[00:00:00] Rob Brighton Good afternoon and welcome to today’s session, where we show you a little bit under the bonnet, some of our secret sauce, of what we’ve been building at The Coders Guild in our journey to become an AI native organisation. As always, Francesca and Alex should be in the LinkedIn comments section, so if you have any questions as we go along, please drop them in and we’ll try to get to them at the end. I’m on co-hosting duties again today with Crispin. Good afternoon, Crispin.

[00:00:42] Crispin Read Good afternoon. Welcome again, Rob. Sorry for press-ganging you into another webinar, but it’s important, I think, because a lot of the things we’ve been talking about recently are operational, so it’s great to have you here.

[00:01:01] Rob Brighton Thank you. It’s good to be here.

[00:01:03] Crispin Read And it’s good that we can make it in around the additional parent admin that comes along with the holidays starting, but fingers crossed we can make it through the whole thing without a request for crisps or extensions on screen time limits.

[00:01:22] Rob Brighton / Crispin Read It’s very much holiday mode at the moment, hence the obligatory plaid shirt, a bit more informal this week. We’ll try to keep it as calm and chilled as possible for everyone.

[00:01:39] Crispin Read So, a little bit about who we are. I’m Crispin, CEO of The Coders Guild, and Rob runs operations for us, so this is very much an operational discussion in AI adoption. A bit of history on how we got here: The Coders Guild started as a grassroots initiative to train software developers in things that matter. That hobby project got out of control and became the real business, but the things that always stayed with us were my obsessions with how people learn really technical things really fast, and how we align vocational education to business value. Those two things have led me to a path where all the skills I’ve been building up for a decade are now super useful.

[00:02:52] Crispin Read In particular for this webinar, this is a thing Rob and I talk about a lot: how the organisation is evolving. A lot of the work I did last year was with Skills England, the Alan Turing Institute and the DfE, creating the Level 4 AI and Automation Apprenticeship standard. In doing that I spoke to dozens of high growth SMEs about their barriers, the projects they wanted to do, and where the business was going. Innovate UK also did a bunch of research for us and collated existing research into how we make this possible.

[00:03:55] Crispin Read From the past six or seven months of helping businesses raise AI capability in their own teams, really flexing the domain knowledge their existing people already have, I’ve built a matrix for understanding a team, an individual or a business’s appetite, readiness and maturity in AI adoption. It’s a six point scale, similar to ones Anthropic and OpenAI have put out. You start off watching, not really doing anything. Then you’re experimenting. Then you’re adopting, using tools in a more strategic way. Then you’re connecting more things together, other systems into your AI driven work. Then you’re embedding, constantly looking for improvements, ways agentic and automated work can improve productivity.

[00:05:47] Crispin Read That’s usually where these frameworks stop, but we think we’re on a new step beyond that. We’ve done all those things, and very recently we feel the business is now at a new stage: AI native. Everything we do now starts from an AI enabled position. It’s different to having agents and loops that do bits of people’s jobs, and different from just using AI in a sanctioned, policy driven, collaborative way. This webinar is about how we got to this point, and it came mostly from Rob and me talking through this one process, plus a load of others like it.

[00:06:55] Crispin Read It’d be good to get some feedback on where our viewers believe they sit on that six point scale. Let us know, and we can help get you towards six, where we’re currently sitting.

[00:07:07] Rob Brighton And having a lovely time with it, looking at my job with renewed vigour, having an interesting, exciting time.

[00:07:25] Crispin Read The other thing about this AI maturity model is I’m also thinking about it as a maturity curve, not just the linear scale. If you’ve seen my other stuff on how people learn, the Dunning-Kruger curve should be familiar: really exciting at the start, loads of fast progress, then you think you’re really good at everything, then you realise how much there is to the whole area, and that drops people into what we call the valley of despair, where it feels overwhelming. The good news is that’s the point where you realise these are exactly the steps you need to take to get better. We don’t want people staying in that valley, we want to help them onto the steady incline of getting more productive.

[00:08:40] Crispin Read I’m doing another webinar just on this, because there’s a lot to unpack. We’re working with people at all stages of this curve, and I’ve got an AI enabled process to map your team, department or business on this matrix. If you want to be part of that, drop a comment and we’ll be in touch. It’s completely free, we’re just collecting data at the moment, and you get a skills gap analysis of your team plus suggested projects out of it.

[00:09:40] Crispin Read So, what we’re actually here to talk about is the process that’s evolved across this journey. Before I hand to Rob, a bit of narrative: when this was a hobby project, we were a bunch of developers writing curriculum and session content for each other, training each other’s junior developers. We used GitHub for version control, so when someone ran a session they’d check it out, review it, and check it back in. What we had in GitHub was a constantly reviewed, always current version of the training.

[00:11:04] Crispin Read When The Coders Guild became a proper business, with non-technical people who weren’t comfortable in Git, we needed a new way of working. We created a format where everything lived in Google Docs, more clunky to work with, but we still wanted that currency and constant review. Rob, can I hand over to you?

[00:11:51] Rob Brighton That’s fine. It hadn’t really changed much. We brought things into Monday, used forms, so the trainer would do a pre-session check, deliver, then provide feedback. As the team grew, someone had to collect that information, take the suggestions, and put them into a workflow. That would fall into a sprint cadence, which could take a couple of weeks to join, and then the deadline for the change could be another two weeks after that.

[00:12:47] Rob Brighton It worked, it didn’t need changing, until it did. The problem was we’d have multiple feedback forms coming in, requiring manual collation and triage before adding into the workflow. It was slow, time consuming, and a lot of manual labour.

[00:13:11] Rob Brighton Why did we need something different? As Crispin said, we helped design the standard for AI and Automation Level 4, so we had new course material we’d want to keep as up to date as possible, iterating on every session. AI is a very fast moving sector, and while the fundamentals of the standard hold true, a lot of the tools and terminology are shifting, and those small changes need constant updating.

[00:14:01] Rob Brighton The old cadence was fine, but because of high demand for the AI apprenticeship and the short courses, we could see these courses potentially running monthly, so we had to speed up the pace at which changes got into people’s workflows. We identified that a new process had to be created so we could manage our workload with a smaller team.

[00:14:51] Crispin Read One of the things that reminds me of: when we first did this with software dev back in 2012, 2013, people always said “this is moving so fast, how do you keep up.” Software dev is probably the thing that’s changed the most, and more often than any of the others. We had to have a system to stay current, but Rob’s right, it was fine, every couple of months we’d change a thing. Whereas with the AI course we run, we start a new one every four to six weeks, and I don’t think a single session has stayed the same through even one iteration of the course.

[00:15:51] Rob Brighton When I think about the course development we did on the AI skills boot camps we delivered last year, compared to what we’re delivering today, it’s ancient. The fundamentals are the same, but the tools we’re talking about are different. If we look back at previous sessions where we talked about migrating into Claude, the concepts, projects, skills, routines, different platforms have different names for these things. These aren’t core changes we’re making, but to keep current there are tools we need to tweak and terminology we need to adjust, so people don’t feel out of date as soon as they’ve left the session.

[00:17:00] Rob Brighton The inputs have stayed the same: we take the transcript and the video recording that comes in through Google Meet, our preferred workspace tool, and that gets put into a Google Drive. The forms come in via Monday, and that drops into a board where we collect that information.

[00:17:50] Rob Brighton Here’s where Claude, the loop, comes into effect. We’re running it in Claude Code, where it’s called a routine, and Claude Code Web calls it a scheduled event, essentially a skill, a set of instructions for Claude to follow, on a repeating schedule, daily, weekly, monthly. Every working day, Claude goes off and checks the calendar for the events we’ve taught sessions in over the last 24 hours.

[00:18:30] Rob Brighton Once it knows which events happened, it goes to the Drive and checks for the transcripts saved from the previous day. Then it goes to the Monday boards, the feedback boards we’ve defined in the skill, and pulls the sessions attributed to the last 24 hours, bringing all of that together.

[00:19:03] Rob Brighton Once we’ve got that information, we generate some rich data, including whether a session landed well with learners, whether it overran or the timing was too short, particular topics that caused confusion, and insight into learner and coach engagement.

[00:19:42] Rob Brighton We always try to keep a human in the loop, so it generates a draft report that goes to the owner’s DM as a double check that the information is correct and we want to proceed. The owner then confirms in Claude that we want to write the session intelligence and summary report off to the individual session stored on Monday, and we send the summary report into a team Slack channel so that information’s available to everyone.

[00:20:28] Rob Brighton This is only possible because of the skills we’ve set up and the connectors to these tools we’d already set up, made sure are compliant, and documented, so there’s transparency. But the routine gives us this rich data output in multiple formats.

[00:20:50] Rob Brighton The daily Slack report comes in ahead of any standup sessions, so if there’s anything of particular note we can talk about it in standup, or in the various meetings we have through the week with the delivery team and senior leadership. It surfaces that information immediately so we can act on it.

[00:21:15] Rob Brighton The Session Directory is our library, where we keep all our sessions linked and detailed, and the report drops onto a particular session. Each time it runs, a new report is built, so we’re building up a really rich picture of information, whether reviewing a particular session or doing a retro on a whole course.

[00:22:06] Rob Brighton There’s another human step we can use Claude for. It can add a subitem to the session delivery board if it’s urgent, or drop a new task into our delivery board with the exact content and links, so it’s not a human copying and pasting all of that. It’s minutes saved, but that stacks up over a week or a month.

[00:22:51] Rob Brighton One of the things we’ve found is it’s amplified our reporting capabilities, which helps us report to the wider team on attendance, and we’re looking for more opportunities to use this data in future.

[00:23:28] Crispin Read The important thing for people watching us talk about this, because it is quite technical and invested in our own domain knowledge, is that if you step back and look from above, the difference is that it used to be humans setting the trigger, doing the review, setting the actions, doing the checks, always thought of as “if this happens, do this thing.” We’ve been building automations and bits of AI into that for a long time.

[00:24:20] Crispin Read The shift, the AI native bit, is that we can reimagine that entire process, evolving it to a new level. We’re not looking for triggers anymore, like an automated thing waiting for a new transcript or a new form submitted, hold this here, do something later. It opens up a real difference in our operating model: it’s no longer “where are the little bottleneck inefficiencies to correct,” it’s “where can this take us.”

[00:24:56] Crispin Read It’s as if we’ve got observers on every step of the journey, and there’s a lot more under the hood than we can explain without boring everyone to death. There are great implications, like for safeguarding flags. If people seem disengaged, we can anonymise those things, we’re not watching our learners, but a learner ID could be flagged as more disengaged than usual because they weren’t as involved in the transcript. We haven’t done any of this yet, there’s a bit more governance that would go into it, but the possibilities in that shift are huge.

[00:26:23] Crispin Read A couple of weeks back we were talking about this shift: go back a decade, Google Workspace was the centre of our tools, then things like Airtable fed into what we did in Workspace, then the next stage was a task management tool like Asana, Trello or Monday.com becoming the centre once we started doing more automations. Now I feel like Claude is the centre, and Google Workspace and everything else is peripheral. I think that’s the shift that made us start thinking about AI native as the thing.

[00:27:37] Rob Brighton Reflecting on what you’ve just said, a lot of these tools independently we were never really a hundred percent happy with, and couldn’t quite get them to work the way we wanted. We’ve put Claude at the centre and it’s kind of moulded to the problems with the platforms we were using, and just glued it all together, made the platforms much more usable.

[00:28:18] Crispin Read And it’s freed us up. I’ve always had this thing I bore everyone with, about how I’m against software telling people how to do their job or how to work. That’s always been the case with SaaS tools, you get to a point where you have to do it a certain way because that’s how the software works. This is really freeing, because we’re actually able to think about different ways of working, based on what we actually want to do, rather than what the software wants us to do.

[00:29:05] Crispin Read Right, sorry, I’ve waffled too much and left us no time for questions. There’s a QR code with a link, and Francesca’s also shared a Calendly link to book in that assessment. We do have one question, if you’re happy to field it.

[00:29:31] Crispin Read I’m laughing because it’s a question from my wife: aren’t you worried you’ve handed your competitors the blueprint? No. I’ve always worked on the assumption that what we do should be copied, and I’m often surprised people don’t copy the stuff we’re doing, like keeping sessions as up to date as possible. I’ve been vocal about that, spoken at conferences, written blog posts. I’m expecting people to copy it, because it’s a good idea and you could do more with it. I don’t want us to be the only people training businesses on how to use AI, but I do want to be the best at it.

[00:30:43] Rob Brighton That’s a noble goal. Right, we’ve overshot. Upcoming webinars: we’re having August off from webinars. It’d be great if anyone rewatching on LinkedIn or YouTube could drop a comment on what you’d like to see us cover. We’re planning on coming back with some Claude Design sessions and Claude Code, but if you’d like more introductory stuff, let us know, we might ask some questions in LinkedIn posts to tailor the webinars to what people want. The link to our webinars is up on the board.

[00:31:31] Crispin Read Have a great summer.

[00:31:33] Rob Brighton Thank you very much.

[00:31:33] Crispin Read See you on the other side.

[00:31:35] Rob Brighton Cheers.