AMR Network: Portraits | Cohere

[00:00:00]AI is really enabling [music] humans to to work with technology in a way that's in a human to human interaction, the same way that we're [music] talking.

[00:00:06]We're able to now talk to technology in natural language, and and that's really exciting.

[00:00:11]>> [music] >> So, Ryan, talk to me about [music] agentic AI. It's a very much the buzzword of the moment. People are excited about what this technology can bring, but specifically when we're looking at agentic AI and how it can transform engineering in Formula 1, where do you see the main points it can be applied to?

[00:00:31]>> I think there's a it's a really exciting opportunity for agentic AI and and F1, and just especially when you think about the engineering opportunity as you mentioned. You know, firstly, what we think about is the the vast amount of data that F1 has, right? So, there's telemetry, there's diagnostic, there's simulation. And folding that into agentic ecosystems, not just with a single agent that you're deploying and doing these things sequentially, but doing these things in parallel and having a level of orchestration that's of course led and partnered, you know, by uh with humans, uh can drive really significant outcomes because you're not just comparing one single source of data, you're now working with multiple sources and multiple experts to deliver an outcome to the human counterparts that are orchestrating this, you know, this this big pool. And of course, you know, when you think about how this can be deployed, doing this in a setting that can access your most sensitive data is incredibly important. When you think about engineering specifically, that's that's a critical piece of it because some of this data is incredibly sensitive. It could be sitting in multiple formats and multiple languages even. And having the ability to to work with that in such a way we can gain insights faster, you know, to have better precision and making decisions, you know, that can impact the performance of the team is a really great opportunity for us as partners.

[00:01:43]There's ample opportunity. There's a couple things we're working on right away which we're really excited about, but then there's a lot of opportunity down the road, uh no pun intended, you know, that's going to come up, uh and I think engineering is going to be a piece of that.

[00:01:54]>> And talk to me specifically how multi-agents work together. How do you sort of see that playing out?

[00:02:00]>> Going into 2025, the world became, you know, very obsessed with agents. And now that we're firmly in 2026, you know, that of course has scaled to the point where enterprises have deployed these things into production setting and they want to run them securely, you know, in a sovereign environment, but also have control and flexibility over the the architectural decisions that are being made. That's exactly where Cohere is fitting. We want to give customers that full view of the situation.

[00:02:27]Uh especially when we're thinking about multi-agent ecosystems, the governance layer that you're that you're that you're getting from North out of the box, which is our Intrinsic Platform, gives customers the ability to see how the agents are performing in their setting and have true traceability, auditability of the agents, you know, that are that are performing. I think also to understand, you know, to do AB testing, for example, of which model is going to work the best or which agent is actually performing and driving results.

[00:02:52]That is really where the world I think needs to go in order to scale these systems. Having a multi-agent ecosystem can just make you more effective as an organization because you're not relying on single agents to to be an expert on everything. You're giving the power of data to specific agents that are becoming experts and then synthesizing those views together. But again, those things can't be accomplished if you can't do these types of things in environments where the data is incredibly privileged and in such a way where the systems could understand it.

[00:03:20]But AI is really enabling humans to to work with technology in a way that's in a human to human interaction, the same way that we're talking. You know, we're able to now talk to technology in natural language and and that's really exciting.

[00:03:32]>> And how will this multi-agent approach change the actual role of the human engineer?

[00:03:37]>> Yeah. Yeah, it's great question. So I think about that a lot. And for us that means that we are empowering engineers to do their most productive work and to empower them to be better engineers in the in the first place, right? So, you know, by empowering them to leverage technology in such a way that's taking off taking away low value tasks and you know giving them the ability to uh to gain insights from data that they would have normally been be able to to access.

[00:04:01]Uh that has been a huge way that we've been helping engineering teams certainly at Cohere and in a partnership with Aramco, but for us, you know, the focus on sovereignty, the focus on empowering individuals, the giving them the modularity and the the flexibility to customize the technology alongside with us is also a huge boost to engineering teams because they they demand you know the ability to to have these things so they're not they don't feel like they're locked into a single ecosystem, but they feel that they have the the independent nature of of building together.

[00:04:31]>> And tell us about North, your agentic platform.

[00:04:33]>> So North, as you mentioned, is our agentic platform. What that does is it folds in our large language models, so both our generative models and our retrieval models alongside proprietary search and retrieval capabilities, which by the way search and retrieval has always been a huge part of what Cohere has done.

[00:04:48]Uh searching across multiple languages, modalities, you know, wherever the data may be sitting, whether it be in the cloud or completely on premise.

[00:04:55]North is able to deliver those capabilities into private and secure uh deployments and you know in and for regulated enterprises, for high stakes enterprises of course like Aramco, but also for the public sector. Uh what we care a lot about is serving these organizations in such a way that it provides sovereignty both at the individual level and at the enterprise level, but even at the national level.

[00:05:19]You know, these these are real considerations for organizations because when you deploy AI in such a way that it's touching mission critical information and is running mission critical workflows, you can't ever have the threat of AI being shut off, right?

[00:05:33]You need to have control and the be the ability to you know to understand these systems in such a way that's inherent and second nature to your organization.

[00:05:41]And that is what we're doing with our partners is empowering them to deploy not only the platform, but the models that underpin it.

[00:05:47]>> Data is such an important part when you speak about AI and it's so important when it comes to Formula 1 teams. How can they best use this data so they get the results that they want?

[00:05:57]>> When we think about data within organizations, within F1, there is a tremendous amount of data that's tucked away in every corner of F1 and accessing it sometimes can be a challenge because it's very sensitive. You need to keep in secure environments. Sometimes it's going to be sitting potentially in different modalities as well, or even different languages. We touched on the fact that we can do that, but you know, as AI is able to access not only just a a few sources of data, but a broader pool of data, the insights become greater, right? So it's not necessarily just about giving the teams the ability to make quick decisions based on a few sources, but it's about making sense and cross-referencing different modalities and types of data to give you insights that weren't normally surfaced to to teams. So the fact that you can do that today empowers F1 to make faster decisions, obviously leading to more precise results and driving driving performance of the team.

[00:06:49]>> Ryan, thank you so much for joining us today. [music] >> Yeah, of course. It's a pleasure.