Conversational First: the next shift in how we use software
Every decade, something fundamental shifts in how humans use software. We are at the next shift.
I keep watching the same move. A company takes the software it already sells, bolts an AI assistant onto the side of it, and calls that the future of the product. I understand the instinct. I also think it repeats the mistake the industry made with mobile, and the way to see that is to back up and ask what actually changes when software changes.
Every decade or so the assumption underneath the interface shifts. Not a feature and not a faster chip, the assumption itself about what an interface is. In the 1980s it was the command line, and you learned the machine’s language. In the 1990s it was the GUI: point, click, drag. In the 2010s mobile changed it again. We’re at the next one now, and most of the industry is about to repeat the mistake it made with the last one.
The AI-first trap
The current narrative is “AI-first.” Every SaaS company is racing to bolt an AI assistant onto the product it already sells. Xero has “Just Ask Xero,” Salesforce has Einstein, Notion has Notion AI, and in each case the assistant sits in a sidebar beside the interface it was added to.
This is the responsive design of the AI era. A chatbot in the corner of your accounting software isn’t a new way to work. It wraps the same old interface in a new technology and leaves the work feeling exactly as it did before.
What Conversational First actually means
Conversational First is a design philosophy, not a technology choice. It starts from the assumption that the primary interface is a conversation, and that visual elements exist to support that conversation, not the other way around.
AI-first accounting: you open a dashboard, navigate menus, and there’s a chatbot in the corner to help you find the right screen.
Conversational First accounting: you open the app. The system says: “Twelve new transactions came in since yesterday. Gamma, 340 euros, looks like workshop supplies based on your usual pattern. Want me to book it there?” You say yes. Done. No menus, no dropdowns, no codes to look up.
The conversation is the spine. Everything else is a rib.
Why this changes software economics
Traditional SaaS is built on a hidden assumption: complexity is a moat. The more features, the higher the switching cost.
When the interface becomes a conversation, switching cost evaporates. There’s nothing to learn. The user’s relationship is with the conversation, not with the navigation, not with the button placement.
This means incumbents face a classic innovator’s dilemma. Their entire business model depends on interface complexity as lock-in. They can’t go Conversational First without destroying the very thing that keeps their customers paying.
The infrastructure gap underneath
The models aren’t the bottleneck. The bottleneck is memory.
A human bookkeeper’s value isn’t knowing accounting rules. It’s that after a year, they know your business. They know that Toolstation orders go to workshop maintenance. They know your Q1 cash flow is tight because annual subscriptions renew in January.
Current AI systems can’t do this. Every conversation starts from zero. This is the persistent context problem: the infrastructure gap between “AI chatbot bolted onto software” and “genuine Conversational First experience.”
This is the 4G of conversational software. The shift is technically possible now, but it’ll take purpose-built memory infrastructure to work at scale. Whoever builds that layer enables more than a single product. They enable the whole shift.
Where the shift begins
Big changes never start in the center. The first Conversational First applications will emerge where existing software is terrible and users aren’t technical: small businesses drowning in Exact Online, freelancers who hate their invoicing tool.
These users don’t want better software. They want no software at all, just something they can talk to that handles it.
That is a different test than the last thirty years rewarded. It favors newcomers over incumbents, conversational depth over feature breadth. The incumbents will see it coming and still struggle to follow, because following means dismantling the complexity they sell. The opening is for whoever builds the memory layer first.