The durable change underneath was quieter: distribution stopped depending on physical proximity, and that reshaped whole industries over the following decade.
Mobile was the same shape. The loud version was "there's an app for that" and a scramble to ship an app nobody opened twice. The structural version was that computing became something people carried everywhere, always on, and that changed what customers expected a business to be able to do.
Cloud, again. The loud version was "own nothing, everything is a service". The structural version was that capacity became elastic and cheap, so a small team could run infrastructure that used to need a server room and a specialist to look after it.
In each case the hype attached itself to the surface — the visible, sellable application. The change that actually mattered was structural, slower, and considerably less exciting to talk about at a conference.
What that pattern says about now
Agentic AI has the same two layers. The loud claims are on the surface: software that builds itself, teams replaced wholesale, businesses run end-to-end with nobody at the wheel. Some of that will look as dated in five years as a Flash intro does now.
The quieter change underneath is that the cost of certain kinds of cognitive work — research, drafting, triage, first-pass analysis — is dropping sharply. That's less dramatic than "AI runs your company", and it's the part actually worth planning around, because it changes the economics of work your business already does and pays for today.
If you only track the loud layer, you'll swing between two mistakes: buying things that don't survive contact with a real workflow, and dismissing the whole thing as noise. The structural layer is where the durable decisions live.
Where experience helps, and where it doesn't
I want to be careful here, because "I've seen this before" is also how people got the last few cycles badly wrong. The same instinct that lets you ignore a passing fad lets you dismiss a real shift. Plenty of experienced people called the web a toy and mobile a distraction, and they were confidently, expensively wrong.
So the value of having lived through prior cycles isn't a verdict. It's a filter. The question I've learned to ask is whether a claim describes a structural change in cost or capability, or whether it describes a shiny application sitting on top of one. The first tends to be durable. The second tends to be a phase, and phases are fine to wait out.
By that test, "AI writes your marketing autonomously" is a surface claim — it's an application, and a fragile one. "The cost of producing a competent first draft has collapsed" is a structural one. You can plan a business around the second. The first will either mature into something governed and useful or quietly disappear, and you lose little by waiting to find out which.
What we're doing with it
We treat the current wave as a real change to delivery economics, because by the structural test it is one. We've moved work onto it where it removes a cost we were genuinely carrying. We've held back where the claim runs ahead of what we can measure or govern, because an impressive capability we can't stand behind isn't worth much to a client.
That's not caution for its own sake. It's the same reading we'd have applied to the web or the cloud if we'd been more disciplined about it at the time — separate the structural change from the marketing around it, and act on the first.
If you're trying to decide how seriously to take any of this, I'd skip the question of whether AI changes everything. The more useful question is narrower: which specific cost in your business is actually dropping, and does that change a decision you'd otherwise be making? If the answer is yes, that's worth a conversation. If it's not, the noise can wait.
We're always happy to talk through where that line sits for a particular business — including the places we think the honest answer is "not yet".
