00:04:09 · Bernardo Aceituno, Ruby Justice Thelot
Today we got into a point where for the purpose of these workflow automation tasks, they can go to, like, their latest open way model. Let it be let it be InkLink. Let it be Muse. Let it be Beam. And they can have just as good a performance as they would have expected in February this year when, for the purpose of enterprise automation, these models achieved this general intelligence capability. [00:04:30] S1: What's still not working? I think that
00:04:30 · Ruby Justice Thelot, Bernardo Aceituno
[00:04:30] S1: What's still not working? I think that [00:04:32] S0: when you look at, you know, going beyond the domains of enterprise world automation, and you look at operating in domains where, like, long term memory is needed, when you're looking at domains where very deep visual understanding is needed, which often are not pressing in the world of automation space to a level a human does, at those set points, both the open way and the frontier level models, that's not available. I think that there's a big topic on benchmarking as well. I would find that over the past three months, there's been a corrosion and the trust towards benchmarks in the enterprise just because of how much different providers benchmark their open way models to perform well in the benchmarks. And that's scary for them, because they don't have a good way to operate on it. And unfortunately, I have to rely on, you know, my favorite benchmark, is the BiTES benchmark. And at that point, you know, there's a big trade off to make into, like, do you trust the results of this benchmark, or do you, like, actually rely on a model? And I think open close frontier models and frontier labs still have that trust into what state of the art performance looks like. But I think that's gonna get closed, especially as better benchmarks and better consensus of the community is achieved.
00:05:38 · Ruby Justice Thelot
Ad[00:05:38] S1: I want to shout out our sponsor, Lovable. You know the app you've been meaning to build? The internal tool, side project or product you'd otherwise lose a weekend to? With Lovable, you get software you can ship today. That includes an editable code base you can inspect and change, plus two way GitHub sync. Lovable also handles the backend and infrastructure, including managed Postgres, auth, storage, hosting, payments, and more so you don't have to wire it all up yourself. And through Lovable's MCP server, you can create and deploy your project directly from the agents you already use. Turn ideas into software people love with Lovable, Lovable. Dev. Now onto the episode. So you're constantly talking to enterprise companies and understanding how they're trying to apply AI in their businesses. What's a trend you're noticing that isn't obvious?
00:06:27 · Bernardo Aceituno
[00:06:27] S0: That isn't obvious. Well, when I look at, like, the history of AI agents in the enterprise, I see how there's essentially three stages. I would say there's a four, but yeah, three stages. 2023, we saw the arrival of the chat assistant, just a tool that could help you brainstorm ideas, work through them and out. 2024, saw the advent of chat assistants that were augmented with retrieval, augmented with context and knowledge, and our help kind of bring the enterprise knowledge and enterprise search as a part of the AI assistant. 2025 was the year where we saw the admin of AI agents, the language models that could properly use tools, could properly orchestrate those. 2026, we look at operations where we are starting to look at trends where AI agents are actually operating outside those constraints of MCP servers and assistive work, but more towards visual interfaces and browsers. I would say computer use was one of the big unlocks of AI agents in this year. When I look at most close weight models, the fact that Rock Band music is so popular is a big part because they're becoming so good at computer use. I think open weights have gotten to a point where, for enterprise purposes, they are often reliable enough because the environments are more predictable, but I think that as that behavior keeps improving and visual understanding keeps improving, we will see more adoption in the enterprise.
00:07:44 · Ruby Justice Thelot, Bernardo Aceituno
[00:07:44] S1: When it comes to these personalized agents, you talked about GrokBot, you talked about Muse, they're still using closed source models. So are you noticing any personalized agents that are using more open weight models? [00:07:56] S0: Well, OpenClaw is probably the first one that comes to mind. [00:07:58] S1: But are enterprises using that? [00:07:59] S0: So enterprises try. I think they're very afraid of security. Like, I think that the fact that these tools are very tailored towards the consumer, make it such that some of the things that are critical towards the enterprise are not present. The key governance, residency, the key data controls, and robot access controls, want are not there. At Stack AI, we're helping them get to a level in which they can have some very personalized workflows towards the user. It's not yet at the level in which actual employees and enterprise do have a personal companion that helps them operate in their day to day beyond their AI assistant, but I think we'll get there very soon. And of course, when it gets there, the first choice will be to pick an open wave model because for them, it's a very easy trade off. They have a lot of performance that they can trust. They have absolute control on where it's hosted, price, cost, and they can easily get to a point where usage is there. I think they are very similar to Linux versus Mac and Windows to a point where probably very soon, upwards of 90% of enterprise usage of language models will be open weight.
00:07:59 · Bernardo Aceituno, Ruby Justice Thelot
[00:07:59] S0: So enterprises try. I think they're very afraid of security. Like, I think that the fact that these tools are very tailored towards the consumer, make it such that some of the things that are critical towards the enterprise are not present. The key governance, residency, the key data controls, and robot access controls, want are not there. At Stack AI, we're helping them get to a level in which they can have some very personalized workflows towards the user. It's not yet at the level in which actual employees and enterprise do have a personal companion that helps them operate in their day to day beyond their AI assistant, but I think we'll get there very soon. And of course, when it gets there, the first choice will be to pick an open wave model because for them, it's a very easy trade off. They have a lot of performance that they can trust. They have absolute control on where it's hosted, price, cost, and they can easily get to a point where usage is there. I think they are very similar to Linux versus Mac and Windows to a point where probably very soon, upwards of 90% of enterprise usage of language models will be open weight. [00:08:57] S1: Yeah. What I'm noticing is that, of course, people care about open weight models. They say they care about privacy and token cost in theory. But then when it comes down to it, it's still really hard for a lot of these enterprises to actually integrate AI into their processes. And so they end up just going for whatever tools are easiest to use despite it may be being closed or open.