Inside Jensen Huang’s Son-in-Law’s Rise at Nvidia, Anthropic’s Sizable IPO Charity Bill

Akash Pasricha (0:13)

Welcome everyone to The Information's TITV. My name is Akash Pasricha. It is Monday, October 5th. Today on the show, we are talking to Cohere's Chief AI Officer about the company's new enterprise AI agent platform.

President Trump has also finally appointed a new AI czar. We'll talk to our tech and politics reporter about what the new appointment means. We're also going to unpack a new profile on Nico Caprez, who recently married Jensen Huang's daughter and has quietly become one of the Nvidia CEO's most trusted executives. And to close out the show, we're going to take a look at the extraordinary wealth that philanthropies and nonprofits are about to run into as a result of Anthropic's eminent IPO. It's going to be a great show, so let's get right on into it.

Cohere unveiled a new version of its enterprise AI agents platform, North, or what the company is calling North 2 The new product will let companies automate tasks with agents while giving them strict guardrails. I want to bring on Joelle Pineau, Chief AI Officer at Cohere for a conversation. Joelle, welcome to the show. It's great to have you here.

Joelle Pineau (1:23)

Thank you. Great to be here.

Akash Pasricha (1:25)

So walk me through what North 2 is. You just unveiled it this morning. What should we be paying attention to?

Joelle Pineau (1:32)

Well, really, North is an agentic platform that's designed for enterprise governments, really enterprises that both want state-of-the-art technology in terms of their models, in terms of ability to build and deploy agents, but also want the ability to really govern the behavior of these agents, whether it's from a cost perspective or whether from a security perspective. So we offer both of that and as well as support for all their use cases and so on, whether it's connectors and ability to do all of the work that you need to do and take the full advantage of AI.

We talk a lot about the models, but often we forget, like, what does it take to actually make that productive and secure in an enterprise context?

Akash Pasricha (2:17)

And so when you guys say agents, I mean, just put it into tactical terms for us. This is like enterprise search, looking through my data. This is automating employee onboarding function. What is this exactly?

Joelle Pineau (2:32)

Well, it's really moving from just the chat loop, which we still support, of course.

A year ago, North One was really designed around that experience. Now, combinating all of that, it's the ability to, yes, pull in the information that's needed, pull out the information you use in doing your work, and turning that into productive work, whether it's doing deep analysis, whether it's producing reports, whether it's actually congregating across several pieces of work to produce a new artifact, whether it's coding abilities, and so on and so forth. We have different use cases that we identify depending on people's work profile, departments, and so on. But it's really the ability to do work alongside the human workers.

Akash Pasricha (3:16)

Right. When you think about the competitive environment for these types of tools, I mean, you've got Microsoft with their Co-Pilot Suite. I think they've also got one called Autopilot now.

There's Google with their enterprise platform, Glean. Another company we've covered on the show with their search focus tools. How do you differentiate yourself from all those competitors?

Joelle Pineau (3:41)

Right now, it's still very early days. What we find is a lot of companies who work with us are really interested in having very strong security guarantees. Many of them want to deploy completely on-premise air-gapped.

Some of the solutions that run on cloud are really not an option for them. The fact that we build our own models, but we have a harness that's model agnostic, means that we both have that expertise and we can produce models that may be the best for their use cases. But there's flexibility if they want to use a third-party model. They still get the governance layer that we provide in terms of security and cost efficiency.

Akash Pasricha (4:20)

Is that how we should think about Cohere? Because I've been trying to figure out, okay, Cohere once used to talk a lot about its own models, maybe two and a half years ago when we used to cover them.

It seems like the company has transitioned a bit into this conversation around security, sovereign AI, privacy, owning your own stack. I mean, should we be thinking about Cohere as the own your own stack player in the AI lab race now? Or where do you see yourself?

Joelle Pineau (4:55)

You should definitely think of Cohere as a company that builds models. That is part of our DNA. This is where the company came from and still to this day, yet very important part of our technical work. But as the company grows and matures, yes, the goal is to provide the best enterprise platform for AI that provides security, cost efficiency, that really builds what companies need. In contrast to most of the other AI players that are really in the consumer space.