Listen as Gareth Llewellyn shares Brew Digital’s framework for team-wide AI adoption—closing internal skills gaps without compromising client data.
In this episode of Marketers of the Universe, we launch a new miniseries on AI, captured live from Brew Digital’s recent event in Norwich.
Managing Director Gareth Llewellyn takes the stage to address the practical operational challenges of AI adoption within agencies. He shares our six-step playbook—a framework designed to upskill teams, ease workplace anxieties, and ensure no one gets left behind.
Key topics covered in this episode:
Managing the operational hurdles of AI, including data security and rising subscription costs
Protecting brand quality and avoiding low-value 'AI slop'
Transitioning your agency from ad hoc experimentation to structured, measurable AI integration
Listen now to discover how you can move past the hype and implement a sustainable AI strategy for your business.
Further reading
You can download the six-step playbook here.
Is your B2B marketing team ready for AI? The case for equitable skills training
AI will empower a new generation of creatives
Your first-party data is the ultimate defence against AI browsers
Episode Transcript
Tom Inniss • 00:08
Hello, and welcome to Marketers of the Universe. My name is Tom Inniss, and I'm the content marketing manager here at Brew Digital. Today, we've got something a little bit different for you. So recently, Brew Digital hosted an AI event in their office at Norwich. And a number of senior leaders within the team did talks around how they feel that AI to impact the world of marketing, impact the world of work, and what we are doing to prepare for it. So over the next couple of episodes, you will get the recordings from those sessions. Today, we have Gareth Llewellyn, who is the managing director at Brew Digital, talking about how we at Brew are implementing AI across the whole team and ensuring that nobody's left behind when it comes to that rollout. It was a really interesting talk, and hopefully you enjoy it. See you on the other side.
Gareth Llewellyn • 01:23
I want to talk to you about what are the challenges that we faced at Brew Digital when it came to AI, and really about how we close the skills gap when it comes to AI enablement that we're starting to see. But more specifically, I really want to talk through the plan, the playbook that we put in place and share that with you so that if you're seeing similar challenges in your business, then hopefully there's something useful that you can take away. So first of all, the problem we're really dealing with is, obviously, as we all know, the AI tools have really come out of nowhere over the past couple of years. They've progressed super quickly. And like most businesses, we're starting to use these new tools that have amazing capabilities. They're moving so quickly. And there's really a level of adoption sort of faster than any other technology that we've had before. And really in our business, we want to be using those AI tools. We want to make sure we're making the most of them. And really we want to make sure we're not getting left behind and being overtaken by our competitors. But the big challenge we've had with the speed of the AI rollout has been quite a lot of the staff don't know where to start with those tools. Maybe they're experimenting with them a bit at home. Maybe they're not. Or some are using the tools, but then a bit nervous about it because they don't know what tools they're allowed to use. They don't know how they're allowed to use them, what kind of data to use with them, and things like that. And also, there is a lot of worry as well, in terms of, is there a security risk? What company data can I share with this? But also, things like reputational risk as well, because if we're suddenly turning out loads of AI slot, that's gonna have an impact on our business. And then there's the cost as well. I'm sure we've all heard the sort of horror stories about people running up huge AI bills and things like that. So again, how do we sort of keep on top of that? So to address these problems, what we wanted to do is really put together a playbook, a plan, that clarifies how we want to use AI as a business, ensure people know what tools they can use and how to use it, but really the key thing is getting everyone comfortable with using the tools and understanding the different ways that we want them to use it. And really one of the key things is, is about closing that skills gap and getting everyone to a good base level of understanding with the tools, a base level of experience and comfort. With the speed of change that we have with AI, we really wanted to make sure that whatever we do is really adaptable so that we can change as we need to. So as new tools come along and as existing tools maybe get dropped because they get out of touch or whatever. And then really importantly, we wanted to be able to measure what impact AI was having on the business. So we want to make sure that, OK, if we're spending all this money on AI, whatever tools we're using, what are we getting in return for that? What cost savings or quality improvements or whatever? So what I'm going to do is take you through the six-step plan that we had. The very first step was around getting the exec team, the senior leadership team, fluent with AI themselves. Because even within that small team, And we found there was a very big difference in terms of people's skills and their level of comfort using the tools. Some of us were already using the tools day in, day out, and others were still really wary of doing anything. So what we did, first of all, was make sure that we all have some time to really start using the tools, start experimenting, try different things out, and get comfortable with what we're asking everyone else to use. We really encouraged everyone to, like I say, try different tools, but also do some of them slightly more advanced things. Not become really expert level, but at least try things so they're getting a good understanding of the capabilities. So we suggested things like building a custom skill, setting up an agent or an automation, so again, they can start to understand some of those capabilities. Do things like vibe coding a small app, or maybe like the complex spreadsheet or something like that. All of these sorts of tools that are now, all of these sorts of things that are really easy to get AI tools to do and can really give you an insight into the capabilities. And through that, also start to understand what are the security risks. What kind of data do we feel comfortable sharing with AI tools or not? So step two is around surveying the rest of the company about what they're doing already with AI. Because again, it was one of those things- some people are super comfortable, some people weren't, and so we were seeing a big skills gap. Now to do this, obviously you could run quite a formal survey, use like Google Forms or something like that to ask a bunch of questions. However, we specifically chose to not do that. We didn't want to go to all of our staff with something that looks really formal, particularly given that some people might be worried about the impact of AI. And we didn't want people to suddenly think, "Oh, wait a minute. I'm being surveyed about AI. Is my job going to be replaced or anything like that?" Because that absolutely wasn't the intention. So we wanted to take a softer approach with it. So instead, we had our line managers discuss these topics with their staff in one-to-ones and things like that. So we didn't get hard data, but we had that fed back to the management team. So really, we were asking about if people were using AI already or not in their work. If so, what tools were they using? Were they using work or personal accounts? Because again, that was something that was a big mix. A lot of people using their personal Google accounts for things and stuff like that, which obviously isn't ideal. We were asking them what they're actually using it for. So which parts of their jobs were they experimenting with AI for? And then also what benefits they were seeing. And also at this stage, it was just anecdotal feedback we were getting, but the main benefit that was being reported back to us was time savings. But again, none of that was really being measured. And then also we wanted to find out what people's fears and worries and that were so we could make sure that we address that in whatever we did next. And some of the things that we found out was actually the main thing people wanted was guidance. They wanted the stuff that we were putting together because they wanted to be using the tools but they had a lot of uncertainty about were they using the tools in the right way, were they using the right tools. So they wanted some guidance, some training and basically just know they were doing the right thing. And something else that came out of the survey, which was really interesting for us and really kind of gave us confidence in what we were doing, was we didn't get any resistance when we were talking about using AI more and using it more systemically. If I think back 12-18 months or so, we definitely had a good number of staff who were really resistant to AI, really anti using it because obviously there was lots of uncertainty about it, people obviously worried about the impact on their roles and things like that. That general feeling seems to have definitely softened over the past 18 months or so, to the extent that we didn't have anyone at all sort of having negative reactions to these sorts of questions. So the third step was for us to create some AI guidance. And this, so this exists as essentially a section in our company handbook, which is on our intranet, so available for all staff and new starters and things like that. And really in there, that's just a case of writing down a bunch of the key guidance that people want. So what tools we're encouraging people to use, because obviously they've got access to a whole bunch through their company accounts. Really giving them some guidance as to where we see AI adding value to our business, but making sure it's not everywhere. We still need to keep the human in the loop and what that means for us. Being clear about which data we're happy for people to share with AI tools versus what we're not. So we're really trying to minimize the security risks. But then also, when we're talking about the tools, also being clear about how the costing works. So obviously, some tools are on monthly subscriptions where we know it's a fixed cost. So people can be confident using those tools as much as they want. And the worst thing that will happen is they might get throttled a little bit in terms of their usage. But they know they're not going to run up big bills. And then for the tools that do have token-based pricing, being clear about that, but also pointing people to where they can find out how much they're spending each month so they can keep an eye on that and make sure they don't have any kind of runaway agents running up big bills and that. We also include things like how to access some of the shared skills, which is one of the things that we're working on more improving how that works. But also just some general hints on how to use AI well. And also pointers on how to get help with the AI tools and how to stay current with- obviously, things are changing so quickly. And particularly for when it comes to getting help, this step four is probably one of the most- probably the most important step here. And this is all about providing support to everyone within the teams. So what we wanted to do was assign a number of what we call AI champions. So these are people within the teams who are basically there to help everyone else. So we wanted to make sure we had people throughout the business, and not just within the technical teams, but across the whole business. And we wanted to pick people who are naturally curious, who ask questions, the kind of people that are already experimenting with the tools and are not afraid to try stuff out and just sort of have a go with stuff. But really importantly, because they're there to support everyone else, they need to care about people, about their fellow colleagues. They need to be really approachable and easy to talk to, so that other people don't feel nervous going and asking what they feel might be a stupid question or a real basic question, things like that. So for the AI champions, they don't need to be someone who's super technical or the person who's the most advanced with AI or anything like that. And in fact, I think it's probably best if they're not, because if someone goes to someone super technical with a very basic question, they might just get overwhelmed with a super technical answer, where actually they just need some basic handholding or whatever. And one of the key things we did is made sure that they had some protected time in their week to account for this work, so it wasn't being impacted by their regular work. So we allow them half a day a week for this, essentially spread over the week as they need it to support everyone. Now, step five, and this is something that the AI champions can really help with, is to roll out micro-training. So like I said, when we did the survey, training was something that people wanted, that they wanted to learn. But as I'm sure many of you know, developing proper training courses takes a long time, takes time to deliver as well. And more often-- well, particularly with the speed of change with AI, we thought if we do formal training courses, by the time we roll it out, it's probably already going to be out of date. So instead, what we did is create micro training. And this is something the AI champions have helped with and have been doing. And these are really just very short videos that demonstrate one or two concepts. So they're normally just like a screen share. There's someone talking over the top of it. So they can literally just record it on their laptop with Loom or Zoom or any of those sorts of tools. And just talk about one or two very basic things. So for example, we've got one that simply goes through the steps of going through our single sign-on tool to go to the AI tools that we already have with our company accounts and go through to do your first prompts, essentially. So real basic how to get into the tool, how to start using it. We also have stuff around understanding the different types of the LLM model to help people understand where they should be using Sonnet versus Fable and things like that. And again, that's an area that's changing so quickly. There's new models coming out every month. We don't expect everyone to keep up with that stuff. So instead, we can have this micro training that people can refer to whenever they need to. We also have stuff like what different tools are good and bad at. So Rovo, for example, is the AI tool that's built into Atlassian apps, so Confluence and Jira, things like that, which we use a lot. It's really good at some stuff, like search is quite bad currently at other stuff. So again, having a bit of guidance as to where it works well is really useful. So the key thing is with these micro training videos, they're super quick, cheap, and easy to produce. So there's very little kind of invested in them so we can easily scrap them when they come out of date or replace them or whatever. So we can just update them whenever we need to and add more stuff as sort of people request specific things that they want more information on. So that's the first five steps of the plan. And those first five are essentially considered to be the sort of phase one, putting everything in place to be able to bring everyone in the business up to that base level of comfort, understanding, and experience with the AI tools. The final step, step six, this is more of a phase two type activity. And this is about building on that, and then essentially applying it to the business so that we can then really start to make improvements that are more business-wide. So here, this is essentially a case of picking one process or workflow or whatever that's where you think AI can help. And this is one of the areas where, again, the AI champions can be a great help with- they can be a big part of this. What I suggest is that you measure what happens today with that particular workflow or process, understand where the pain points are, and then see how long does it take, or are there any quality measures that apply to it. From there, build an AI-based minimum viable product. So it doesn't need to be the complete thing, sort of like an 80% or 90% solution. It doesn't need to be perfect, because initially it's just to kind of trial it out on a couple of projects. And then it really is testing it on a couple of projects, but really importantly, measure how it performs in those projects. Is it saving time or is it improving the quality or improving the consistency, those sorts of things. If it is, brilliant. Perfect it, do a bit more work, take it to the 100% solution, and then scale it out and widen adoption, apply it to all your projects. But really importantly, if it doesn't bring improvements, then kill it and move on to something else. We don't want to be forcing AI onto processes for the sake of it. We only want to be adding it where it's really adding value. So ideally, over time, by doing this and doing this across a number of different process areas and things like that, you'll start to understand the areas of your business that AI can really add value and where it doesn't, where you still need to really focus on the human doing the work. And then hopefully, if you've been doing the measurements, then for the processes that do work, that do get scaled up and rolled out further, you can really measure what is the improvement that's happening. Is it a time saving? Is it quality improvement? But then you're starting to see, OK, we're spending this much on AI, but we're also saving this in terms of costs, or we're improving profit, or quality, or whatever. So we can really get the ROI for our investment. So that's it. That's the six steps. We have, like I said, we've got a little handout here that just gives a summary of those, and essentially breaks it down into a 12-week plan with that phase one sort of taking the first sort of six, seven weeks, and then the phase two sort of going on beyond that. Obviously this is what we've done for Brew Digital, won't be perfect for everyone, but hopefully there's some stuff that will be useful for your business. Thank you very much everyone.
Tom Inniss • 18:27
That's all we have time for today. If you are interested in that handout that Gareth mentioned, you can find it on our website broodigital.com/resources. There you will also find a myriad of articles and other podcasts that we have written and/or recorded over the years. There's something for everybody there from design to web development to digital marketing and strategy, so definitely something you will want to check out. Thank you for everybody's hard work in making this podcast possible at Brood Digital. I really appreciate it and I couldn't do it without you. And thank you, listener for continuing to come with us on this journey as we try to discover what the format is that works for marketers of the universe. Just to throw forward to the next episode we've got Rich Harper, the head of digital marketing at Brew Digital, talking about AI, slop, and how your clicks are likely to change in this new era of search and AI. So it's definitely something worth keeping an eye out for in your feed. Thank you so much for listening and have a wonderful day!