Shownotes

Internal Conferences, Team Topologies, and the AI Learning Organization

Anne Currie speaks with Matthew Skelton, co-author of Team Topologies from IT Revolution and CEO of Conflux, about why AI is exposing weak organizational design and making intentional knowledge-sharing more important than ever.

They chat about how internal conferences, shared language, and clear team boundaries help organizations align faster, reduce duplication, and spread useful practices across the business. The conversation also introduces Matthew’s upcoming book Adapt Together and a new leadership program, Engineering the AI Native Organization.

Key topics

In this episode, Matthew explains that AI is acting like a bright light on organizational effectiveness, making unclear operating principles and vague responsibilities much harder to ignore.

He connects Team Topologies to the broader challenge of organizational architecture, arguing that value flow depends on clear team boundaries, aligned workflows, and shared understanding.

Matthew and Anne discuss how internal conferences at places like Financial Times, Metaswitch, and Klarna helped teams share knowledge, avoid duplication, and accelerate cloud native adoption.

They emphasize that the real value of conferences is not just the talks, but the hallway track, where context, nuance, and shared meaning are built through conversation.

Matthew introduces an innovation and practices enabling team, or IPET, as a way to spot emerging good practices, curate them, and help them spread across the organization.

The episode highlights the importance of intentional knowledge diffusion, especially now that AI systems depend on human-generated context and well-aligned terminology.

They discuss why leaders must define what good looks like, rather than assuming work is obvious or that teams will naturally converge on the right approach.

Matthew argues that organizations need a continuous refresh of context because what worked three months ago may already be outdated in fast-changing environments.

The conversation closes with Matthew describing Engineering the AI Native Organization, a leadership program for helping non-engineers understand concepts like APIs, decoupling, and asynchronous working in organizational terms.

Timestamps

00:00 - Casual warm-up and setting up the conversation

00:58 - Why AI is sharpening long-standing organizational problems

03:32 - Introducing Matthew Skelton and his work on Team Topologies

05:27 - How Internal Tech Conferences grew out of real organizational practice

07:20 - Why FT and other early cloud native adopters invested in internal conferences

09:08 - Cloud native success, innovation diffusion, and intentional alignment

10:36 - Why AI now affects the whole knowledge-work organization

13:03 - Internal conferences as a deliberate business tool, not just a nice event

15:02 - Public speaking, career growth, and organizational benefit

16:25 - Adapting the message for different audiences and stakeholders

18:18 - The missing assumption in Team Topologies: cross-team sharing is rare

19:44 - Team Topologies as organizational architecture for value flo

21:01 - Introducing Adapt Together and the need for intentional sharing

23:02 - Why AI systems need human-generated context, not recycled output

24:26 - Domain language and terminology alignment as an AI requirement

25:35 - Guardrails, metrics, and value flow without constant manual checks

26:40 - Why AI struggles when context is unclear or historical code differs

28:16 - Knowledge lives in groups, not just in individual heads or AI systems

29:39 - The hallway track as the real source of conference value

31:23 - Why listeners, not editors, do the quality control in live talks

32:38 - AI-driven change is spreading beyond software into every profession

34:23 - Why organizations need to define specific problems for AI to solve

36:12 - Nudging people toward the future state instead of enforcing it

39:41 - IPET: the innovation and practices enabling team

41:40 - Helping teams package and share their practices effectively

44:04 - Cross-functional learning, like HR adopting Kanban from IT

45:43 - Using internal events to prepare people for external speaking

46:39 - Organizational self-awareness as a competitive advantage

47:08 - Matthew’s brain science background and the “nervous system” analogy

48:18 - How Team Topologies and Adapt Together fit together

49:26 - Why leaders must define what good looks like

52:11 - Why “good” must be revisited continuously as conditions change

54:30 - Effective leadership starts with accepting that work is not obvious

56:29 - Matthew reveals Engineering the AI Native Organization

57:39 - Explaining engineering concepts to non-engineering leaders

58:36 - Closing thoughts on AI as an accelerator of organizational problems

Transcript

Anne (00:00)

So welcome to episode twenty-two of Asynchronous and Unreliable, a weekly podcast where we discuss the latest ideas and concepts in tech. I'm your host Anne Currie, co-author of Building Green Software, The Cloud Native Attitude, and author of the science fiction Panopticon series. Today I have the great pleasure to welcome Matthew Skelton, co-author of the award-winning book Team Topologies and CEO of Conflux, who wants to provide leaders with the levers and language to enhance decision making and create cultures that enable high performance without burnout. Welcome, Matthew.  

Matthew Skelton (00:33)

Thanks very much.  

Anne (00:33)

It's really nice to see you. I'm a fan of multiple of your books. We met at a conference many years ago, which is where I met half the people who are on this podcast. But I really love Team Topologies. Just yesterday I read Internal Tech Conferences for the first time, and I really enjoyed that as well.  

Matthew Skelton (00:59)

Well, that book came out just before Team Topologies, actually. For anyone watching the video feed, it's a purple book just above my shoulder. For anyone not watching the video feed, it's a purple book. I'm the co-author with Victoria Morgan Smith, who used to be at the Financial Times in London. We both had experiences running internal conferences.  We can talk about that a little bit later, but the way Victoria and I both framed it completely separately—and arrived at the same point, which is why we wrote the book—is that there's a way of bringing people together for events like internal conferences that is super valuable to the executive of the organization.  We maybe didn't emphasize that as much in that particular version of the book. There's a possibility for another version of the book coming out at some point soon where we're adding more stuff. But it's an example of the kind of activity that I've been doing in this space for at least fifteen years, thinking about organizational effectiveness, clarity of responsibility, bringing people together, avoiding duplication, and increasing alignment.  It fits in with a lot of the themes I've been focusing on for a long time. Even though the most well-known thing I've done recently is Team Topologies, even that is really about alignment, ways of working, effectiveness, clarity, and responsibility boundaries. It comes from the same kind of place.  

Anne (02:49)

Internal Tech Conferences studies three case studies of really sophisticated internal tech conferences. One was the FT, which oddly enough, I also studied because they're one of the top companies in adopting cloud native very early within their enterprise. I don't think that's a coincidence. Oddly enough, I interviewed Victoria, your co-author, for my book The Cloud Native Attitude, and I also interviewed Sarah Wells, who you work with quite a lot these days.  

Matthew Skelton (03:28)

Sarah Wells, yeah. Quite right. Sarah was also heavily involved in making the internal conferences at Financial Times work really well. It's great that you're able to acknowledge her contribution to the Cloud Native success at FT, because there was a core group of people. Victoria's one, Sarah's another, and there's a bunch of other people too. The comment you made there is very accurate.  The cloud native period is never going to finish, right? There's always going to be something continuing, but the real learning curve for cloud native was 2008 to 2022/2023 for the early adopters. The organizations that really managed to adopt cloud native very effectively and rapidly were those that had a very clear approach to innovation, discovery, knowledge diffusion, and internal alignment. 

It's not surprising that many of those organizations ran or still run internal conferences.  Klarna, for example, the payment organization, is one of the examples in the Internal Tech Conferences book. They are super ahead of the curve in terms of adoption of new approaches and ways of working, and their internal conferences have been running for ages. That raises a really important point. With cloud native, as you know, there was a real need for organizations to deduplicate, align, adopt things quickly, and have an intentional approach to innovation and practices diffusion. You can sort of argue that cloud native is only applicable to the IT department—from the perspective of some execs, they don't really care about cloud because it's just an IT thing. Now with AI, it applies to the whole organization.  

Suddenly now the entire knowledge work organization needs approaches to deduplication, rapid discovery of innovation, and sharing of practices—all the same stuff these cloud native organizations have been doing for 15 or 20 years. Now the entire organization is feeling the need to do this. I was talking recently to a CIO of a bank in the Middle East, and he said, "Matthew, AI is forcing me to make my organization a learning organization."  The whole Learning Organization book came out 20 years ago, but AI is forcing every organization to become a learning organization finally. Without it, you are going to get left behind or burn through so many LLM tokens that you won't achieve anything. It's forcing this shift towards effective learning, discovery, and standardization of practices in a way that is very engaging and human-centric. That's the picture I'm painting, and that's where I'd like things to get to.  

Anne (07:10)

Internal conferences are, as you say, excellent for alignment and engaging people in getting involved, thinking, and learning new things. You might not have realized, but the other company you did a case study on—Metaswitch Networks, before they got bought by Microsoft—I worked there for 10 years.  

Matthew Skelton (07:37)

Did you experience some of the conferences that were talked about in the book?  

Anne (07:42)

Well yes, because oddly enough I am obliquely mentioned in your book as a friend of the company that came in and spoke at the conferences.  

Matthew Skelton (07:54)

I didn't realize that! I'll have a look later—I have the book in my hand. I hadn't clocked that at all that you were involved, which is really cool. You can actually speak firsthand about the practices and approaches that make an internal conference actually valuable at the whole organizational level. It's not just about running an internal conference, it's what's the intention behind it. 

When I was leading these conferences at a company based in London back in 2012-ish, we had a deliberate approach that was technology-centric but outward-looking to the rest of the organization. For example, one talk we curated was the head of the database team and the head of the sales team talking about an Oracle database upgrade. On paper, that should be the most boring talk in the world. But because they came together, made it really funny, and had a live query showing old versus new—a technologist and a salesperson standing on stage together—it had a really powerful effect. 

We did similar things across lots of different dimensions.  To the point where the CEO, after we'd run five of these every six months, stood up and said that these internal conferences had been the single biggest thing to align business and technology in the organization, which is amazing. For an organization doing 150 million at the time, we probably saved them five or ten million by running those conferences—genuinely CEO-level business value. But it has to be very intentional and connected through to what we're trying to achieve, rather than just getting people together to nerd out about LLMs or cloud. It's fine to nerd out, but what's the actual business outcome we're driving for?  

Anne (10:16)

It's not just good for the business, it's very good for the individual. Nearly a third of the folk on this podcast were part of the Metaswitch conference. Jon Berger tuned all the material that's in his new book on stage at that internal tech conference. Yanqing Cheng was very good on stage at those conferences. Liz Rice was at Metaswitch with me before we did all of that stuff, but the culture was there to start with. It's fantastically good for individuals to get used to public speaking and shaping their thoughts.  

Matthew Skelton (11:12)

Even beyond that, if you can get individuals and groups who can give a really good talk at an internal conference, they'll be good for giving a talk at a meetup group and starting to spread out practices that are useful for hiring and for sense-checking whether what you're doing is relevant in the tech cycle. There's no point spending time getting excited about something that has already been solved and that you can just buy as a service.  

If people are capable of presenting in an engaging way—particularly being able to change their message for different stakeholders by diving deeper for specialists or keeping it high-level for non-specialists—that skill is hugely valuable for internal comms. Whether you're doing a status meeting or trying to persuade someone to join an initiative, the ability to articulate benefits with the right language is super valuable for the whole organization. There's an outsized benefit from nurturing these things, resulting in companies like Metaswitch or Financial Times being leaders in adopting new approaches.  

Anne (13:08)

I'm calling you back to a book you wrote a long time ago, but it's relevant today because AI is forcing fast change on organizations, and we need to find new ways to get everybody involved and aligned.  

Matthew Skelton (13:37)

Let me make a confession. When I co-authored Team Topologies with Manuel Pais—writing it in 2018, publishing in 2019, and releasing a second edition in 2025—one of the assumptions I made was that organizations would have a vibrant dynamic of people sharing things across team boundaries. Turns out most organizations don't do any of that whatsoever. 

That's partly because I've been lucky in the organizations I've worked in, so I assumed it was the same everywhere.  Most organizations would benefit from a substantially increased amount of cross-team sharing, learning, and innovation diffusion. That's particularly true where you've got an organizational architecture optimized for value flow. That's effectively what Team Topologies is: organizational architecture for value flow—the teams, systems, and workflows that support value flow. 

That organizational architecture was needed pre-AI, and it is absolutely needed with AI because that's how AI agents navigate different domains and contexts to orchestrate tasks. You set up your organization for value flow by thinking about an architecture that works, and Team Topologies is a great starting point. The danger is that those boundaries become too impermeable or watertight. So we need a way to share across the boundaries.  

I'm in the middle of writing a new book about exactly this. Once we set up the organizational architecture for value flow, how do we make sure information and knowledge don't get stuck inside these flow-oriented boundaries? Let's make it intentional. 

The organizations that succeeded in the cloud native era were already very intentional in actively diffusing knowledge across boundaries and running internal conferences. I think it's a predictor for success.  Now with AI, every organization needs to do this. You've got your organizational architecture in place using Team Topologies; what are you going to do to make sure you're intentionally sharing, deduplicating, and aligning?  

The realization that most organizations don't do that intentionally made me think I should write it down, since I've been doing this for fifteen to twenty years. The book is called Adapt Together, because it's about how organizations adapt together. The "together" part is about an intentional way of bringing people together rather than leaving it to individuals. It enables us to adapt to new technologies, new compliance, or new market conditions. It's a collection of patterns, case studies, and examples. 

Being very intentional about deduplication, looking for innovation, sharing practices, and creating shared standards ends up being essential for an AI-based approach to value delivery and product creation. AI agents need human-generated context. You can't put AI output back into the top of the funnel, because that leads to model collapse. It has to be human-AI intentional input. A side effect of doing this for organizational effectiveness is providing additional high-quality context into the AI systems you're using.  

Anne (19:14)

You're using it to constantly realign the AI with what you actually need.  

Matthew Skelton (19:23)

Yeah, and surfacing any disagreements, misunderstandings, or terminology that is not aligned. If your terminology is misaligned, AI is not going to play well with that at all. For those who are more technically minded, domain-driven design is essential now. Going back 20-odd years to when the book was published in 2004, it has always been relevant, but it's absolutely essential now.  

Underneath, we've got organizational architecture. On top, we're layering a deliberate approach to creating context by learning and sharing across the organization, setting up conditions so leaders can shape what's happening without micromanaging every detail. What kind of metrics and guardrails can we put in place that enable rapid value flow without a whole load of manual checks all the time? It's been proven in IT delivery since 2010 when the Continuous Delivery book came out. We're taking those same principles and applying them to the entire knowledge work organization. A key part of that is an active approach to diffusing awareness and knowledge.  

Anne (21:09)

It reminds me of a conversation I had with Sarah Wells: where context isn't clear, AI really can't handle that at all. The examples were code languages. Java is difficult because there's so much Java written 20 or 30 years ago under different contexts and needs, compared to modern Java. The same with C and C++. AIs can't handle that very well.  

Matthew Skelton (21:53)

Unless you can somehow codify that back then we used Java in one way, but now we wouldn't dream of doing that. That's a really important point about AI adoption. Knowledge exists in the heads of humans and in the shared experience of groups of humans.  Knowledge and intelligence don't exist inside LLMs or non-transformer systems; knowledge exists in the collective understanding and experience of people. 

That's why approaches like conferences work so well. You know the InfoQ website and the QCon conferences—bringing together that shared awareness is how we create meaning and generate context. Context is stored in the shared experience of people who had the conversation. Making space for people to come together and learn from each other, whether in person or online, is part of the secret sauce of organizational success because that's where actual knowledge lives. It doesn't live in a wiki.  

Anne (24:33)

The value in a conference is often more in the hallway track than the speakers.  

Matthew Skelton (24:48)

You need the speakers and sessions because that draws people in, gives you recordings, and forces speakers to condense their thoughts. But a big chunk of the value is in the conversations sparked in and around that. You need to look at it holistically—the value emerges from the dynamics of the whole event.  

Anne (25:31)

I had this conversation with Charles Humble. He was saying at a conference, the talks aren't edited the way articles on InfoQ are. So the task of fact-checking and evaluating ideas falls to the listeners talking amongst themselves in the breaks.  

Matthew Skelton (26:46)

The software industry had so many conferences like QCon because of the pace of change, and that's not going to decrease anytime soon. What's happening now is the pace of change for every other part of the organization is catching up and becoming very rapid.  Organizations need to realize that if they don't take an intentional approach to curating how learning happens and shaping the direction they want to travel, they are going to have whole departments pulling in different directions.

Think of a big container ship with a load of tugboats: if each tugboat pulls in a different direction, the ship stays still or moves the wrong way.  There needs to be intentionality. What specific problems are we trying to solve using AI? If you can't demonstrate alignment to solving one of these problems, don't bother using it. 

We don't need more reports or code; we need things that solve specific problems. Measuring people by how many AI tokens they burn per day is like measuring workers by how much coal they shovel into the furnace in the Industrial Revolution. It's easy to measure, but it doesn't shift anything.  It comes back to intentionality around purpose, ways of working, technology usage, and team stewardship boundaries. How are we going to nudge people towards the future state? Put in place ecosystem mechanisms like conferences, lunchtime talks, or common standards. If someone finds a much more effective way of peeling a banana that saves a million pounds a year, what is your organization's knowledge diffusion approach to ensure everyone learns it within six months? If it's not intentional, the organization burns millions of dollars on inefficient practices.  

Anne (32:33)

When I read Team Topologies, one of the things I liked most was the idea of cognitive load—trying not to distract people with things that aren't germane to what they're doing. Enabling teams spread best practice and manage cognitive load to keep people aligned. Conferences put you on the edge of what may or may not be germane. How do you control cognitive load so people hear useful ideas without getting overloaded?  

Matthew Skelton (34:14)

Ideas are cheap, especially with tools like ChatGPT. In the new book, Adapt Together, we have a concept called an Innovation and Practices Enabling Team, or IPET. It takes the idea of an enabling team from Team Topologies but gives it a specific focus.  Inside an IPET, you have a mix of people looking for early emergence of good ways of working that point in the direction the organization wants to go. 

They are very aware of what good should look like in twelve or twenty-four months, and they actively look for early seeds of better ways of working across the organization.  They help teams showcase their innovations. An IPET can work with a team on writing engaging conference slides or coaching them on giving a talk. If someone doesn't want to stand up and present, the IPET can interview them and write an internal blog post. 

The IPET curates these insights, hosts weekly lunchtime talks, and shares code or approaches.  For example, if the HR department starts using Kanban for onboarding employees after collaborating with an IT team, HR can give a talk sharing their perspective. This cross-departmental sharing gives others the opportunity to learn.  

By running this cycle repeatedly, you spot what works, spot the problems, and gather material for internal events or external conference proposals. Getting accepted at a conference validates your internal practices. Alternatively, if you get feedback that an issue has already been solved by a low-cost subscription service, it prevents you from spending millions building it internally. 

It builds organizational self-awareness—acting like a nervous system for the organization to sense its internal and external environment.  I did a master's degree in brain science (neuroscience), so I think of an organization as an organism. 

Adapt Together will hopefully be published in the first half of 2027. At Conflux, we're bringing together Team Topologies as organizational architecture for value flow, and Adapt Together as cross-organizational intentional sharing to build continuous context. The combination is very powerful.  

Anne (43:46)

All the companies that adopted Cloud Native successfully defined at the C-suite level what "good" looked like, providing a vision of the future to align against.  

Matthew Skelton (44:35)

I agree. Organizations that find this difficult often have leaders who assume the work and ways of working are obvious and shouldn't be hard. The ones that succeed expect to define and codify ways of working, language, and standards.  The work is rarely obvious. Articulating what good looks like is necessary at multiple levels. What's the right way of working? We need the human intelligence in the organization to understand what good looks like, set that direction, and continuously challenge it. What was relevant three months ago might not be relevant now. It's a continuous process of rediscovery. You can't assume that people's knowledge at the hiring point is enough for the next five years.  

Anne (48:09)

Everything important has trade-offs, and setting directions for people to align to is a high-skilled job.  

Matthew Skelton (48:29)

An important starting point for effective leaders is realizing "the work is not obvious." Let's find out what it should look like for the next three or six months, while expecting to revisit and reshape it as conditions change. Continuous refresh is the price of an organization existing in the ecosystem.  

Anne (49:23)

Internal conferences force you to talk to people and find out that it's not obvious.

Thank you very much for being on the podcast. I wasn't expecting you to announce your new book today!  

Matthew Skelton (50:30)

There is something else I'm working on called Engineering the AI-Native Organization. It's a leadership program for taking people from non-engineering backgrounds through concepts like decoupling, APIs, version control, and asynchronous work. We are applying engineering principles to the entire organization. How do you explain decoupling to a Head of HR, or APIs to a Head of Finance? This program addresses that so leaders can have better conversations and make effective use of AI.  

Anne (51:59)

AI has amplified difficulties within organizations by accelerating things, exposing bottlenecks and choke points. 

Thank you very much for being on the podcast.  

Matthew Skelton (52:39)

Cool, thanks for the chat. Loved it.  

Anne (52:43)

Thank you very much to our listeners and watchers of Asynchronous and Unreliable. I will catch you again on a future episode.