Subscribe
Share
Share
Embed
AI First with Adam and Andy: Inspiring Business Leaders to Make AI First Moves is a dynamic podcast focused on the unprecedented potential of AI and how business leaders can harness it to transform their companies. Each episode dives into real-world examples of AI deployments, the "holy shit" moments where AI changes everything, and the steps leaders need to take to stay ahead. It’s bold, actionable, and emphasizes the exponential acceleration of AI, inspiring CEOs to make AI-first moves before they fall behind.
Andy (00:00)
This is AI First with Adam and Andy, the show about what happens when AI stops being a tool and starts becoming part of your workforce. I'm Andy Sack, and alongside my co-host Adam Bratman, each episode we talk with business leaders on the front lines of the Agentic era. What's working, what's failing, and how AI is changing the way companies operate. No hype, just real world lessons and ideas you can put to work.
This episode came out of a debate inside Forum3 this week about how quickly the AI tool landscape is changing and whether the way companies have been thinking about enterprise AI is already outdated. Rose is joining us to help guide the conversation and bring in the practical questions we're hearing from companies trying to roll this out at scale. Rose, take it away.
Rose Kelly (00:46)
Hi guys, I'm excited to join you guys again for another very interesting conversation. So, Andy, as you mentioned during our leadership team meeting this week, we had a really healthy dialogue and there was a little bit of tension, which was I think kind of fun, about talking about tool use and like what does it look like? And you know, for the last few years we've been recommending to teams, we've been very clear, like
You want to pick one general purpose AI tool and you wanna stick with it, get your team trained on it, start building things out in it. And I think that, you know, Adam, you made a really interesting point that like we are in a different era than we were in even 60 days ago. and so I just wanted us to kind of bring that to the audience because I think it's really interesting thing to think about.
Adam, so we had this conversation during our leadership team meeting. It was super interesting. It was some really healthy debate. and I just thought we should bring that to the audience. And so if you can kind of like set the stage of like what the conversation was, I mean, we've spent the last
Adam Brotman (01:42)
Yeah.
Rose Kelly (01:43)
few years recommending to people they check one tool, and I think that we're having some questions about that. So can you kind of set the stage?
Adam Brotman (01:50)
Yeah. As if this space wasn't moving fast enough already, because we went from the chatbot era to the agent era and we've talked about that a lot. And Andy and I wrote the book about it. But when we kind of lived our life in this AI bubble that we're in, so to speak, from January until August of this year.
It was like, okay, is my company a Claude company? is it a ChatGPT company? Is it, you know, what system am I using? should I do both? Which one do recommend? but that was one kind of a framework, which is like if I go to ChatGPT or Claude, I've got usually some kind of an enterprise account, I should, like either team or business or enterprise or whatever. I'm like used to
Looking at this thing that's got like chat history, I've got connectors, I've got plugins, I've got skills, I've got projects, I've got folders. You gotta learn all this stuff, you gotta know what it means, and then I've gotta figure out my model picker and all that. That's the stuff that we train people on. And you might have a preference for Claude or ChatGPT or whatever, right? And then what changed, or at least got layered on that we were talking about is a more, I don't wanna call it more modern, because that's not quite fair, but like this new.
Sort of paradigm or framework came along with what people would call like personal agents, meaning Grokbot, Muse, Instinct. And what how are they different? There's no chat history, right? There's just like me and my agent. And my agent's got in the claude, has a computer and has tools and has a brain, which is like a frontier model. and I might.
Create
multiple these agents, or one of them, it doesn't matter. And I just have this like one conversation to rule them all, and it's like my teammate, and I just go. And what's happening is Andy and I, in this conversation, the story is we were sort of talking to our team and saying, Hey, I'm not sure. And it's literally, I'm not sure. I wonder what's easier and better for a company to use to just go get in this sort of non-chat history, non-doc.
Bells and whistles kind of version of a ChatGPT or Claude, or should they get into kind of personal agent mode? And it's like streamlined, it's easier in some ways to know how to use it. And you just kind of go. And that's the conversation we were having. And you know, they're not mutually exclusive, you can do both, but that was the story, right?
Andy (04:07)
Yeah, well, I mean, I think what we're trying to share with the audience is given the pace of development and the evolution of these new tools, specifically Grokbot, Muse, and Instinct, what's the implication of that to our advice to our customers? And we're just sharing an internal debate. So let me try and ground it with two very concrete stories. One is a customer of ours that chose
the company, Claude as the tool. And the entire team relies heavily on all of the Claude suite of tools. And they've invested in that really all year long. I don't think they're going to other tools
Unless absolutely necessary, they're not going other tools. They're staying very focused on the claude suite of products for their entire company. As a result, they've developed both individual skills and team skills. Their pace of development is faster. Everybody understands it. Like everybody knows the right same tool. And they're not lifting their head going what's going on in the marketplace and looking at other tools.
their productivity is really off the charts. And so that's really the benefits of choosing one tool.
Rose Kelly (05:16)
I would add one more thing is that from an admin perspective and like a security,
Andy (05:20)
And a cost.
Rose Kelly (05:21)
yes, from admin, there's like the cost of it. But I'd also say like it is hard being an admin of one of these systems. And for everyone who is just of like an end user of it or like an individual user of it, the admin side of it is really complicated. but it also gives you a lot of control and some sense of visibility into what's happening. And so I do think there is something to be said there in terms of
being
able to like control it and they just have one system and they have a good understanding of what everyone has access to and what people can and can't do.
Andy (05:50)
Yeah, that's two parts of a story of one company choosing one tool. And I chose Claude, but it easily could have been ChatGPT. it's just the story of a company choosing one large language model as their primary provider. Alternatively, suddenly there are these new agentic tools, which we've mentioned, and we're wondering.
Is that still the right advice? How model agnostic should we be encouraging companies to be? How easy is it to switch out of different tools? And so this all started because Adam, about three weeks ago, started mostly out of his own curiosity on a weekend, started playing with Grokbot. And he started using it for work, but also for his personal life. And
He's saying to myself and to Rose, guys, have you checked out Grokbot? Holy cow. And it took me two to three weeks and I started using Grokbot. and I'm three weeks behind Adam. I'm purposely leaving out Muse and Instinct because I haven't really digested those, but I've started using Grockbot and I am blown away both by the total change in my own personal workflow.
in the last week to ten days have developed twelve different grok bots. And so these are like digital members of my team. one as a travel agent, one as a builder. Like I've developed all these roles and it's so much fun and it's lighter and it's easier to use. And I during the AgentSync book, I was all in on Claude Cowork.
So my own personal work has changed and I'm now scratching my head going, What's the advice that I would give the one hundred, five hundred employee company about how to use agentic tools?
Rose Kelly (07:35)
Yeah, it's hard when it feels like we're giving
people one advice and then we secretly know there's these other options that are really easy. And I'd say, you know, one of the things I think we really run into with bigger clients or with most of our clients is that they have some people who can like really figure it out that are willing to get in there and they're able to figure out skills and they're getting the, you know, not maybe the most, but they're getting a lot out of it. And then you have others who are just like still just don't really know what to do. And you do think about Adam, I think you made a great point that like for those people.
Having something really simple could actually like, you know, really accelerate their use of it and like the value they'd be able to get out of so yeah, I
don't know, Adam, I feel like yeah.
Andy (08:12)
Well, and that's that that's
that that's the important thing to know about the current tools, is they're easier and more fun to use. And as a result, they're more consumer friendly. As a result, I ended up, even though I was a Claude Cowork user, I ended up turning to Grockbot to do all sorts of tasks that I never thought to ask cowork to do. And it's because
Adam Brotman (08:33)
And that's
Andy (08:34)
it's simpler.
Adam Brotman (08:34)
yes, and that's and I would say it's still a hypothesis. So that's why there's a debate or you know the tension that Rose mentioned in our team meeting, because it's like, I don't know what I'm about to say is true, but I'm gonna pick up on something Andy just said. The hypothesis that I pondered out loud and we had a good discussion about is if you just have one mega conversation going with an agent.
And that agent has all the tools it needs, all the connectors it needs, all the LLM brain power it needs. And you're just like, hey, agent, whatever I've named you. this is the goal or the output I'm trying to accomplish. Can you help me with this? let's work on this, let's go. And then it just does it. And then it comes back to you. And you're like, I verified and
inspected what you did, that was close. Can we try it again? And okay. Well, because what Andy just said is right. Because I don't have chat histories and project folders and tasks and plugins and connectors and skills and scheduled tasks and all this other shit to worry about. That I just have a conversation with it and I'm thinking like a manager, I'm thinking like a coworker with my agent and I'm just going, this is what I'm trying to accomplish. Can we do this? Yes, we can do this.
Can you do it for me and then I'll inspect the work? Yes. That's much more the way we encourage people to use AI, even with all the complicated stuff. but when you're presented with this huge dashboard of stuff, chat history and this and that and the other stuff, I think it could be overwhelming to people. Like Andy was just saying. So my hypothesis that I was pushing the team on is I don't know this, but I wonder, power users or not, if you just gave them this like,
simple tool, gave it a role and a goal as a tool, not just for a chat, but for the tool. And you're like, let's go. You have the connectors you need, you have the background you need, yes, I do, okay. I think like Andy's story is gonna become the norm that people are just gonna naturally start using it in a more proficient way. Ironically, because it is simpler in its interface. But I don't know that's the case because Rose
You and our team have done such a good job and we've done this. We've spent years now training enterprises on how to use these more complicated interfaces on ChatGPT and Claude.
Rose Kelly (10:43)
That
just keep getting more complicated every month, too.
Adam Brotman (10:46)
Yeah.
And they're more powerful and they're more complicated, but we've taught them how to use it and they're coming back to us going like guys we're transformed. This is amazing. So there's a part of me that's like having cognitive dissonance because I'm like, is it bad to like give them the simpler, you know, way to go? But I don't know. I think that's gonna be to be determined, but it's a fascinating conversation.
Andy (11:04)
Yeah, and it's particularly fascinating for the enterprise, right? Like it's easy when it's with the individual consumer, but when you're a hundred or a thousand employees, what to do? and we're giving you a glimpse behind the curtain of really our own struggles
at the forefront of trying to give advice on AI transformation and helping people make that switch. There's no question that these simpler tools I think play a role. And what that means for the enterprise and what that means for our advice, I think we'll need to check back in sixty or ninety days.
Rose Kelly (11:37)
Yeah, I think
there's something interesting too, when we say like one tool, what do we even mean? Like Adam, when I think Suhail on our team, who's like in our CTO, the most technical person
Adam Brotman (11:45)
Yeah.
Rose Kelly (11:45)
on our team, he made a really good distinction that he's like, I feel like we're pulling apart two different things. And so it's like, are we talking about the model? Are we talking about the app? Are we talking about the agent harness? Are we talking about the company interface, the end user interface? Because I think those are different things. And so I think maybe we all need to stop trying to
feel like a tool is all of those things and it has to be all those things and it has to be the same for every person on your team. And so I think since that conversation, that has really opened my eyes to thinking differently about everything that happens behind and there's like the intelligence layer and all of that. And then there is like what your, you know, all your employees are actually interfacing with doesn't, you know, there's something different maybe.
Adam Brotman (12:22)
Yeah, I think the future, and this is the part I can't tell. Ironically, we're talking about cutting edge agent stuff, which Andy and I were just talking to a friend of ours the other day that's really into AI, and he was like, You guys are living in the future. And like, and we're talking about just using these agents, right? And including like cowork and codecs.
Andy (12:42)
No, well what what had
him say that was I showed him my twelve Grokbot agents and he
Adam Brotman (12:46)
Right.
Andy (12:47)
and he was like, you know, A, he was not comfortable at his company, even though he's in the AI space and very AI knowledgeable. he doesn't do that because he doesn't think they're safe and isn't willing to do that.
Adam Brotman (12:57)
That's right. And
he's super AI proficient, but the reason he doesn't think they're safe and it's worth mentioning is because and Andy, you said this to him on the call, if you're gonna use a simpler interface like a Grokbot Muse Instinct type interface that we're talking about, you're gonna have to use connectors. Like you're gonna
Andy (13:14)
That's right.
Adam Brotman (13:14)
have to connect it. So that's in other words, that's how it works. Like
what's
Andy (13:18)
Well, and that
point goes to Rose's point, which is if you use that as your front end and then you're using connectors and you're connecting to either Claude or ChatGPT, like us getting language clear for the audience on these things is really important.
Adam Brotman (13:32)
Yeah, so we can talk about it in more detail. I know we gotta wrap this episode. I will say one thing, just to give our audience a sense of, and maybe this is too futuristic, but just to give you a sense of what we're talking about, because we might be confusing you. Imagine if you had one agent, agent meaning just like one little chat bot. It looks like a cute little chat bot to you. That's all it is to you, right? Like, okay, I don't get why is this an agent? Why is this so powerful? Imagine if it had
All the connectors to your email, to your whatever systems you normally use on your computer at work, to like you know, various SaaS programs and browser things that you normally do. imagine if it, your agent, could actually talk to other agents of yours that also had connectors, but were using different models, cheaper models, easier models. You just have one agent that you're talking to.
That's got all the right access to all the stuff you need. And you're just having a conversation with your chief of staff, let's call it. Okay. But it's like chief of staff is an understatement. And it's like I'll go talk to the connectors, I'll go talk to the other agents, and you just have a conversation with me, Adam, and we'll go get work done. And you don't have to worry about all these complicated things. and by the way, I'll keep the cost down, I'll keep you secure, and I'll make sure the work gets done. Like that's
That's sort of the future we're going to. it's not here yet because not all the connectors are safe.
Rose Kelly (14:48)
Yeah,
it's like what's the closest experience to that today? And that's the question
Adam Brotman (14:52)
Correct.
Rose Kelly (14:52)
I think.
Adam Brotman (14:53)
But it's ironic because I think we're heading towards a simpler user interface, is what Andy and I and you are talking about here, because more of the complicated heavy lifting is happening in the background. but we're not there yet. So we're sort of in this moment where like we might be going from this like complicated ChatGPT and Claude interfaces to something much simpler as the interface and all this like.
Secure but connected and computer use and all this crazy stuff we're talking about is happening it's obscured from you and yet it's more powerful. And you end up with where Andy said in his story, where you're just like, This is more fun, this is easier, and I can just sort of like work with my agent. And by the way, we wrote about this in our book. We said in our book, Agents Inc. that the future of AI work.
Is
gonna end being people just managing their agents and steering them and then verifying them, as opposed to like worrying about all the details of like what connectors and data and tasks and prompt you know engineering I need. it melts away and becomes simpler. But we're heading towards that era, we're not quite there.
Rose Kelly (15:58)
Yeah. I mean, one way to think about it is like we have been living through this with ChatGPT and Claude. And we've had to see the sausage being made. And it's not always pretty and it's not always easy versus some of these other new things that are coming in who get to kind of leapfrog all that and just give you the nice shiny thing at the end without you having to see, you know, it all being made. so I don't know. for me, I'm really hoping that.
ChatGPT and Claude are gonna come up with something similar because I do really like their enterprise systems. But we will see and I'm sure we'll come back and revisit this conversation sometime soon. Andy, you wanna wrap things up or any last things to say?
Andy (16:29)
Yeah, I mean I I
no, I think that I mean, yes, I do wanna wrap things up and I may have something last to say, which is
Rose Kelly (16:37)
I actually do
have a sorry, I have another question.
what do you really recommend to a CEO who is maybe wondering the same thing? let's say you're a 200-person company. So you're not super small, you're not as small as us, and they have ChatGPT or Claude and it's working relatively well. What do you say about these other tools?
Andy (16:54)
I think
at least today. I think today's episode highlights in a dialogue a question that the three of us have related to one tool versus many for the enterprise. And our advice historically has been choose one, evangelize it, go fast, and it doesn't really matter whether you choose ChatGPT or Anthropic.
but choose one of the two frontiers. We think that choosing one of them is probably the right thing to do. Microsoft is obviously in the mix, but choose one, place your bet and go. and have your AI council be aware of other tools, but choose one and go. I'm not ready to recommend anything other than that today, but I'm increasingly aware as a result of the introduction of these new tools, Muse, Grockbot, Instinct.
That A, I think that the large language models will be coming out with simpler agentic harnesses and interfaces that make it easier for me to engage. And I think both Meta and Grok and Instinct did a really nice job of shining that light and it's making us scratch our heads, but I don't change my advice.
Rose Kelly (18:01)
Okay. Adam?
Adam Brotman (18:02)
I agree with Andy, and I would just say last word is it used to be that Andy and I, and Andy was particularly good at this, would be talking to CEOs, and our advice a year ago, same advice today, but a year ago was you as a CEO got to use these tools. We were talking about ChatGPT and Claude. We still believe that that's the right advice. However, I think Andy and I would now say, And you might want to try Muse, you might want to try Grok bot.
As a CEO. Just on your own, just to play with it. You don't need to do anything crazy with it. Like just understand that like
Rose Kelly (18:36)
Even just for personal
use, just so you can see what the future is gonna look like.
Adam Brotman (18:39)
Correct.
We've always we've always given CEOs that advice to like roll up your shirt sleeves and use some of these things. look, don't do that instead of using ChatGPT or Claude. Like they gotta do that first. But by now, by God, I hope they are doing that, okay? and so that's the first advice. But once you're doing that, that's not enough anymore. as a CEO, you gotta like at least play with some of these tools. And they're not hard to use. they're just like texting, right?
Andy (19:02)
I think really at root, because I think Rose said it earlier, really what we're seeing in the marketplace is a battle for who's gonna own the customer, both in the enterprise and then the consumer. And you know, we haven't heard from Apple if like Siri suddenly started to work and was agentic and like you could just pick up your phone and tell Siri to do everything.
I mean that's what we're starting to see and because of the introduction of these tools, it raises questions for like I again, just using my anecdotal story, data point of one, I'm definitely AI forward, if not AI first. I work in an AI company and
I was eighty percent Claude and Claude co-work usage ninety days ago. And today I'm eighty percent Grok Bot and you know, ten percent ChatGPT and ten percent Claude. So my own personal use in work has completely shifted in the last really last 30 days, but I'm just comparing it to 90 days ago. and I think your question, Adam, go to bring us all home and I'll close it out, is
that you phrase to the entire Forum3 leadership team is gee, is this the new way of working and that we're experiencing and you posited your own hypothesis and we just thought we would share that with the audience. So thank you to the audience for letting us
Give you a glimpse behind the curtain of what's going on at the leadership meetings at Forum3 and wrestle the topic down of one tool versus many tools and how to how to think about that. Hopefully this conversation was helpful. thank you for listening to AI First with Adam and Andy. For more on how to build and run a company in the agentic era, head to Forum3.com.
You'll find case studies, research briefings, executive summaries, and our book, Agents Inc. How to Rebuild Your Company in the AI Era. We truly believe you can't overinvest in your AI learning. Onward.