69. Building an AI First Organisation

Hello everyone, and welcome to another episode of The Operations Room, a podcast for CEOs. I am Brandon Bensinger, joined by my amazing costar Bethany Errors. How are things going, Bethany? I don't know. So one thing is I don't think Instagram is good for me. I suspect Instagram isn't good for anybody, but yet despite that, I still am on it and I'm still on it probably too much. And it really feeds my health anxiety.

Health news and all the ways that we are killing ourselves and all of the rules that I'm not following and all of the reasons why when I die, it's going to be my fault. And even though I'm aware of it and I have blocked as many health news things as I can, any time it comes up, I can't help but watch. And then the algorithms like, what she cares about is health news. And then it just comes back in. So the most recent one was actually not specifically a health news influencer account.

It was a data account and it was about how women are I don't know if it's dying of cancer or just getting cancer at a tremendously higher rate than men. And then you had comments from random people trying to decide why it is that women are getting cancer at much higher rates than men. And those comments have seeped into my brain and now I'm freaking out. So it was that women wear a lot more cosmetics than men and they're not properly tested and a lot of them are made with petroleum products.

And so we're just like smothering oil all over our bodies. And that causes issues. Men tend to wear more cotton than women, and women wear more special fabrics, and all of the special fabrics are made out of plastic. And as they degrade, you get microplastics in your body, sports bras directly linked to breast cancer. I mean, all of this is not real. And these are people in the comment section postulating as to why women have higher rates of cancer.

But my ability to cheat death and do all of the right things, follow all the rules and never die. That's a lot of pressure on somebody's shoulders. At some point you will die, Bethenny, I suspect. No, no. If I follow all the rules, I'll be fine forever. So sometimes Instagram's not bad. I was served a video earlier this week That was Jane Fonda talking about old age. And the opening of the video was the best because she said, I am Jane Fonda.

You may know me from being Jane Fonda for the last 86 years. Exactly. And growing up, it's literally impossible somehow to not know who she is. In the video, she it's like really. Well, production is video. And she's working with a scientist in the scientist is talking about the facts. And then Jane Fonda is adding in her bits. And one of the things she said like trying to banish senior moments as a thing. And it's just like what we're telling ourselves is that we're

losing our memory. Like when we can't find our car keys when we're 17. We're not saying, it's a senior moment, we just can't find our car keys. And to embrace aging and not be so afraid of it and also say that we all focus on ending up in nursing homes and freaking out about nursing homes. But actually only 3% of the population will end up in them for that. True? 3%. I mean, that's true according to Instagram and Jane Fonda. I have no idea if that's true. We could go ahead and verify that.

Jane Fonda said it has got to be true. Jane Fonda and the scientist said it. So, you know, I choose to believe that that is true, that 3% of us are going to end up in nursing homes and the rest of us are going to have some combination of home care or just dying doing okay. So anyhow, that's the inner workings of my brain. So how am I doing, Brandon? I don't know. All right. So we have got an amazing topic for today, which is making AI happen in 2025 for your organization.

We have an amazing guest for this, which is Charlie Cullen. He is an AI strategist and has singlehandedly developed a product called Koala using AI tools behind him in terms of development tools to make the app happen. Being a non techie, which is phenomenal and quite an amazing story. So before we get to Charlie, the first question I wanted to ask you bit of a broad question, which is what is the opportunity right now for scale ups to go horror on being AI first, as it were.

I made a face when you asked that question because I was wondering if you meant I first in the product or I first within the company. I mean, I think there's the bigger existential question is, is your product shit and do you need to tear it up and just started again as an AI product. And if you're more on the startup and scale up side, I would highly encourage you to do that. I'm just pushing that one out.

There is like a little moment of of compelling thinking or controversial thoughts, although it's interesting because everything is so new. There's a lot of scope for improved products. Like for me, the really big gap is a next generation Zapier like something that is really easy for non-technical users to you to use to automate the soul destroying work in their lives.

And there are loads and loads of companies trying to figure out a tick, but they're still like technical for technical people rather than no code. Easy automation for non techies. So I'm very excited about that space. And once that. Unlocks. Then everything in your business should be your first.

It's a bit of a question of agents which are going to be verticals right now for very specific functions as opposed to agents working together in tandem to actually have more of a net effect for an organization. And I think that's definitely out there in the future. I think we're going to get first is more siloed agents for Realm A versus Run B? Yeah, but like a lot of those agents are just siloed. Vertical agents are not interesting

for me. My Nirvana. And I think at some point this year will be I have all of the CVS to sort through. I don't want to sort through them. I tell the agent, these are the topics that I'm looking for on the CVS. This is what good looks like. Go and read all of these and bring out the ones that I'm looking for rather than a rules based. And then you're like, No, no, you didn't quite get the CVS, right? Really? I want to have blah, blah,

blah, blah, blah, blah. And, you know, like actually doing the boring work for you, but you're prompting them in English. But I would consider that to be a vertical sized agent for the people recruitment function as an example. I think that's definitely in sight right now. I think what I was thinking more in my head is more around broader activities where, you know, it's not just simply a recruiter and CV thing that they're doing, but in fact it's maybe the entire people function.

Multiple agents need to work together and cross-pollinate to come up with better outcomes, I suppose. I think you're right in terms of exists now. But the problem is, is every single one of those things exists as a $10 subscription and I want to pay $130 subscription and do it all where, you know, there's like one tool that the people can use to do their stuff and accounts payable can use to do accounts payable and sales can use to automate their stuff.

And you have like a new automation layer that the entire business can use. I think from an operational standpoint, for a company, it seems like the low hanging fruit that is out there is simply your processes and policies. Anything that's been documented in that form, which is either internally creating a subset of GB TS that pass off different pieces of policy sections or process sections or functions or whatever. Or what I've also seen is using GB to your court or what have

you. But what I've also seen is actual agents that are set up to ingest your policies or ingest your processes and do something more advanced than what you get with an internal chat and GPT that you create yourself basically. So one example of this was the internal knowledge base curator, GPT. So that product and it is a product, you basically take your documentation with your company, ingest it into the CBT.

It's all set up to sort, categorize and provide a more advanced experience for the end user in terms of being able to search for things and so on with a taxonomy that's created and so on. And that seems very interesting to me. Yeah, I just don't want to have to pay 10 pounds for every single one of them or 20 pounds for every single one of them. I'm looking forward to the consolidation and it just at what level does that consolidation happen?

So I was reading Charlie's LinkedIn posts and he created one post around hiring for an AI first organization, and his little description that he put in there was to create a task for a candidate where they need to write a prompt or create a. Claude Church project that can accelerate a key task in the role that they're being hired for and to demonstrate an example of how they would use it. And thus that's really interesting

to me. Or maybe think at the very least in the sense of like, okay, if we want to hire people that are on a I using a, I have some level of competence and skill and interest in A.I.. It seems like at the very least asking an interview question in some form, maybe not a task, but some kind of questioning around this seems to make tremendous sense. I'm just curious what you think. I agree. And I actually think being a task would be totally legitimate. And it shows interest.

It shows aptitude and it shows curiosity. I think there's still a level of resistance to a AI because it's seen as cheating. And so it's actually showing people who have a AI first mindset or forward mindset where using A.I. isn't cheating, it's the new normal and it's the smart thing to do. And so both you're testing for people in who are coming into your organization, who are afterward, but you're also massively demonstrating expectations that in your world, using AI is smart, not cheating.

So when you think about rolling out A.I. within the organization right now in 2025 for a company, we have our standard tools, if you want to call it that, which is clod chatbot, Gemini. What makes sense here? Would you encourage companies right now to say to themselves, okay, look, let's create a policy around this. Let's license some of these tools for different functions within the company or within the entire organization and launch them into the company.

Upskill people in terms of how to write prompts or create projects, create some level of support structure around it where people are being supported to use it effectively, and some kind of reward mechanism to reward those who are actually integrating it back into their jobs. Is that something that we should do now? All the technology is changing so rapidly that I would not commit to a single one.

I would go even if it's a bit more expensive for rolling monthly contracts and continue to experiment. It's too early to sign up for an annual commitment to anything right now. I don't even know if you can, but if you can, I would not suggest annual commitments. It's not worth it. Things are moving fast. Policies, Definitely. Because it makes you think through all of the gotchas and it makes it really clear for the organization just to push Charlie's LinkedIn a little bit

more. He happened to have a LinkedIn post this morning that was a seven page, a AI policy template to use that looked pretty decent for what a lot of the main thoughts if you want, if you haven't put one in place yet or you want to make sure that you're covering the key areas, I suggest having a look at it.

And then the reward mechanism or for us at peak, we have a few different ways that we are focusing on the team using and getting more experimental and interested in generally because we have pockets that are super interested in pockets that aren't. So in our weekly we have like a company all hands that's weekly. There is a five minute segment every single week where somebody does a show and tell on something they've done. The second thing that we've done is we have a Slack channel that's a

Jenn-air channel. We actually used to have two and have just merged it into one because it was getting a bit messy as to which is which. And so all news, all thoughts, all new technology, all uses goes into that channel and it's a self-selecting channel. But I think most of the businesses in it. And then the third thing that we are rolling out this year is for our technical teams, a Jenn-air training course we're running internally.

We've not found anything externally that matches our needs and that is an enablement course. It'll be running for the first half of the year and then we'll see what the second half brings us. So when it comes to budgeting and providing different functions that have different aspirations and different needs around some of these generic tools, what are you doing from a budgeting standpoint? You're you're creating like a budget for each of the functions or how does that work?

So we're doing a bit of of that for experimentation and like what are the right tools, what are the best tools? And then also for all of our systems of records or standard SAS tools as they come up for renewal. We are investigating the market and looking for eye first alternatives and seeing whether or not they're too early stage to move to.

But our general assumption of hypothesis is all of the SAS businesses that have a bit of AI stuck on the side are not going to be as good as AI first businesses. And so we're constantly searching in the market for what is the new system of record. Like in this new world. And for my vision, I think we have one system of record with lots of different elements to it. Because what's a system? A record, but a massive database.

And so and there's always been that the reason why the systems record are different is because of the interface and the tooling required. But if the interface becomes a chat box for pretty much everybody, do we need different systems of record long term? Short term we do. Because I don't think anybody's reached that long term goal. But that's part of why we're not committing to long term contracts right now while we watch the space.

And so for our existing systems of record, when they're coming up for renewal, we're not doing three year renewals for anything. We're doing one year renewals so that we have this choice, even if nothing's fit for purpose. Yet our our guest is something will be fit for purpose in a year. So when it comes to the policy that you now have and you also mentioned Charlie on his LinkedIn profile, putting together a template, a policy. He talked about a couple of

interesting things. One was in the policy itself to provide a list of encouraged uses for each function that are endorsed by the leaders. And I think this is quite important and useful in the sense that the policy itself is not simply there to tell people what to not do, but in fact to do the opposite in this case, because what we want with AI is to truly embed in people's thoughts that we want the company to win. And the way the company is going to win of the future is using A.I.

tools. And we need you to try them. We need you to experiment. And by providing a list of encouraged use cases and it's not to say that it's going to be exhaustive, but you need to put something out there where people like, okay, within my function of sales, here's like the ten things that the cells like right now has tacitly endorsed, basically. So maybe I should start doing something about it, I guess. What do you think of that first one? So we have that, but not in our policy.

We have it in our strategy document. And so what we've done is for each department, we have like low hanging fruit, you know, what are like the immediate things that we can be doing. What are areas where we want to do more, investigate and experiment with, and what is our long term vision? So what is our ideal in each of those areas? What we have also done that I forgot to mention is we have an A.I. evangelist in every team.

So that's the person who self-selected could be very different in each team, but it's the person who just like naturally is reading everything, naturally experimenting. They tend to be the person who's in the all hands showing information, but they're like the go to person of Ah, I've just realized that stars can do whatever.

Or in the second one that he suggested should be in the policy is deciding on a list of company approved A.I. tools in a particular a streamlined process to get tools and versions approved, i.e. somebody see something, they make a request, they get an approval within days as an SLA. And the default is to say yes, pretty much. What do you make of that? So we already have that in place, but not specifically for genitals, just in general for buying for our security reasons.

So it has to go through a security audit, which means that it's not within a couple hours. It just depends on how much information we can get about these companies and how they're treating our data and whether or not they have ISO 27,001, whether or not they're GDPR compliant, etc. And sometimes that can take weeks.

But I think it's worth it because there's so many new tiny companies that the chat you or the Open Eyes and Gemini like you can find out all of their security really quickly and make decisions and understand whether or not your data is training the models, etc.. Tiny company with 15 people, we end up contacting them, getting their certifications, understanding what their policies all are. If they don't have certifications yet, and talking through whether or not we feel comfortable working with

them. Quite often they choose small and haven't not mature enough and don't have it yet, or maybe haven't even been around enough for an audit. But we will have conversations with them and ask them how they're thinking about things and if they seem to understand what they're talking about and be quite forward thinking and have policies in place and a roadmap of where they're going to get to, depending on the sensitivity of the data that might be going there will still say yes.

But if we talk to somebody and it's just like, okay, these people are very immature, they have no idea.

We cannot trust them with any level of data, we'll say no. The other thing that occurs to me for companies right now, irrespective of any policy, are actually rolling on AI at all is just this obvious thing where right now today, if you don't have any AI policy, you've got a problem a little bit because people in your company are using personal accounts to use chatbot to write their emails or whatever that is happening.

So having a very simple statement to the company, irrespective of anything else, just saying, look, if you're using a personal. Howard, please, please, please go into your settings and toggle the data learning off essentially, because if you don't, anything that you do will be sucked into the vortex of the the large language models and we cannot have that. And also then write an AI policy. The last bit I wanted to bring up was best practices.

And you had a couple of nuggets here that I thought were useful. One was when it comes to clod projects and church EPD projects and projects are essentially a framework to allow you to have a consistent set of context for whatever it is that you're asking of the GPU in this case to respond to around the company or around your products or around your policies or whatever, whereby you don't have to enter a prompt every single time to put in that context.

And his best practice was to ensure that those projects and those statements around the company were there to allow any user within the company to take that project, take that context and apply it to their their account very specifically to ensure that all the context was all pre set up. When you think about one. That is beyond what I've experienced so far. So I just be like, if Charlie says it's a good idea, it's a good idea.

So what do we park it here and get on to our conversation with Mr. Charlie Cullen?

Why don't you talk a little bit about Koala? Not so much for a product that people should go and buy, but the journey of creating it. So in August 2024, this is when I was in pink. My goodness. I'm going to set up a i boutique consultancy. And then suddenly, my goodness. Like, who am I to go and tell people how they should be implementing and integrating these technologies if I've not done it myself? That's just snake oil. This was August.

I was on a ferry to a Greek island with my family, and the kids were sleeping and my wife and I were reading and I was reading a book called The Jolt Effect. And they did some research where they looked at why enterprise sales end up with the status quo. And the prevailing thought was that people stayed with the status quo because as a sales person, you had not been able to sell the difference from what they've got today.

So I came back from this holiday and I was thinking, okay, well, I'm going to build an AI type solution for this and it will solve my problem of I've never built anything with these tools and then I'll be able to advise people. And even in September of 2024, I was WhatsApp paying friends. Do you know anyone that is a developer that could know how to write this thing? Because I'm a non-technical founder. Does anyone know anyone that could help me either as an agency

or whatever? But just over the course of the summer, some amazing new tools had been coming out new versions of Code, which is one of my favorite albums, which is like a a new tool called Cosa came out, which is an AI development environment. A new tool called V Zero came out, which is a UI development tool. And I just thought one weekend, you know, I'll just see how far I can get as a non developer. And so I started off with Claude. So if you use EPG, then you know how to use Claude.

One of my favorite features of Tord is that you can set up a project and a project you can think of as a wrapper around a specific subject or topic. So I set up a project called Koala, which is the name of the application that I built, and I gave it some custom instructions. You are my technical co-founder and sure I am. Got a clue what I'm doing.

I've got an idea for an app, but I want you to be my co-founder and to help me to architect it, to help me structure the project, to help me to understand what tech stack I should use. And I'm going to come back to you with questions and you should push back on me if what I'm asking for is a stupid idea. And our goal here is just to get to an MVP where we can get some paying customers. So there's more that goes into it than that. But that was basically the custom instruction.

And then as I started chatting with Claude, I would build out Google Docs. That might be the project overview, it might be the

file structure, it might be the licensing model, it might be some of the go to market. And all of this becomes the context that is in this flawed project. And my goodness, I chat to this thing all day long at the weekends because this is a weekend project from eight in the morning I'm chatting to Claude, say, Hey, Claude, you know what you think about this. What do you think about that? And one of the great things about having Claude as a technical co-founder is you can ask stupid questions.

You can ask the question that you asked Claude yesterday, and, you know, he doesn't complain. And because you give him the instructions, he'll push back and say, I wouldn't do it like that. Keep focused on the fact that you're building an MVP. So I wouldn't do that right now. What about might be so Claude was sort of the first, you know, build out the architecture, build out the project plan, build out the user stories and so on. Next step is going over to this tool called V zero.

It's an AI powered generative UI tool. So here you set up a new project into which I upload the context which has come from code. So Claude's given me this whole page product requirements document that goes straight into V zero, and now I can start, you know, can you design the homescreen for me? I want these sections. Here is my brand colors. Here's I want a dark theme.

And without me having to write a very long prompt because it's already got the entire payload, the outcomes, the sidebar outcomes, the project, upload document outcomes, the chat screen. I'm not were even better than that. What you can put into V zero is screenshots of other apps that you like and say I want that but in my project context. So why do I need to reinvent the wheel? Code and chat have got the kind of chat interface fairly well nailed. If my product did that, I'd be

happy. Screenshot. Drop into V0. Give me like that. And so as a non developer or non designer, I should say at this point, you're now designing screens that are rapid, right? And not only does it do the design, but it then creates all of the code, whether that's TypeScript JavaScript in the back end.

So how much did you spend on it to build koala? So when GPT launched and they pulled a number out of thin air of $20 a month and as Sam Altman has said, we basically pulled a number of out of thin air. We're going to charge $20 a month. Well, that's what everyone else has had to follow. So clod is $20 a month. I pay $20 a month for GPT. And I use them both interchangeably throughout the day. Rappler, I think, was $120 for a year's license to call V0 $20 a month.

Kazu I think I pay maybe 30 or $40 a month and then I'm paying for the others because I'm testing them out. That loveable as well. But they're all in this order. A couple of things that are a little bit interesting. Just when people are thinking about pricing models for these things, two things that I've seen are quite interesting v zero and lovable. You can set up a free account. But when you have a free account, your designs are public and so you pay to go private.

And I remember when I was first designing some of the UI screens for Koala, I'm sure no one would have been interested in what I was doing. But for me, I was. I felt compelled to make this thing private. I'm not seeing that before. You could imagine that for Salesforce or HubSpot, you can use it for free, but your contact records are public. Do you want to go private? So that was an interesting dimension. And so you did all of this from September to Wednesday launch. December. December, yeah.

So it was 12 weeks from having never written a line of code to having a production app. I would say that and this was all done during the weekends and a few evenings. There were three weekends when I literally broke down in tears in front of my wife and I was like, Who am I kidding? I don't know what I'm talking about. I should stick to doing what I'm good at. You know, what was I thinking?

Like, there's a reason why developers get paid a ton of money is because I this is complicated and I don't know what I'm doing. And this is one of the problems with AI development tools, is that, yes, it gives you this acceleration from not to amateur, but when something goes wrong, you don't know what's gone wrong because you don't actually understand at the start what you're looking at.

And I was trying to deal with quite complex user authentication through a third party and I just couldn't understand why when certain people were logging in, they couldn't see certain things. And I literally broke down in tears with my wife going, I don't know what I'm doing. And she'd say, Go for a walk or go for a run. That always helps you listen to a podcast.

And so I'd put on a Lenny Chatzky podcast and I would come back inspired and I got I'll just try this one thing, and that one thing always solved it. And I was like, Yes, I'm back in the game and then would carry on. So I finished each weekend on a high. But Saturday afternoons were my my low points.

There's some really interesting conversations happening on X or on LinkedIn where you've got real developers that actually know what it takes to build a production enterprise application and backend infrastructure and scaling versus this whole new breed of indie hackers or non-technical founders that are now spinning up these PCs and the people that have been doing this for a long time and actually know what they're doing.

They're like, You think that's development, That's not development, You haven't got a clue. And the same thing happened when Canva came out and real designers who are using Photoshop or whatever are looking at people using camera going, that's not real design. And I know this trend will continue, but the tools will improve.

Like I said, I listen to this Lenny Rich Jet Ski podcast and always out of my runs and there's two episodes or two sort of sound bites that really stuck with me when I was going through this process. The first and I'll have to remember his name, he was the founder of Right, which became Google Docs, and he's telling Lenny about the process of writing original Google Docs.

And he said, We didn't know what we were doing, but throughout my entire career, I've just got to the edge of what I know and fucked around by getting to the edge of what you know and fucking around. You end up learning stuff. If you stay within what you're doing, then you don't learn. So that was one thing on those days when I was like, I don't know what I'm doing, I don't know what I'm doing. I had his phrase in my mind like, No, but this is where learning is happening.

And the second episode that really stuck with me was a guy called Nikita Baer, who as well a number of B2C apps that have gone viral and be bought by Facebook matter. I think 1 or 2 of them, he said this phrase stuck with me is like, you know, one of the things no one ever tells you is the moment you hit virality, you have to build the whole thing again because it isn't built for the scale you need. And I took the positive

in that. I was like, So what you're telling me is I don't need to build an enterprise solution. I just need to build something that gets to the first 5 or 10 users and solves that problem. And if I can get that far and then prove the value in the use case, then I can get proper people that actually know what they're doing and then build for the next stage.

So that gave me this kind of license to like, I'm not purporting to be an enterprise developer that's going to build a SOC two compliant, scalable thing, but I don't need to build that yet. I just need to prove that I can solve a problem for someone and that they would come and log in. And if I can prove that, then solve the next stage later on.

So this is me pushing your LinkedIn content, but there's probably a piece of content a week that you write that I end up sharing internally. And this week's piece of content was around public versus private data and information with chat shaped. So you want to talk about that? One of the things that I'm always having to remind myself as someone that would now consider myself to be in the AI bubble is that the majority of people are not in the AI bubble.

They're just, you know, going to work doing the work the same way they have. Maybe they've asked a couple of questions, you know, what is or how should I do that? But they've not really embedded it in their work. And to that extent, a lot of the companies I speak to either have not got an AI policy or they've got quite a pessimistic and restrictive AI policy. You know, you must not use GPT on your company laptop. You must now upload company data to GBG.

And definitely because I've come across almost no companies that have this, they have not got a team or enterprise account for GPT or Clod.

So if you're asking people how are you using it, I say either I've got a free account or I pay for the pro account on my own personal card. Now, this is where the data privacy issues come in. So if you're on a free or a paid personal account for GPT by default chat GPT open, I can use what you upload in terms of your content. So that can either be what you write in the chat or any files that you upload that is defined as content. They can use that to train the model by default.

You can go into the settings and that is a I love the way they phrased it. The setting does not say train our models. The setting is would you like to help improve the model for everyone? It's a really nice one. I would like to help improve the model. It doesn't say what that actually means. Your data then goes into training the model by name. You can turn that off. But that is you have to opt out of that and out of the people that I speak to, No one does that because I haven't thought about

it. Most people are low tech users of it. They just start using it and uploading things on the teams. This is GPT specifically when I say teams or enterprise, but the same thing on code as well. If you are on one of those accounts and by default, your content is not used to train the model unless you explicitly opt in as part of their feedback model.

So where I find this is quite amusing in a way that the companies that are most risk averse that say, you know, you must not use AI and we're certainly not going to pay for an account for you to have actually push everything under the radar and under the carpet to people that are still doing the work. But they're using a personal account which by default gets all of your company data up into these models. And so this is no time to be an ostrich with your head in the sand. La la la la la.

Nothing's happening like it's happening. People are using this and you can choose to ignore it, in which case you create a data risk or you can choose to embrace it. And actually the default is that your data is not going to be used to train the models.

So the question of that and this idea that, hey, I'm a CEO and I'm coming into this organization, venture capital, back to 250 people, and I want to have a real I push in this organization holistically across the company. So the question to you as a CIO, is it? What is it that I should do or what should I think about if I really want to transform this organization? Some of the most common first use cases when you talk to companies about using AI. It comes down to sort of productivity efficiency.

And there are two ways that you can approach that. So if you imagine you've got a 5000 person company and you can say, if we use AI, we could be way more efficient. And therefore, instead of having 5000 people, we could have 2000 people. And so that's quite a defensive cost saving increase. Our margins way of looking at it. The other is to say we're 5000

people. But actually, if we used AI, we can immediately act like a company of 50,000 people by giving everyone that we've already got just way more capability. Benioff was talking about there not hiring any more software engineers this year because everyone has a sort of wry smile. Benioff The great marketer, by his point is we don't need to hire any more engineers because we can get the engineers. We've got to be way more productive. And you're seeing that across so many companies.

So the first thing that I'd be canceling any senior executive is to say, right, just imagine that you could supercharge the people that you've got and get them to be way more capable to be able to work in different countries in different domains, to know what way more than they do today. How would you approach that? And a lot of this comes back to not trying to do anything overly

complex. So even a couple of years ago, when you talk about AI, people are, you know, we're going to use AI to completely disrupt our supply chain or we're going to use AI to completely rebuild our forecasting of footfall in a retail environment or something. It's quite a complex thing. Machine learning, data science. The big win right now is just get every single person that works in h.R. Is a financial analyst. That is a marketer that works in customer support.

They have work that they have to do as a human being. And i sort of split this down into three buckets. There's there's work that you receive from other people, so someone gives you some requirements. A customer gives you a request for proposal. Someone gives you a quote or a contract. Maybe there's a new regulation or new compliance rules. There's something that someone else has given you.

And as a human being, you have to understand that, analyze it, summarize it, work out, what are the most important bits and the risks. That's work that's given to you. You've then got the work that you actually do as a person. So, you know, I need to research accounts. I need to prepare a proposal. I need to manage a direct report. Those are the things that you have to do yourself. And then there's work that you pass on to other people. So I need to create a proposal.

I need to create some requirements. I need to create a summarization for my manager. I need to create an investor briefing, whatever that might be. And so these are very human things that you have to do. And how can you bolt on the skills where they are to be able to accelerate that much more capable at ingesting information and new requirements, much more effective at doing the job that I've got to do and a much higher quality of output and speed of output that I'm delivering to other people.

And in it, you talk about the ten x developer or the ten x engineer, you know, what is the ten x h.r. Admin. What's that? Ten x customer support representative. There's people that are kind of biochemically enhancing themselves with these tools suddenly way more capable than the person that sat on the desk right next to them. So I think that's a great vision. There are a lot of people where all CEOs, we're like, Yeah, okay, whatever vision, how do we do it?

And dropping it down a level, I can explain some of the ways that I use cloud or chat type of ended up buying Claud instead. Because of the ability to mimic your writing, I find that Claude creates my writing better than chatbot, creates my writing out of the box, so that's why I subscribe to it. I use it for things when I have to be creative and I'm not being very creative. So I want an image for a presentation that evokes a certain feeling. And in the old world, I used to have an image and I

couldn't find it. And I talked to our designer and he has the entire Adobe back catalog and he gets me some photo and it's done. But now I can either describe an image and have it generated or I can just say I don't really know what image I want for exceeding our target this quarter. You know, like I can only come up with like some really generic ones and I'll say, give me ideas of images that'll work and I'll come up with loads of images. And I'm like, No, I don't like that one.

I do like this one. Okay, yeah, that sounds good. Edit it this way. And now create. And then Claude doesn't I? She creates images, but Chatty Betty does. Or Dolly does. And so then I'll say. Write a prompt for Dolly that will give me the image. And then I'll put it in, create the image, and then edit between the two. And so and and all of that can be way faster than it used to take me to trawl through images to try and evoke the sensation that I want to evoke for customers.

Or I need to write the weekly report or the weekly customer update. And there's bits of it that are written for me in bits that are end, and I have writer's block. And I'll just say I want to talk at the end of the year, end of the week, I've actually created a project for this. It's the end of the week. I need to make people feel this way. These are some topics that I'd like to talk about. Everything I'm writing is really stupid. Write me a paragraph. That's not shit.

Obviously it's a better prompt than that, although sometimes that is the prompt and it'll at least give me something I rarely cut and paste and use that itself. But it's a way of like whenever I'm stuck with the too hard project, that is me having to think. I turn to Claude to help me think. And oftentimes it just releases the writer's block in me to produce something. And that's not an efficiency thing. That's actually like a thinking

thing. I mean, it is efficiency, but it's not like summarizing my emails.

I'll give you a specific example, which for SEO is a really sort of tangible way of using this. Anyone that's in a sea level role has got direct reports managing or leading those direct reports. Probably when you became a leader, you had these great aspirations of how you wanted to be, you know, a great leader and a coach and advisor.

And too often we get dragged down to just being a manager, a boss, and you're just going through the day to day exercise, take a code project or just launched projects just before Christmas. Create a project one for each of your direct reports. So I create a project for Bethany As and in that project I can upload certain documents. Now, I would add in my regular 1 to 1 notes, I would add in maybe the performance review that came out of Workday last

year. I might add in some chaos that have been set. Maybe there was a desk review assessment or something like that and maybe just do my chat. So my knowledge of working with you, but I understand a little bit about your sort of personal career aspirations or maybe personal goals around houses or family or whatever that might be. And having done that, you can then start to ask this project about those kind of creative things. You know, I'm preparing for my 1 to 1 with Beth.

Can you help me think through some stretch goals that would support what best career aspirations are? And I need to provide Beth as some challenging feedback around the way that she handles a recent project. Can you help role play with me? How I could deliver that feedback to Beth? Understanding Desk and the way that she is going to respond to that.

Suddenly you've got ten individual leadership co-chairs for each of your direct reports, and you know, that's useful for managers, but it's about thinking about what's the work that someone is already doing. They have to do it. And how do we use AI to support me and being better at that specific role? It might be managing someone. You might have a project for a customer that you're selling or supporting into. It might be a project for a country that you're looking to open up.

It might be a project, in my case, with a product that I'm trying to build. So it's thinking about breaking down. What are the bits of work that you're already doing and how do we give enough context around that to be able to give me what I want?

Can you give us one example? I need to ingest a wide swath of information, either summarize it or pull out some takeaways in terms of something that I'm trying to analyze or what have you. I'm going to mention a tool called Notebook Alarm, which went pretty viral in the autumn of 2024. It comes from Google, so you can just go to notebook alarm dot Google.com. And what they've done is take a very sort of intuitive approach to using

AI. So you can upload a source and a source could be a document, it could be a YouTube video, it could be a website. And what notebook will do is ingest that source and use that and only that to drive a set of AI outputs and responses. So it's not accessing the wider Internet. And a couple of things that it creates off the back of that document is a briefing document.

It creates some fake news and the to resistance, it creates a podcast interview between two AI generated hosts and an example that a provider may. A little link to some of the recording of this. But is the the new EU Air Act, which starts to come into force in February 2025. So in just a month or so from now? Now the EU Air Act is 144 page document, very dry as you'd expect, and difficult for most people, if not legal, to quickly figure out. Does this relate to me? How must I consider it?

So what you're able to do in this example is upload that document and then suddenly it generated for me a 31 minute podcast interview with these two hosts discussing it. But even better than that, you can join that podcast. So as you listen to it, you can say, Sorry, I don't understand that, or just help me to understand that. So suddenly you're having a conversation about this complex subject. Right, Right. So it's it's right. You can interrupt the podcast hosts and basically ask a question

midstream. You're like, Hey, podcast host, I have a question 100%. And they go, Hey, what's up? You know, what's your question? And then you also you say, you know, you talked about the air coming in, but like, what's the first deadline that I have to be considering? And they go, you know, well, February the 2nd, 2025. And so you're having this conversation now. This is about the Air Act. But think about all of the things that you have to consume.

Maybe it's the Mssa that's been sent to you from a supplier.

Can I give you an example of what we're looking to do? So we're also it's interesting that you've mentioned it because my first experience with it was yesterday when one of our Gen I people pulled together information on ideas of what we're going to do as a podcast. Not only does it do it as a podcast, but it can also play on your voice. So it's his voice doing the podcast. And what we're going to experiment is, do you remember years ago, I can't remember which company it

was. It was a company that I worked with that would give all of their sales team CD's to listen to for sales enablement while they were on the road. It was either Microsoft or IBM, I don't remember. And so what we're going to experiment with is taking all of our sales enablement content and turning that into an internal podcast for our sales team so that while they're on the road, not that people are on the road as much as possible can consume that content.

And we're looking at a couple of tools to basically create an internal playlist where you can find all the content as a podcast. But I didn't realize about the interactivity, so that's super cool. So yeah, it's only just released maybe a week or so ago that they put the interactivity in. It's amazing. There's a maybe a two second delay after your question wallet reconfigures, but it's amazing. So my one question is Gemini. It was shit have heard that Gemini two is not shit.

Have not played with it yet. We're going to experiment. Worth it or not. So when I'm talking to clients, I talk about three things. I talk about Gemini chat and Cod as being those two providers that people should be thinking about and definitely want to be watching, especially if you're a if you're a Google workspace customer, then that has to make sense. I just say on that, it's one of these things that is so different to SAS five years ago.

If you're buying a CRM system or you're buying an ERP system, you go out to market, you pick which one and you pick one. That is not the way that it works here. People are not going out to market to pick one Al-Alam. You will have this composable architecture. You will have multiple, maybe some closed source, maybe some open source, and you'll use different ones even within the same workflow. You know, this is doing the research and this is doing the writing.

And so from a CEO perspective, from a CIO or CTO perspective, I think about Lego blocks. What you're building is a box of Lego that people can build what they want at the time. This isn't like a traditional procurement, so you would have Gemini plus the others in your box of Lego.

If our listeners can only take away one thing from today's conversation, what is it? It is to plant the seed of use and every single one of your business teams. I say business to me. This is not an IT project. So for every one of your teams, how can you give them just one way of using publicly available tools called Gemini to improve one aspect of their work? Because it is those people that are closest to the business problem and the business process.

And if you can show them one way of using this, that is where you will suddenly find all of these new ideas come from. So figure out how to get this groundswell going at the core of your business and bring that back up to the top rather than going top down. Lovely. Thank you, Charlie Cohen, for joining us on the operations room. If you like what you hear, please subscribe or leave us a comment. And we will see you next week.