Being a co-founder has its perks. My favorite is probably how often I’m able to trick smart people into talking to me about their work.

Case in point: I recently sat down to chat with Danny, who leads IT at Cursor, and Johnny, one of their systems engineers. Cursor is, obviously, one of the hottest tech companies in the world. And they’re building one of the best AI coding tools at a speed and scale that’s almost unheard of. 

Their IT team has remained small, even as they’ve scaled to support well over a thousand employees.

We covered a lot in our chat: offices spun up overnight, a Wi-Fi crisis, a playbook built in an hour. It was a fascinating conversation, and I encourage you to give it a listen. 

Near the end, Danny said something that captures our product thesis pretty well:

"AI models are obviously going to become far more capable. But I really do believe that the truth lies in the context."

Of course, I agree. It’s why we built Consoles the way we have. And it’s why Console doesn’t need to wait for improvements from frontier models to get better. 

The context alone is transformative. 

The requests within every request

There's an assumption baked into a lot of the excitement about AI right now: that the intelligence of the model is the thing that determines whether an agent is useful. 

It isn't. 

Frontier models are extraordinary and getting better extraordinarily fast. But the truth is, raw reasoning resolves almost nothing in a real IT environment.

An IT request always carries a hidden clause. "Answer this question" really means "answer this question for this person, who is on this device, in this office, with these access permissions, inside this org's identity structure, given how we've decided to do things here."

Strip that context away and even the smartest model is just guessing. Add it back and a fairly ordinary request, "my Wi-Fi keeps dropping," becomes something an agent can actually resolve: pull the device, pull the user, pull the network logs, recognize the pattern, and respond in the team's own voice.

As is so often the case, simplicity belies complexity.

Why not just build it yourself?

So, if it’s so important, why not just build it internally? I love pointing at Cursor, because if there's a company on earth that could roll out a serviceable IT agent over a weekend, it’s them. 

So I asked Danny and Johnny directly: you have Cursor, you have the talent, why not try to build your own Console?

The answer is, again, in the context. Building that context from scratch would have taken months. And months are the one thing a small IT team doesn’t have.

"There are tools that you should build. And then there are some that you should let others build, if they do it really well." Danny said.

In other words (or at least my words), you don’t need to reinvent the wheel. 

Or, as Johnny puts it, “Console already has all the context it needs." 

Where this is all going

We recently talked with a range of leaders about how the IT function is changing, and the throughline was that IT is moving from the back of the house to the front. For a long time the job was defined by response: something breaks, someone escalates, a person triages. 

AI is dissolving that tier-one and tier-two work faster than almost anywhere else in the org.

The same people once boxed into manual ticketing work are finally getting a seat at the table. IT has always been the shepherd of an organization's tools. Now, they’re becoming the infrastructure that can decides how the whole company operates.

Models will keep getting better; that's the safest prediction in technology. 

But capability alone was never the bottleneck for internal support, and it won't be what separates the winners from everyone else. 

That's the bet we're making at Console. And anytime I talk to the brilliant IT teams at companies like Cursor, I'm reminded it's the right one.