Most leadership has no idea what most of their team does all day. Folks come to us and say, we want AI in our business. And I say, where? And they say everywhere. And I I remind them that it's not some magical silver bullet that just, you know, pierces through every workflow and every piece of data and whatnot.
SPEAKER_01How do I know when a company is ready for this multiplayer, this shared context environment?
SPEAKER_02Well, everyone is technically ready for it. Challenging part is how do you want to apply AI to it? AI, if it's done well, should kind of be invisible in the work. Because half the time people think they need AI, and the answer is just, hey, this process and this workflow was created 20 years ago. That person has since left, and no one has ever thought to question it. And something that takes us a week and a half to do across five people and 90 emails could actually be done by two people in two days with four emails.
SPEAKER_00With or without AI.
SPEAKER_02Exactly. And they're like they got, you know, 20 hours a week for five people back just by reinventing the process, regardless of AI. Well, what about data and permissions and you know all that fun stuff?
SPEAKER_01Yeah. Hey everybody, and welcome to another episode of Using AI at work. My name is Chris Daigle, and I'm the host of the show. And today we're going to be talking about something that I think is probably interesting to every single business out there. And that is how do I know when I need to take AI from individual licenses into exploring perhaps a deeper approach to it? Our guest today is going to introduce us to this concept of single player AI and multiplayer AI and what that means for your business and help you kind of figure out which one of those that you're working with if you're even using AI at this point. Our guest today is Justin Watt. He's the co-founder and CEO of Switchboard, and that's a company building custom internal AI tools, softwares, and automated workflows for businesses. Now, with a background like Justin's, where he's had operations, process improvement, automation, systems integrations. With prior experience at MetaLab and IBM, we've got the right person on the call today or on the video if you're watching this today. And their lane really with Switchboard is implementation rather than spending time or even wasting time on long-term AI strategies. So, Justin, welcome to the show. Thank you so much for taking some time out of your day for uh sharing your wisdom with us.
SPEAKER_02It goes both ways. We'll see if it's wisdom, but thanks for having me, Chris.
SPEAKER_01As a listener, I want you to know that I see it all the time. We work with businesses, we talk to businesses that are uh trying to answer the question, what are we going to do about AI, right? And a lot of the leaders have already bought that Chat GPT or that Claude license or something like that. But they're still asking, wait a minute, why hasn't this transformed how the company operates? And I think that we're about to get a clear answer on that. So Justin, I want to jump into this concept that you mentioned of single player mode. What does that mean? And what does it look like inside a typical company that you're talking to on a daily basis?
SPEAKER_02Yeah, I think we're, I mean, what we see is often folks come to us and say, we want AI in our business. And I say, Yeah, where? And they say everywhere. And I I remind them that it's not some magical silver bullet that just, you know, pierces through every workflow and every piece of data and whatnot. And so I think understandably, people are confused because the world was introduced to AI through the lens of a chatbot that they use, and it's progressed over time and the models have gotten better, and you can integrate it with your tools. But it's still fundamentally that single-player experience, just like a video game where you know Mario or Zelda or whatever it might be, you you go through a story by yourself and you've you meet these predetermined characters versus you know multiplayer uh for folks who play or have kids who play, you know, Roblox, Fortnite, Call of Duty, those those things where you're on an internet connection and it's you and a hundred other people in the same environment doing stuff together. And that's the fundamental difference I think that people are understandably having a hard time grasping is when you're in a chat bot and you say, look at my email and my calendar and my SharePoint for XYZ information and tell me what to do. That's very different than let's take uh your average sales process. You know, you've got probably someone in business development, but you've also probably got someone in you know, service delivery, if it's a professional service firm or product, and you've got legal, finance, you know, whatever it might be is involved. And the sales process is an entire process, but there's many different parts and pieces and component parts like the actual workflow, but also the people involved. And so if you say spin up the sales process in your chatbot, it's only going to have the information for you to use and for you to access, but that's not how businesses work and that's not how teams work. So I I won't ramble on too much to answer your question. Like that's the fundamental difference that people are struggling with.
SPEAKER_01Well, I think that's um something that I've uh an awareness that I'm making now is that when we work with companies, a lot of times we we do introduce those connectors so that they have a more bespoke experience with the models. Um but unless collectively everything's being shared in that SharePoint or Google Drive or you know, whatever that that server environment is, they're only able to access the the models are only able to support them based on what it has access to, which is my stuff. Well, maybe somebody handed that lead off or talking about sales or whatever. Um I don't know what that that unless that's somewhere I can access it, I don't know what that initial conversation looked like. I don't know what kind of rapport was built and those sorts of things. So um yeah, yeah. And I would imagine that most companies that I'm seeing and speaking to, uh they haven't even made it to the individual connector stage yet. So they're a long way from this multiplayer environment. So let me ask then the obvious uh next question would be how do I know when a company is ready for uh multi this multiplayer, this shared context environment?
SPEAKER_02Well, everyone is technically ready for it as in they already do work together as teams, they're already collaborating and whatnot. I think the challenging part is re-orienting or reimagining, not to be hyperbolic, but like you kind of have to reimagine how you work. If you overly simplify a business as an assembly line and think about, again, maybe a sales process or a customer support request, it's usually an assembly line of things that happen. It's not, you know, one person does one thing and it's done. It, you know, that sales process, someone has a call and then they do a proposal and then a quote, and then there's negotiation and handing it off to a team to kick things off, whatever it might be. And so are teams ready for it on the technical side and the process side? Probably not, but conceptually the work is there, it's just how do you want to apply AI to it? And so that that comes back to that kind of the silver bullet that people are looking for with AI. AI, if it's done well, should kind of be invisible in the work, it should help augment teams, it should help um do certain parts of a workflow for people. So if you go back to that assembly line analogy, it's not that AI gets applied to every step, it's probably that AI helps in step two, step four, step nine, step twelve, and it does parts of it for you, and you start to automate. So that's kind of the way I encourage folks to think about it is what is the quote unquote assembly line of process or workflows in your business and where can AI help in those? Instead of just, well, are we even ready? Probably not. We're not even gonna care about it. We're too scared to start.
SPEAKER_01So the reality is if we've got some well-defined processes and we know the handoffs and the constraints and the the you know all that stuff, that we might be ready for this multiplayer AI injection into that process.
SPEAKER_02Yeah, and then it becomes a technical and process decision of well, which processes have AI, and then technically how do we weave it in? And that's where I think you know folks, again, understandably, I'm empathetic to it, um, but understandably get lost in the sauce of, well, what about data and permissions and you know all that fun stuff, which we can get into, but there's a lot of component words to think about.
SPEAKER_01That does sound fun. So, okay, as I'm thinking about this, uh, you know, through the lens of the listener, they are experimenting with AI. Maybe they've got a couple of power users, I think most companies that have it, they've identified two to five percent that are the power user. Um they're ready to take it further because that two to five percent is certainly having a uh an impact on their own individual roles, but they may not necessarily be uh moving the needle uh enterprise-wide or you know company-wide. So what should a CEO expect to change when AI would move from that individual productivity environment into the shared workflows? Like what is that you you you mentioned that there's some technical considerations, you mentioned that there's some uh data governance uh considerations, but what does the what is the investment of thought or time or expertise into preparing my stick with the sales process, preparing my sales process for this introduction of AI along some or all of that process?
SPEAKER_02Yeah, I think they need to think about I mean, first off, the question is always what is what are we expecting the outcome to be? I know it's an overly simple question, but most people I find that I chat with aren't even thinking about it. They just are caught up in the hyperbole hype around AI and just say, we just want AI. And I say, Well, what would it solve for you? What would be different a year from now if you add it in all the places you want it? What would be different? And often the the answer or the ones that resonate the most for me and that I've seen firsthand where everyone gets excited about it. It's not this job doomerism of well, we want AI so that we can replace all the humans. More often it is we have a huge amount of people doing a huge amount of busy work and the admin work. Like we all grew up with hope of astronaut, doctor, lawyer, or not moving spreadsheet cells around all day and updating CRM entries. And so A, are people getting time back for more important work, which can be hard to quantify. Where I do see um C-suite quantifying it more is well, we've increased revenue 20%, whether or not AI helped with that, but we didn't have to hire, we thought we'd have to hire four more people. We don't have to. And so you can literally start to feel it with your growth not equating to headcount um needing to be hired alongside it.
SPEAKER_01So, you know, when we're talking to companies, uh there's usually three main constraints that we hear. We'd love to use AI more, but we don't know where to use it. We'd love to AI use AI more, but we're not quite sure of the risk, so we're not ready to go all in. And we'd love to use AI more, but we don't have that person that we don't have Justin on our team, or we don't have Chris on our team. So let's start with that that first thing. Don't know where to use it. What is the first company-wide workflow you would usually connect, or what's the and or if they're a different answer, what are people usually asking for?
SPEAKER_02Uh what's fascinating is they're often asking for how can you help our revenue grow? And that is usually not a good place to start because AI kind of sucks at it. Like the AI outbound SDR or you know, doing out everyone can feel it, and we all get 90 emails a day with flop that we just archive. So the thing I always encourage folks to look at is like your internal operations is probably there's probably a huge amount of tax that you're paying on your business where your margins can be a lot better, but you're wasting a lot of time doing busy work. So often where um where we'll look first is where in the revenue generating parts of the business, where do things slow down? So a good example is we uh we worked with a logistics company recently that they do all of these quotes and proposals for the logistics work that they do. Sure. But there was no source of truth for rate sheets. And so they had I think it was eight account managers, all had their own version of rate sheets, they all had their own kind of system for doing it. Yeah. So they said, we want AI to help with our proposals and our quotes. And we said, How? Like you you have you have the same job, but you have nine different rate sheets for it. So how would an AI figure that out? But we went through the exercise of okay, let's get it all together in one place. It ended up being in one spreadsheet, their rate sheet was equivalent to 14,000 rows in a spreadsheet. And so we we that's our part to play, is to challenge them to say, well, how could this be 40 rows instead of 14,000? And so taking that time to bring that down, then it was very easy to apply AI to it where their quote request came in, AI would look at the request and say, What are the the things that they're looking for? Look at their rate sheet and say these are the things that they're you know, it was what it corresponds to in our services, and it would create a first draft of of the quote. And so these eight account managers went from spending about half their week, so 20 hours a week times eight people, 160 hours a week that people were spending doing quotes, and it went down to about three hours a week. Because they're you know, we're not taking the humans out of the loop to review them, but it's only you know 10 minutes here, 20 minutes there to to correct it. So it went from 160 hours a week to three times eight is 24 hours a week, roughly. And that team, it's not like they pared down that team and said, great, we don't need them anymore. Team, their revenue started to grow because that team had way more time to invest in customer relationships, actually closing the work, et cetera, not the busy work of generating quotes and proposals.
SPEAKER_01So I want you to think about this as a listener. Um, you know, hey, we want AI in this department. Well, great. Just like Justin said, how and where. So on that where, before you ever invest in the AI person or whatever, I would suggest that you sit down with the person that owns that process and say, turn on the recorder, the voice memo, the native app on your iPhone or something like that, or get them on Zoom or Loom and have them record what they're doing. Say, okay, and then what? And then what do you do when that email comes in? Okay, and where do you put that file? And really get granular with that. And it's not sexy, it's not AI magic, it's not the shiny stuff, but the AI magic is gonna fail and you're gonna go, ah, we tried AI, it didn't work, if you don't do this step. So think about not only where you want that impact, but understand you don't just reach out to Justin necessarily and say, hey, hey, do the thing. He's gonna he's gonna ask you the same questions I just talked about. Do you agree with that, Justin?
SPEAKER_02Yeah, the one thing I would add too is I'm I'm finding a trend chatting with execs where they feel like they have to have all the answers for their team. And so the thing that I encourage uh many folks in leadership to do is ask your team. Like it can be a simple analogous question to not set their alarm bells off because a lot of people hear about AI and they think, oh, what about my job? But going into your team and just saying if you could wave a magic wand and have a uh a robot helping you tomorrow, do like what parts of your job do you think that they could handle? Or what are sometimes it's uh a fascinating outcome has been they will ask them what are the least fulfilling parts of your job? Like what are the things you know we need to do as a business, but you hate doing in your job? And if you ask 20, 30 people in the company, you will start to see trends and you will probably have the answers given to you on where to focus.
SPEAKER_01Yep. It's uncanny. We do something very similar to that. And it's what do you not like doing and you're not good at as part of your job? And let's kind of focus on that area because you know, from a change management perspective, you mentioned it. Like if I come in and say, hey, we're gonna have AI start to do some of this stuff, people are gonna go, like, okay, who's losing their seat, right? Um, but if we can come in and we can introduce AI into things that they don't like doing and that they're not good at, that starts working, they start looking around going, hey, this AI stuff's pretty good. Let's do more. Or it's it it's at least not as intimidating. Do you do you have that experience as well?
SPEAKER_02Oh, 100%. I was actually just chatting with uh someone who is a managing partner at a P firm earlier today, and he was bringing up like, How do we how do we get people to start thinking about this better? I was like, just start small, like pick one thing that everyone would see value in, tell everyone about it. And it's been wild to watch. It's almost like um I think most people have seen the matrix, so hopefully this analogy lands of like when he starts to see the real world or the the world in code and the green kind of code, and it feels like the same kind of thing of you look at a business and go, where do I use AI? And then once you use it in one, two, three places, everyone starts to look at everything. Why don't we apply it there? Why don't we apply it there? Oh, why don't we think of that sooner? It just starts to become second nature. Kind of like, you know, it's funny to think that we use typewriters and had no email at a certain point in time, and now can you imagine looking at a business from that that lens? And I think it's just an evolution of this is how we will run our business, and we have to see it work in one or two places to understand how it could work everywhere.
SPEAKER_01Yeah, yeah. We're seeing the same thing. We we actually call that phenomenon when somebody like they do something and they're like, oh, wait a minute, maybe AI could help with this and this and this. Like we I I use those words to say the matrix unfolds for them. And uh we call that they they start thinking in AI. They start any business issue that that comes up or or an opportunity or whatever, they don't just use brute force. They go, Oh, how could AI help me with this, right? So you we're we're spot on, we're seeing the same thing. So let me move on to this question about, okay, great. And you mentioned it earlier. We where do you want AI? Oh, well, everywhere. Okay. Well, they got to start somewhere, the journey of a thousand miles, right? So who is the best person in the organization? Is it the the I mean let's say we start with operations. Is it just like you said, it's that frontline user where, because you know, as a chief AI officer, we bring stuff into companies, and we've worked with companies where it's top-down. Hey, everything that that maps to strategy, that's what we want to focus on. But when you get down to the actual users, they're like, that's easy, right? But like that's not really the thing. But if you just start with the the frontline users, they may have something that doesn't necessarily map to a strategic target or doesn't necessarily map to what's important to the leadership. So who is the best person in the organization to make that determination on this is where we're gonna start, this is the process?
SPEAKER_02To put it this way, we've we've turned down a few folks working with them because they have said to us, oh, you only need to talk to leadership and we'll we'll help you figure out what needs to be done. And we always say no. Like that's that's a that's a quick path to a disappointing outcome or no outcome realistically. Because to your point, at the end of the day, most leadership has no idea what most of their team does all day. And it's not a reflection that they're doing a bad job, it's more so once you get to a certain size, how could you know? And so we always say pick the challenging or problem child department or the department that you feel moves slowest, and then we will go talk to the individual contributors, or if they want to do it themselves. But talking to the ICs is where to start to figure out how does it work currently? Because half the time people think they need AI, and the answer is just hey, this process and this workflow was created 20 years ago, that person has since left, and no one has ever thought to question it. And something that takes us a week and a half to do across five people and 90 emails could actually be done by two people in two days with four emails. With or without AI. Exactly. And they're like they got you know twenty hours a week for five people back um just by reinventing the process, regardless of AI. And then the question, the other part of your question, I think, too, that you're getting at correct me if I'm wrong, is like in leadership, who should be making that kind of call?
SPEAKER_00Um department.
SPEAKER_02Yeah, I have found operations with technically minded people is usually the best outcome. Uh because usually everyone else has their own like CFOs, achieved growth, sales, you know, related folks, they've all got their bent on it. Most, I find most well operating operations department act a bit like Switzerland. Like their goal is to serve the business, not whatever their interests are. And so operations tends to be a good department if you look at a department level. At a role level, that's harder because a lot of people think that it's you know gotta be a very technical person. But does that technical person have the business acumen to understand why to do something that way or what the ROI would be doing this versus that? But on the business side, it can be tough too, because a lot of business people have no idea what an API or an MCP is, and so they they don't know what a potential solution might be. So how how could they say this is our solve when they don't even know what the possibilities are? So I know that's not a perfect answer, but I lean towards operations as a department. Yeah. And it's technical people with business acumen.
SPEAKER_01Okay, I totally agree with that. Uh operations understands holistically what's happening in the organization and and that sort of thing. Sales is focused on their domain, finance focused on their domain. They're not necessarily paying attention to the fact that customer service has a backlog or whatever. Okay, so we've got a listener, they're operations influencers, or they might be the head of operations, but they don't have that technical uh understanding, which I agree, you don't you don't necessarily need to understand how what machine learning is or how data science works. It's much more important that you understand process, truly, because generative AI can adapt to that stuff. But if I want to make sure that I'm as that operations listener, that I'm covering my bases and I do have that perspective from a technical side, what would what would be your recommendation? Is it approach IT? Is it approach somebody in the organization who has already been like, oh yeah, I've been using AI for a couple years and I'm I'm building stuff at home or I've built an app to help me with this sort of thing? Is it the the grassroots person or is it the traditionally trained technologist?
SPEAKER_02Depends on who you have in your org. Usually uh a a good rubric or a good test of this that I have found is ask, you know, whoever it might be, or maybe the leadership team gets together and says, who's the most technical person in our organization? And then ask them what they're capable of. And a lot of times, like we ask that question too, because we're told, you know, oh, we've got someone technical who can help you. Right. And then we talk to that person and everyone else in the company is blown away by this person, and we talk to them, and their peak top-tier technical ability is creating pivot tables in Excel. And everyone else views that as a high technical competence, and that is not the kind of I mean, bless a building pivot table is confusing, but that's not the point of this kind of work. So that's a good rubric of do we have that person internally? And so if your answer is, you know, building pivot tables or you know, they vibe-coded a to-do app for themselves, that's probably not the person. Uh or the person probably that isn't in the org. If you do have someone in the org, then I would have them involved. And so that person tends to be a bit more classically trained, so to speak, IT or software developer kind of mindset. Because the biggest thing that you need to think about on the technical side is in the multiplayer environment, it's not just one person accessing the data, it's many people and who can read from it, who can write from it, what are permissions, structure, etc. A good example is we we had someone come to us after a bit of a blunder on their end where they put operations in charge with no one technical. And so operations built a great resource for themselves to plug into all their their HR data. And so they had this internal chatbot where they could figure out stuff very quickly, and they said, Great, let's release it to the whole org, not realizing that the operations team had different permissions than everyone else, but they gave those to everyone else. Interns were able to say, How much does this CEO make and colleagues were able to say, oh, what did my colleague in other department get as their performance review? And they didn't think about that. So it can it can bite you if you don't have someone technical involved.
SPEAKER_01Agreed. So listeners, if you're thinking about using AI and building tools like that, a major consideration is that permissions thing. Your IT team, if you leave it up to them, they're probably going to lock it down as strict as possible. That may not necessarily be the answer. So don't necessarily I would reach out to somebody like Justin that understands both sides of the table because you have a technical background, right, Justin? Yeah. So like a very good fit to answer this question. I want to take a couple steps back real quick. You mentioned that there was a couple of companies that have reached out, and their approach was going to be, well, we don't need you talking to our people. We're just going to have leadership handle it. From what you understood from those conversations, was it um that they didn't trust the acumen of the staff level individuals, or they didn't want to necessarily introduce this specter of AI replacement into the conversation?
SPEAKER_02Yeah, I mean it's anecdotal, it's an you know an N of you know eight or nine. Yeah. But the trend has certainly been uh folks who have been there at the business for a while and they think that they know, you know, because I built it or I was here when we were 40 people and we might be 80, or I was here at 80 and we're 160, whatever it might be, they feel that they understand the business intimately enough at all levels to be able to describe all the processes and all the data and all the things. And if they can, so be it. But I have I've yet to ever see that be true.
SPEAKER_01Yeah. So I've seen the same thing, Justin. So audience, I want you to think about this. When you're thinking about AI, there is there's very common for someone to be like, oh pff, I know AI, right? You we don't need the we don't need the expert, or I can handle this, or all you got to do is put it into Chat GPT and ask it what you should do. Like, not the right person to lead. Perhaps they're part of the AI council or the the the steering committee on what we're gonna do about AI for sure, but I wouldn't um I'd be very careful trusting that internal evaluation of an individual's acumen for sure. So with this, um I I I heard this concept this morning on another podcast called uh Artificial Intelligence Show with Paul Reuters, so I don't know if you listen to that, but it's a great one. Um and he introduced me this term that I hadn't heard before. I think you for sure and most of the listeners have heard of this term human in the loop, right? But he introduced this idea, there's like uh uh a sister uh acronym, which is human in the lead, right? And if you think about that for a second, like human in the loop just means like, yeah, I'm seeing what's happening. Oh yeah, that looks good, that that sends. But it's not necessarily indicating that the human is making the decision. So human in the lead, like I'm waiting for the output before I tell it the next step kind of thing. So um and and that is probably much more aligned with my next question, which is what signals tell you that a handoff or a process is ready for automation besides the I don't like doing it, I'm copying spreadsheets. Are there any other key things that you would look for that maybe a listener could say, you know, I I don't know where they're pay copying and pasting 40 times a day, but I I do see this.
SPEAKER_02Right. Yeah, I think that is about deterministic versus non-deterministic outcomes. I think that's that's the same reason we've seen software engineering be one of the core competencies that is kind of solved for with AI to a degree. Like human in the lead is what all software engineers are doing now because the code is deterministic. You know that if you don't put you know a column there, it's gonna fail. And so because it's so deterministic, it is kind of in solved for to a degree. Um and so I mean we see it with our own software engineers where you know they're in the lead as in they've given it here's a goal and an outcome I want, write the code to get there and then I will review it. But as long as it runs and I can check the security of it, it's probably good. And that's very different than human in the loop, where it's like I have to review everything and I I know I'm gonna need to make a bunch of changes. So I give that analogy to say processes that have deterministic outcomes are the ones that are ready for human in the lead. Whereas human in the loop is more about where you still need someone legal is a great example of human in the loop. Yeah. Where an AI might make calls on, well, this clause actually isn't needed, and our indemnification can be this and our liability can be that, but it still needs someone to sign off on that, uh, both for liability, but also you can't trust it, you know, to that level with something that's so non-deterministic as legal interpretation.
SPEAKER_01Sure. Um, okay, this is uh a a little bit of a parallel track that I want to ask here. And it has to do with, because I we we do a Friday call with all the certified chief AI officers and that came up this morning. Um, the difference between like some clients, they're looking for, you know, hey, we just I'm looking for a prompt library. If you're new to to AI, yeah, that may be the case. But with the advancements of the capabilities of the models, you already referenced it. We don't tell it what to do, we tell it the outcome. And then we let the the smart version of the model, the 5.6, I what is it, Terra or from ChatGPT or Astra now, I guess.
SPEAKER_02Astra as of yesterday, yeah.
SPEAKER_01Do you have access to it?
SPEAKER_02Uh no, I should by the end of today, but not yet.
SPEAKER_01Okay, good, because I'm I'm waiting to. Or, you know, Fable 5.1 was released yesterday as well, maybe?
SPEAKER_02Uh earlier this week, lastly. That one I that one I had for a while, so I can't remember the actual release. But yeah, that's a good example.
SPEAKER_01So, and for the listener, I want to make this distinction here. Uh, human in the loop, human in the lead. In the old way of working with the models, my engineered prompt was very important. But I still needed to necessarily like make sure that it was following my instructions. This is what the prompt said. Did that come out? I'm the human, I need to review that stuff. Whereas with the current capabilities, it's moving more into what you referenced. Your engineers say, This is this is the outcome that I'm looking for. And so that's now human in the lead. You're telling it the outcome, but you're not necessarily investigating every step that it's doing to make sure that it followed your instructions, because your instructions were pretty simple. They were, this is the outcome I'm looking for, you know, some beefed up version of that requirement. So am I on track based on how you're seeing things as well?
SPEAKER_02Yeah, I the way I've been thinking about it, it's kind of like bowling. Uh imagine if you were bowling and you had to literally roll the ball slowly down the lane into each um oh my goodness, what's the word I'm looking for?
SPEAKER_01Lane, pin.
SPEAKER_02Pin, thank you. Uh imagine you had to roll it into each pin. That that was kind of the first era of AI where, to your point, very clear instructions, you have to baby it along. Uh we then moved in the last call it six months where bumpers became a thing. And so you can roll that ball down and you've got the bumpers as a guardrail, and now those bumpers are kind of like you everyone doesn't have to be an expert bowler, they're probably gonna hit the pins every time. Uh hopefully that analogy is not falling apart on itself, but that's kind of the stages that we've been through. And a lot of people use Chat GPT when it first came out, and it felt like a Google on steroids that was about it, and it kind of sucked. The ability for you to describe something and say, because you've got access to my calendar, my email, my meeting transcripts, um, I need you to analyze my last week and tell me what my priorities are for next week. And if you did that a year ago, you would laugh at the output, and now it's pretty damn. And that's why I use the bowling analogy because it is able to reason for itself and parse through what it needs to figure out and tell you. And the big part why it's so important for companies to understand that distinction is the important part now is the context you give it. So that's why kind of in that multiplayer environment, you need to establish what is our process, our workflows, our requirements, our standards. Because you those are the bumpers and bowling that you give it to say, whenever I roll this ball down the lane, know that we are this type of company in this kind of industry. This is our org structure, this is how we approach sales, we approach legal, we approach finance. Because if it has that context, it can usually figure out the rest. Um, and it's only getting better at that.
SPEAKER_01So I I know that the context um uh acquisition, context structure, it's a big topic. Um I first got exposed to this idea on its importance with uh Jack Dorsey, he released the from uh hierarchy to intelligence paper and that sort of thing about this intelligence layer that as long as you're capturing all the artifacts of your business, like agents or automations, like they can do most of this stuff that we're uh approaching Justin for in the first place. Hey, this process sucks, it takes too long, whatever, right? So is the best place for a company to start with like the the the establishing some context? Is it those those connectors or those plugins or integrations to Gmail calendars and the note-taking tools?
SPEAKER_02Yeah, I think most people start with a single player experience of giving it context. So to your point, you know, the MCPs or the integrations in Cloud Chat or ChatGPT chat and saying here's access to calendar, email, etc. etc. And that's personal context that you're giving it. Um and that's why the multiplayer environment is so different, as in yes, sure, AI can read the equivalent of a Harry Potter book in about a minute. But if you've got, you know, a hundred people and all of their email and all of their calendar and all of it, et cetera, et cetera, um, you really want to think about well, how do I give preference to the latest version of a standard document we have or the latest template, or if we've got two departments disagreeing on what a term means, like you know, EBITDA for uh revenue metric or a financial metric, like what do we mean? What is the calculation behind that? You've got to make sure that those things are clear because if AI looks around your business and has nine, nine formulas for how you calculate your sales pipeline, how's it going to be able to tell you when you ask it what your sales pipeline is? So that can start.
SPEAKER_01Or eight different logistics price quoting companies or individuals who have thousand rows of rate sheets.
SPEAKER_02Yeah, exactly.
SPEAKER_01So if you're listening and that's your situation, don't worry. The example that Justin gave, and I doubt it's the same company, but another logistic company, like that's how they were doing it too. And it was a bunch of people spending a bunch of time. So if you've got these in your business, don't feel like, oh, you know, wah, why it's not the case at all. Um, so okay, great. We've got the individual who's dialed in their stuff. And I'll give just for the listener, I'll give you a perfect example. The other night I was up with uh one of my kids was sick long night, woke up, had three meetings, and I was like, uh-oh. And the first thing that I did was I said, hey, Claude, take a look at my my calendar, see the events that I've got, look at all the email threads related to that, and kind of give me a TLDR of each meeting. Look, I I don't need a script, I understand my business. I just needed those primers, those talking points, so that when I got on the call, I didn't look unprepared. And that that's that's the context application at a personal level. Now, if I've got a hundred people across the department and they've all got, you know, they're all getting a hundred emails a day, and they're each doing several Zooms or Google Meets or Teams meetings, and and I want to capture all of the conversations that were happening in those, that becomes a huge data set. And you mentioned reading Harry Potter uh in a minute. Well, great. That's as page one, page two. Like that's an easy process to follow. But if I've got all of that context and I'm now like, hey, I want you to take a look at the last 90 days worth of calls, and I want you to extract some language from those calls that you're starting to see a pattern that maybe our product isn't addressing or those sorts of things. And and guys, this might get a little geeky for a second, but it's very important because this context concept is going to be something that you're gonna hear more and more about as the models get uh more powerful and that sort of thing. So how how do how do you handle that much data and allow the models to be able to access the context without it taking 20 minutes of reading all the Harry Potters sequentially?
SPEAKER_02Uh we often encourage teams to start small. So I think um there is a proclivity for for many teams to think about AI across their entire business, and we have to have the perfect data lake and all the context of everyone's head for every role in every department from day one, and then we can start using AI. And I would think about the opposite of that of pick one area of the business. Yes. The technical thing you have to think about is the knowledge and context in a place where other departments will eventually plug into. But don't start with we have to get everyone in one place and then we start using AI. Start with usually just one workflow, like one key workflow, not just you know, someone replying to an email, but kind of a core area of the business, or pick an entire department and all of their work and start there and then start layering in because then you start to establish norms because most businesses are made up of a culture of what's expected, because it's humans deciding that this is something that can only be one version of, whereas this thing everyone has a version on their C drive somewhere of the same file and everyone edits it and no one actually has a source of truth. And so if you start small and then establish norms and then add on to it, you end up in a much better place because if you start to break those norms, you will you will notice AI's outputs are worse and it kind of acts as a self, self-healing slap on the wrist to say, oh, we added this department and now it's not working as well. Oh, it's because they have six versions of the same template or whatever it might be.
SPEAKER_01Yeah. So look, as a listener, if you're hearing this stuff and you're like, I'm never gonna touch that stuff. Maybe that's not the point of this conversation, but but the point of this conversation is that you are now more educated when the vendor approaches you or when the board says, hey, this is now a mandate, you're not gonna be going, I I guess that's the right thing to do. So I don't want you to necessarily think like, whoa, that sounds heavy, it's a heavy lift. I I can't lead that. No, but you now are more prepared to participate in the conversation in selection of the approach or the vendor or that sort of thing. So thank you for sharing that, Justin. So I want to, I want to kind of shift gears now to some fun stuff, let's say. Um as much as it doesn't necessarily impact the business, anytime I'm working with leaders or executives or decision makers, if if I start talking about tools, boy, they're really interested. So let's talk about a couple of tools that you as a business owner um uh would like have have really made a big impact in your uh ability to close more loops, not go to bed with all this but running in your head.
SPEAKER_02My personal top tools I would say that I rely on is granola is probably my favorite. Um just because so much, I mean, given my role, as you can probably imagine, I'm in a lot of meetings. Um but even then, everyone in our business loves it because Granola is easy to use. They now have an MCP where you can plug it into anything that you want, and you can organize everything in folders. So then to your point of oh, it feels really heavy to think about context. Well, if you've got an app where all your meeting transcripts come to one place, you can just organize folders. There you go. It's not that heavy. Um, but you can plug that into everything else. So Granola, I love.
SPEAKER_01Um, just for everybody, Granola is uh uh an AI meeting note taker.
SPEAKER_02Yes, sorry, yeah. Um I have quickly grown to love Grockbot.
SPEAKER_00So Me too.
SPEAKER_02Yeah, I am extremely impressed. It is Yeah, I encourage folks to play around with it because it to me indicates uh where the models are getting so good that you don't have to worry about so much prompting and figuring stuff out. It continues to impress me how vague you can be in your ask and how much it will figure out for itself and say, Oh, when they're talking about this project with these people, I'm gonna go look at the Slack channel that they're in for that and the notion gauge that they have for that and some email threads and you know granola, whatever it might be. Um, I would encourage folks to to play around with Grokbot. Again, it is very much a solo single player kind of tool. Um, but since your question is what do I use personally? Um that's one.
SPEAKER_01And then the other is Let me ask before you go on to the next one, what are some of the activities that you have the sub agents doing within Grokbot that like are really helpful?
SPEAKER_02For those who haven't had a chance to use it, GrockBot almost feels like a Slack or iMessage app where the sidebar has all the bots that you want to put in and then you chat with them. That might sound like isn't that just ChatGPT or Claude? No. The difference is A, the integrations are way better, B, it's got its own computer, which is the cool part, so it can go and figure out stuff for you more easily and access things that Claude or Chat GPT might say, I don't know how to do that. And then the third thing is those bots can talk to each other. So I organize mine around areas of the business. So I have a chief of staff bot that talks to a growth bot, which is our sales and marketing, our service delivery, which is our our service delivery as a as a consulting firm, uh, finance bot, like all of these can talk to each other. So I have a bunch of routines and starting points with the chief of staff, and it will say, you know, I'll say, oh, the content calendar for next week. I have an idea about XYZ. And the chief of staff bot will go chat with the the growth bot who manages to literally chat with it.
SPEAKER_01Yeah. Yeah.
SPEAKER_02You can see them chatting with each other, yeah, but yeah, you don't have to. The chat chief of staff one will come back to me 10 minutes later and say, Yeah, looked at that. You should move this to two weeks from now, and I can put this one in its place. The the growth bot thinks we should frame it this way. What do you think? Okay, sure. Um, so it's just really impressive the simplicity of the UX and interaction model, but it feels like a step forward. Um I really like that.
SPEAKER_01I know you got one more tool, but I don't want to get off grokbot yet because I'm glad you brought that up. Go for it. I was using it two minutes before we got on here, right? Um, so for the listener, uh you can get a free test. I mean, it you can burn up those tokens pretty quick, but for $20 a month with a cursor account, like you now have access to exactly what Justin's talking about. And it is, to me, Justin, it's like the first time that I've been able to use a tool that was that powerful and that easy. And me to go, oh, wait a minute, let me take a copy of my org chart, give it to the chief of staff, because it's got links to all the job descriptions, and go, how many of these can you build? Right. So, I mean, it's it's like it's pretty incredible. All right, yeah, I know you had one more tool you were about to mention. I didn't mean to interrupt you.
SPEAKER_02No, all good. The last one is not gonna be everyone has heard of it, but Claude Code. And I I bring it up because I think a lot of people are used to Claude Chat or Claude Cowork. Yep. I encourage everyone to try Claude Code because the word code in there is a misnomer. Uh Claude Cowork, for example, is just Claude Code under the hood with a different UI on top. But Claude Code is just so much more capable, and you you can just tell it to sign up sub-agents. So uh again, people think of agents as this mythical like product. Agents aren't really a product, it's just a function of how AI can work. And so you can say, as a quick example, we are gonna be doing a bunch of hiring and a bunch of our job descriptions. We've changed one kind of key thing of how we kind of frame all job descriptions. And so I said in Claude Code, if I did this in Claude Chattercour, it would have taken way longer, probably been less effective. But in Claude Code, I just said, here's the Google Drive link to all the job descriptions. I want to change XYZ about all of them, use subagents to do that. And it did it in about four minutes across a dozen job descriptions because in Claude Code, you can ask it, you just tell it to spin up sub-agents and you can see it doing it. And I encourage everyone to just try it because it's it's it's not really about the code, it's just it's the most powerful version of Claude. Same goes for if you're a ChatGPT user, yeah.
SPEAKER_01Yeah. Totally agree. And for for the listener, like if you have a Claude account, now what I would do is I would download the desktop version of Claude or the desktop version of ChatGPT. And if you have those on Claude, you'll see the link for the navigation link for code in the upper left. And if you have Chat GPT and you load up the desktop app, you have to actually go up to where it says, you know, chat or work, and then click on the little arrow and it'll do a drop-down, and then you'll see codex there. Now, when it opens up, you're gonna be like, this looks like Claude or this looks like ChatGPT, but the capabilities that it has are so I look at it like and I'm I mentioned this a lot, like if you're using chat, it's kind of like a drive-thru window for most of us. Like we go, we get our answer, and we leave, right? We go back to doing the thing. But if you can leverage the tools like like Code or uh Cloud Code or GPT Codex, it becomes more like the AI's now got a seat in your office. It's like it truly is an assistant or a co-pilot. So okay, last question then, Justin. It's a it's kind of a new segment. I know when we did the pre-interview, I introduced you to this idea that like, okay, we're we're gonna be introducing some stuff. So I'm slowly doing that. And one of the new segments that I'm introducing is what I call the handoff. What is something that you think that every listener, um, regardless of where they are, should be handing something off to AI, should no longer be doing it as the, you know, the human doing the work?
SPEAKER_02If you've got your AI plugged into your tools in your business, so that's a big delineation of if it's just a chat window with no connection to anything that you use at work, this isn't as useful. But at the same time, if you're stuck in that mode, you should chat with your IT team or leadership about changing that. Um but at the end of the day, if you are in any sort of role where you are asked to provide status updates or summaries of where things are at, whether that's an initiative, a project, as simple as a task or something that you're working on, um, I would suggest creating a skill for yourself. And then Claude and ChatGPT, you can do this where you say, I want to create a skill that does XYZ. And I would encourage you to create a skill for whatever thing that you're asked for on a regular basis for status updates on. Tell it where to find the information about that, the context it needs to know about how to parse that information, create the status update, and the kind of tone and kind of language you want it in, and then save it as a skill and just trigger that skill on a regular basis. Because then it blows my mind how many people spend so much time being. Chased by other people for where's this at, or where's this biweekly summary or this report that I'm looking for? And it continues to impress me how good the models are getting at building like charts and actual reports and whatnot, too. So again, you can do that if that's part of what you need to do. So if you're regularly being asked for updates on things about the same thing, turn it into a skill.
SPEAKER_01Yep. Great suggestion. And for the listeners at the chiefaiofficer.com forward slash community. If you join our community, it's free to join. Um we do routine trainings on exactly that, like like skill creation because it is such a powerful tool. And look, you know, insider tip, if you're not sure, well, well, where would I start? Go into your LLM and start a chat and say, hey, I want to build some skills, take a look at what I've been doing, what we've been doing together, and maybe make some suggestions. So again, going back to that thing, I'd love to use AI more, but I don't know where to use it. Um Justin, fantastic conversation. Um, it sounds like and now I what we do at Chief AI Officer is not technical. And I know that what you do is very similar to what we do, but you're approaching it with a team of engineers and developers and um more of the builder kind of thing. So uh it was uh great to hear that even though I don't have a lot of optics on the technical side of building, at least um as far as how we're viewing the paradigm, different approaches, but we're seeing the same thing. So, you know, again, listener, like you know, if you know, if if um you're hearing this from multiple sources, it's usually like, okay, that's a signal, that's not noise. And I would certainly suggest that there was a lot of signal introduced into this conversation. So Justin Watt, founder of Switchboard, you can find them at with switchboard.com, a co-founder and CEO of uh with switchboard. Um in our pre-interview, I love the conversation. I think this is a fantastic episode as well. Um, Justin, anything in closing that you want to uh share with people, especially where they can maybe find out. I know you've been featured in the New York Times and on Notions blog and things like that. Where are you sharing your thoughts and uh and any closing thoughts you have for the listeners?
SPEAKER_02Yeah, like you said, with switchboard.com is the site. Um I'm active on LinkedIn, so Justin Watt, WAT T, if if folks want to talk shop or or anything, the DMs are open. Um and the parting thought is just start. I think that's the biggest thing is the amount of people that I I chat with who aren't even starting with AI because there's this overwhelm with it or this fear. It isn't it I totally get the intimidation, but just try it out and just try to your point, a skill. Hop into a chat and say, you know, you're connected to my email and my calendar. What are the repetitive things I'm doing? What recommendations would you have on things that you could help out with? Just little things. Just start.
SPEAKER_01Great advice. So uh thank you so much for being here, Justin. And before we wrap up, folks, I just want to say thank you. Um, because of you, we've uh now reached the status of the top two and a half percent of podcasts globally. Uh had no idea where this was gonna go when we started this in January of 2024. Um, and that's not because of me, that's because of you, the listeners, uh listening to it, showing up every week, uh, learning from it, sharing it, giving us feedback. So if you are one of those listeners who's gotten a lot out of this episode or really any episode, we'd love it if you'd share uh a rating or a comment or um any type of review on the platform that you're listening on. And look, there's people that you work with that need to know the stuff that we're talking about this. So please share the um share the link to the podcast or let them know, or if you you know you want to look like the smarty pants in the meeting and you want to share something that you learned from this conversation with Justin today, let them know. And I I I listened to the Using AI at Work podcast, and Justin Watt was the guest, and this is what he shared. So we'd we'd love any endorsements that you could share. So thank you so much, everybody, and um, we'll be back next week with another fascinating episode of Using AI at Work. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer, for empowering businesses with AI education and training. Visit their website for a free AI readiness assessment and AI strategy guide to help you get started using AI at work. That's www.chiefaiofficer.com. Follow us on Twitter at the handle UsingAI at work and visit www.usingai at work.com for free resources to help you harness AI in your role.