The Roadmap Is the Ceiling
Jack Dorsey has been on a public thinking tour lately. This clip - around the 12-minute mark of a longer interview - is one of the better ten minutes of CEO-level thinking on what AI actually changes about how a company works.
The line that I noticed:
“A company’s ultimate limiting factor is its own roadmap.”
The roadmap has always been framed as the plan (wrongly in my view) and what I hear is him saying it’s actually the constraint. The ceiling. The hard limit on what you can deliver.
He describes the four emergent layers:
- Capabilities are the primitives - the things your company can actually do.
- Interfaces are the products built on those capabilities: specific screens, specific flows, specific navigations - which are someone’s best guess at what customers need, baked in code.
- Proactive intelligence: using the signals you already have (money moves are his example - “you can lie about literally everything, but when a transaction occurs, that tells the truth”) to act before customers ask.
- The world model: the deep understanding of customers and company that makes all of it coherent.
Today, most software companies live in layer two. The interface is the roadmap. The navigation is the roadmap. The features are the roadmap.
Dorsey’s argument is that layer two collapses. Customers will expect to ask for a feature that doesn’t exist and have it served to them - composed in real time from capabilities.
If the system can’t do it, that GAP is really what consititutes your roadmap. “Written” or perhaps conceived by the customer, in real time, via the “conversation” they’re having with your product.
What survives? Judgment. Is the output aligned with what we intend to build? Is it on-values, on-taste, unique? That’s the job for PM but as we’re seeing, the roles are switching around. Everyone is PM now. Everyone needs to think about this.
In this world - the roadmap doesn’t disappear but I do think it becomes more like a “forecast” - a statement of aspiration and outlook, and less like a schedule. There was a day (maybe still today?) when a lot of companies WOULD NOT publish a roadmap, just for this reason. That it really didnt represent a plan, it was more like a “window” into the future. And customers - in enterprise at least, who live in the IT projects world - would routinely misinterpret.
I remember talking to a large customer (CIO) a few years ago who raked me over the coals - could not understand why he had his “roadmap” planned for 3 years out, but I did not. The only explanation I could give - which he didnt love was - do you really want me building the wrong thing, when we (and you) know it to be wrong?
I think a lot of this shift is positive - and will lay bare some of the structural problems we have in enterprise software - because it enables a shift from “we decided what customers need and we are going to build it whether they need it or not” to “customers are telling us in real time, and we are adapting in real time, based on our accumulated judgment.”
That’s a different kind of product leadership. In some ways though, the original spirit of agile, writ large.