Most growing businesses are past the “should we look at AI” conversation. The tools are there. People have had a go. Some are using them regularly, some aren’t, and nobody’s quite sure what’s happening across the team.
That gap between having access and actually adopting – consistently, usefully, across the whole organisation – is what this session was about.
On 4th June, we ran a free webinar for SME leaders on exactly this. I walked through the blockers we hear about most often, and the step-by-step process we used at The Coders Guild to get everyone on the same page. Not just experimenting, but genuinely embedding AI into how we work day to day.
The recording is below. Underneath that, the key points from the session.
What’s actually getting in the way
Across the businesses I’ve spoken to over the past year – e-commerce, consultancies, marketing agencies, retail, hospitality – the surface-level problems look different. But they keep coming back to the same few things.
Overwhelm is the big one. There’s too much to keep up with, too many tools appearing every week, and finding the time to actually prioritise this feels impossible when you’re running a business. Alongside that, there’s genuine confusion – should you be using Copilot in your Microsoft 365, or moving to Claude, or sticking with ChatGPT? Nobody wants to pick the wrong thing and waste time.
Then there’s what I’d call the team going off in different directions. Some people are on Claude, some on ChatGPT, some on Perplexity – doing different things, in different ways, with no shared approach. You’ve lost track of what’s working. And underneath all of that, there’s a quieter fear: if we automate more, are we just going to AI ourselves out of our own jobs?
That last one matters more than people admit. If you don’t address it before you start, it’ll kill your adoption before you’ve got anywhere.
Frame it around people, not profit
The thing that changed everything for us wasn’t a tool. It was deciding what we were actually trying to do.
We didn’t mandate AI use. We didn’t put KPIs against token spend or efficiency targets. What we did was create a shared goal everyone could get behind: more time for people-to-people exchanges. More time with learners, more time with customers, more time thinking together rather than grinding through admin.
Framing it that way – around what it would do for people, not what it could save the business – changed how the team received everything that came after. It’s the difference between “we’re cutting costs” and “we’re making your job easier and more fulfilling.” Same tools. Completely different starting point.
People, process, tools – in that order
This is the thing I keep coming back to, in this session and in most conversations about AI adoption. Most organisations go backwards. They’re sold a tool, or they already have one, and the question becomes: how do we get everyone using the AI bit in Salesforce, or Teams, or whatever system they’re already in?
That approach puts the vendor in control of your direction and your spend.
Going people first means starting with what the team actually needs. What are their fears? What’s the domain knowledge – the stuff they carry about your customers, your edge cases, your processes – that AI can’t replicate? What do we want to do for these people?
Once you’re clear on that, process follows: where are the bottlenecks, what’s the day-to-day flow? And then – only then – tools. And at that point, the tools are almost arbitrary. They can change. If the people and process thinking is right, swapping out one tool for another doesn’t unravel everything. The strategy holds.
How we actually did it
We went through four stages. Not a project with a start and end date – the last stage is permanent. But it had a structure.
Commitment. We’d been experimenting for a while. Different people using different things, learning as they went. That was fine – it’s how you find out what’s useful. But at some point you have to stop experimenting and choose a system. Commit to a way of doing things and start rolling it out.
Pilot. We picked sales and marketing as our starting point – a small group, real work examples. We looked at what we thought would be useful, got feedback on what actually would be, and refined from there before taking it wider.
Build. Rolling out to other departments, but not with a blank canvas. We built in structure – guidance on how to use the system, starter documents, a shared approach – so people weren’t just handed tools and told to get on with it. We paid particular attention to cross-team handoffs, because that’s where most friction lives in any business.
Embed. This is where we are now. It’s not a project any more – it’s how we work. The focus now is on people finding their own wins and sharing them back with the team. Asking for feedback. Iterating. Making it part of how we think rather than something we have to remember to do.
One thing worth mentioning: our quarterly full-team meeting became part of the rollout. Rather than sending people off to an onboarding course, we ran a session together. People set up their custom instructions, connected things they’d been doing in other tools, and then – crucially – worked on something real and useful to them, not a tutorial. They left with actual work done, not just knowledge of a system.
The competitive advantage isn’t the subscription
We’ve committed to Claude as our main shared system. But if Anthropic changed the pricing tomorrow to something out of reach, we could move. Our knowledge base, our skills, our context – all of that lives externally. We could take it somewhere else.
That’s the point. The value isn’t in the tool. The value is in what your team knows: your customers, your workflows, your edge cases. AI doesn’t have that. Your people do. The advantage comes from giving your people the skills to use AI effectively with that knowledge – and that’s something nobody can copy from a subscription page.
The efficiency gains are real – research consistently points to somewhere between 10 and 55%, and our own experience has been 15 to 60% depending on the process. For a business turning over £1.2m, even the floor of that is worth around £120k a year. But the more durable thing is what your team builds over time: the ability to work differently, and keep improving as the tools change.
Talk to us about your team's AI Training
Not sure which route is right? Book a free discovery call and we’ll point you in the right direction
Francesca Roberts
Head of Commercial
Questions from the session
We had some good questions come through in the Q&A. A few worth sharing here.
On time: The time investment feels significant at the start. But it compounds quickly. The first things you use AI for will give you time back the following week – and the week after. You don’t need a full digital transformation plan. You need a first step that saves you time, then another one. Rob blocks out a Friday window for it; I’d recommend the same.
On cost: Going tools-first is the expensive route, because the vendor controls how much you spend. Going people-process-first means you only pay for what you actually need. We don’t use the AI features built into most of our tools – they don’t match what we get from Claude doing the heavy lifting, with specific tools for the things they’re better at.
On jargon: Use the model to explain the jargon. One prompt we gave the whole team: “Assume I don’t know any of the technical terms. Walk me through everything slowly. If you use a word I might not understand, ask me.” You can teach the tool to meet you where you are. Rob’s go-to prompt is “ELI5” – explain like I’m five. Works every time.
On ethics: The ethics of AI aren’t separate from the ethics of how you run your business. Having a new tool doesn’t change your responsibilities. The responsibility lies with leaders – and that’s true regardless of which tools you’re using.
[0:11] Hi everyone and welcome to another session in our series of AI webinars.
[0:15] Today we will be tackling something a lot of teams are quietly struggling with. You've got the access to the AI tools. You've been told how to use them, but the adoption hasn't quite stuck. So Crispin is going to walk you through a practical guide to actually making that shift happen. By the end, you should have some key takeaways that you can bring to your team and try out this week. If you've got any questions, drop them into the LinkedIn comments and Francesca and Alex will be there to answer live and hopefully we'll have enough time at the end for a quick Q&A. But I'll hand over now to Crispin to get things going.
[0:49] Crispin: Cheers. Thanks a lot, Rob. And yeah, so as Rob says, this is about getting your whole team to actually use AI in a productive way. So not just handing them some tools and hoping for the best. And also what I'll do later on is walk you through the exact rollout process we used at The Coders Guild to get everybody on the same page and using AI effectively.
[1:17] So we've got this graph that Anthropic put out earlier this year about the actual adoption of AI in terms of tasks - not in terms of people, but in terms of the things that we're actually doing with AI. That's the red section. And the blue is the theoretical possibilities - not in the future, but right now, what AI is capable of delivering.
[1:43] And the research I want to compare this to: there's a lot of research out over the past 12 months about efficiency gains in AI adoption, and all of them sit somewhere between 10 and 55%. Accenture, McKinsey - multiple studies are all pointing at this same bracket. So if you take the 10% floor - that's really easy maths. If your company's turning over £1.2 million, that's £120k benefit in a year. And what could that mean for your business? Where would you repurpose that money?
[2:35] To make it more real - we've been measuring our efficiency gains as we've been rolling out automation and AI in our operating model. First projects were immediately giving us 15 to 25%, and some of the more recent projects we're doing got 50 to 60%. And that's just through going through the process, understanding what the process is, making adjustments, and using the tools that are readily available to us. So in the next 20 minutes or so, I'm going to show you how we did it and what that means for you as the leader of this transition.
[3:27] So who am I to be telling you about this stuff? I'm Crispin. I've done 20 years in software and digital strategy. Spent the last 10-plus years running a training firm, completely obsessed with how people learn complicated technical things quickly. Also working in the apprenticeship field - how we get people to show value back to the business as fast as possible. And last year a lot of this work had me working with Skills England and the government to develop the new Level 4 apprenticeship in AI and automation. To do that I spoke to loads of SMEs about their needs and their blockers. We also had research commissioned through Innovate UK - to identify what would turn into a rollout plan for AI adoption as a 12-month exercise. So I know a fair bit about what's getting in the way.
[5:03] So where do we start? What's actually getting in the way? I spent a lot of time last year talking to businesses across loads of different sectors - e-commerce, policy, consultancies, marketing agencies, retail, hospitality, all different walks. And loads of different problems on the surface, but the thing we come back to every time is: how do we take that first step? How do we get past the overwhelm - which we all have as people running businesses? The overwhelm is pretty intense, and finding the time to prioritise these things is probably our main blocker.
[6:07] Another one for us was coordinating a project around the adoption of technologies in a useful way. You could be curious but stuck. Maybe you're not sure where to actually start. Maybe you don't know if you should go with Copilot in your 365 or move to Claude or whether ChatGPT is working or whatever - confusion around the tools. Maybe your team has gone off in loads of different directions - some people on Claude, some on ChatGPT, some on Perplexity or Deepseek - doing different stuff in different directions and you've lost track of what's happening or what's useful. Or maybe you had a go, didn't get the responses or the gains you were looking for straight away, and then it felt underwhelming and you kind of moved on.
[7:15] And then underneath all of that, there's a quiet fear - especially when you start talking to your team about using AI more. The rhetoric across media and social media is: are we just going to AI ourselves out of our own jobs? Are we going to replace people with robots? Whether or not that's going to happen is kind of immaterial. The main thing is: if we don't address that fear before we start, it's going to kill your adoption before you've got anywhere.
[7:59] So what do we do about it? The main steps towards team-wide adoption came from creating a mandate. And this is what's really good about it: we're not thinking about tools. We're not mandating use of AI and automation. We're not putting KPIs against people's token spend or what efficiencies they've made. We're trying to do something we can rally around as a group of people. And the thing we came up with is: we're looking to create more time for people-to-people exchanges. More time with our learners, more time with our customers, more time discussing things together. We framed it around people, not profit. Not where we can cut costs, not reduce headcount by three people - but what can we do together where we can make six people's jobs easier, better, more fulfilling. And I think that framing changes everything about how your team receives what comes next.
[9:45] And that leads on to what is my mantra - the thing I regularly bore everyone with - which is: people, process, tools, in that order. This was made popular in the 70s and 80s and described as a triangle, but we always think about it in a linear format: who are the people, what are the processes that they need to do, and then what are the tools that make those processes better for those people. People tend to go about this backwards. You're sold a tool, or you've already got a tool - you've already got Salesforce - and what's the AI bit in Salesforce that now everybody should use? But that's backwards.
[11:00] If we go people first, we're looking at what do the team need, what are their fears, what's the domain knowledge that's the special sauce, what are the things that AI can't replace, and what do we want to do for these people? And then once you've got that, that's your advantage - because your team knows your customers, your edge cases, your business. That reframes the whole process. Then you get to process: how are we going to do this day to day, where are our blockers, where are our bottlenecks? And then tools. And to be honest, the tools are almost arbitrary. They can change out. If the people and process bit is right, the strategy won't change significantly.
[11:59] And that brings on to practice - how do we use this regularly on real work, and reflect on that practice. Which is how we work as an organisation and how we deliver training: centred in practice, reflection, and feedback.
[12:26] Something worth highlighting for younger team members or recent graduates: a lot of people new to the workforce - especially if they've come from education - have been schooled against AI, because the education system has not historically celebrated its use. So people tend to be more wary, more fearful of AI stopping critical thinking. But the same went down with calculators in the 70s when they got cheap. "Oh no, it's the death of maths." Turns out it wasn't. It just made us better at maths.
[13:19] So coming back to the team and the leadership part: this is about creating psychological safety for people to try things, to voice their fears, to get stuff wrong and try again. It's not a big change, but it is one that needs practice and reflection, and constant attention.
[13:41] Let me put this into the practical frame of what we did at The Coders Guild.
The first stage is commitment. We knew loads of people were using different things and that was fine - we'd done our experimenting. But the commitment was: stop just experimenting, choose a system, choose a way of doing things, and start rolling it out.
[14:21] Then a pilot phase - we tested the initial things based on our hypothesis of how we'd use these things. We divided by department: started with sales and marketing, so Francesca and Alex, with a small group. Looked at real work examples. Took what Rob and I thought we should do, and got feedback on what would actually be useful.
[14:56] Then a build phase: rolling out to other departments and putting in some foundational work so we weren't going with a blank canvas - everyone use this system, do what you want. We gave people guidance and a structured way of doing things. We paid a lot of attention to where the cross-team or cross-people opportunities were, because that's where most of the main friction is in business: when things go from one team to another. And that also creates a really good environment of feedback and reflection, because people are able to tell others how they do their work and how that relates to everyone else.
[16:10] When we did the rollout, our quarterly team meeting - where everyone comes to the office for our full team meeting - we designed a bit of training for it. Rather than sending people off to do an onboarding course with Anthropic or whatever tools we had, we devised a session where we wanted everyone to get comfortable with the tools first: setting up custom instructions, understanding how to move things from ChatGPT to what we're doing now. Then we had group tasks that were specifically useful to everybody's day-to-day. Rather than a tutorial step-by-step process, we left it open. We created starter documents for a piece of knowledge or a skill that we'd put into the system, then went through the process of how we interact with the large language model to decide whether it's a knowledge-base item or a skill and how it gets developed and installed. So people got actual work benefits out of that training session - not just knowledge of a system.
[17:59] And then the last part: embedding. This is not a project with a start and end. It has a start and we've started. But now it's much more about communication - people finding their own wins and sharing them back with the team, asking for feedback, growing that, and iterating on what we've already done.
[18:26] Which brings me back to the tools. Now that we've done this process, the tools are kind of arbitrary. That's derisked it a lot for us. We've committed to Claude as our main shared system. But if tomorrow Anthropic whacked the price of Claude up to something out of reach, we could swap out to something else. Our knowledge base, our skills - those are stored externally. We could take them to another tool. The system I've described is tool-agnostic. It's a leadership process. It's about: what are the things that make your business valuable? And that's the domain knowledge your team carries. Your customers, your workflows, your context. AI doesn't have that. Your people do.
[19:35] So giving our people the skills to use AI effectively means we've got something nobody can replicate - and that is competitive advantage. It's not in the subscription.
[19:52] The thing I keep coming back to, after spending the last few years talking to businesses, building the qualification, and running our own rollout: it's not about AI. It's not about the tools. It's about people and how we invest in them. The businesses in this room are the ones that this matters most for. We don't have enterprise budgets or armies of consultants. But we do know our businesses. We know our customers and our team. That's our advantage. AI is just what we can use to get it at scale.
[20:53] The operating system has changed. The question isn't whether to get on board - it's whether your people are going to be the ones who define what comes next, or catch up later.
Q&A
[21:30] Rob: We've got a few questions mainly around what's blocking people at the moment. Pam asks that her issues were time, understanding the best tools, and expense. Can you give some hints on how it doesn't have to be all time-consuming - and it doesn't have to be expensive? Our team is on an enterprise tier and it's not costing thousands of pounds a month in API calls.
[22:08] Crispin: On the time blocker - in all the work I've done with apprenticeships over the years, my main blocker for getting people to take on apprentices was always "the time I'll have to put in is too much." So we try to make everything about how quickly we can get a person to deliver value back to the business. We try to get it down to a three-month turnaround.
It's similar here. The investment you get from taking a few first steps - looking at the quickest time-saving wins with your team - you're getting that time back immediately, and it compounds week on week. You can free up your own time without having a massive digital transformation plan. Just start using the tools a bit. Start talking about how you'd take your first steps - and you'll get that time back the next week and the week after.
On expense: that can get expensive really quickly. And again the tools-first approach loves this - if you're already invested in a system and they say "here's our AI tier, please pay for it," then they're in control of how much money you spend. If you go people-process first, you might not need their AI system. We genuinely can't think of an AI feature built into one of our tools that we actually use. Claude does the heavy lifting. The built-in tools don't match it.
[25:43] Rob: On the time thing - especially in the earlier stages of using these tools, there is an investment. I treat it as a personal investment in my own understanding. I block out a small window on a Friday to upskill on these tools. The more you put in, the more the efficiency savings build over time. It does return quickly once you've got the processes right.
[26:35] Rob: Gary put "vibe coding" as a blocker - or at least something that's distracting. We had a lot of new tools coming out in Q1 this year, Anthropic were releasing things weekly, and it was almost impossible to keep up. How do you not get swept away?
[27:22] Crispin: On vibe coding - you can go sign up to Lovable and away you go. But to do it well - what I'd call meta-coding rather than vibe coding - you're still coding with an AI partner or agent, and the more you control and scaffold what they're doing, the better results you get. The best vibe coders are coders. The best AI designers are designers. That pattern holds.
On Gary's follow-up - the impact on the pipeline of people coming through - I think it's a leadership question. If we're expecting there to be a future for our businesses and our economies, just living for now and not teaching people how to do this work is a folly. It's similar to what happened when Drupal and WordPress arrived and people said "you don't need to know command line any more." Libraries and frameworks followed. It didn't stop people joining as juniors and learning the craft.
We're in a strange state where people feel frozen - they can't hire junior people because the output from a team of developers is so much greater now. But we're still recruiting software developer apprentices. People are still hiring and still making pathways for people. The tools have changed dramatically, but in quite a similar way as they've done numerous times in the past.
[31:06] Rob: One more from Pam - jargon and ethics. Both are real concerns and you see a lot in the media at the moment.
[31:17] Crispin: On ethics - it doesn't make it more difficult to be ethical just because you have a new tool. The ethics are held by people. The responsibility lies with leaders. That's true regardless of the tools.
On jargon - the good thing about having a large language model partner is you just ask them: "What does this mean? Give that to me without jargon." One of the prompts we sent out to the whole team for the Claude Cowork starter - part of it says: "Assume I don't know any of the jargon words. Assume I don't know anything about these things. Walk me through everything slowly. Check my understanding as I go. If you use a word I might not understand, ask me." You can school your large language model to help you with that learning.
[32:48] Rob: One of my most common prompts is ELI5 - Explain Like I'm Five. It's the best way to get common-sense explanations rather than acronyms and jargon. Right, we've run a little over so we're going to wrap up. Please keep the conversations going - Francesca and Alex will be available to respond to any queries. We'll see you at the next webinar on the 18th of June. Thanks very much Crispin. Bye now.
End of transcript
