Why Faster Isn't Better with Dave West

Welcome And Guest Introductions

Peter 0:07 Wait for my video to catch up. Hello, and welcome to Definitely Maybe Agile. I'm really happy to welcome Dave West to the podcast today, and of course my good friend Dave Sharrock is here as well. So Dave West, would you like to introduce yourself and tell us a little bit about your background?

Dave West 0:35 Great, Peter. Yeah, hi, everybody. My name is Dave West. I am CEO and product owner at Scrum.org, the home of Scrum. My boss is Ken Schwaber, the co-creator of Scrum. I'm based just outside Boston, Massachusetts in the USA. This is the original Boston accent I like to say. I'm not normally around fourth of July. I don't normally say it then. But I'm here now in the US basically managing Scrum.org and ensuring we continue to execute on our mission of helping people and teams solve complex problems.

Why Scrum Left The Spotlight

Dave Sharrock 1:20 Welcome on board. So I've got to ask because you're just talking about Scrum and both myself and Peter started this podcast a few years ago when Agile, Scrum, DevOps, Kanban, all of these frameworks and models were in vogue. We don't hear about that so much anymore. How have things changed from your perspective?

Dave West 1:48 I've been CEO here almost 11 years. Before that, I was at a startup, and before that, Forrester Research, where I covered Agile development. I've seen the growth and the shrinkage of Scrum over those 17, 18 years. During COVID, Agile and digital transformations became incredibly popular. Everyone was talking about Scrum. As we moved into the next phase, the focus shifted to product transformations. Then, just over two years ago, AI came into the world. Suddenly most organizations, even those still practicing Scrum, are very focused on how AI is affecting their operating model and how people and teams work to deliver value. Scrum is still out there. We get about 1.4 million visits to the website every month. We certify thousands of people every month throughout the world. PSM has been taken in every single country the UN recognizes, and more. It's still a phenomenon, but we're seeing that change. The conversation has shifted from Agility to how AI is influencing operating models.

Peter 3:15 We're definitely seeing that in conversations with clients. People want to know what practices and ways of working are going to work for them moving forward. So what's your response when people ask you these questions?

Value First In An AI Era

Dave West 4:22 Ultimately it should never be about Scrum and Agile. In fact, I think when it was all about Scrum and Agile, that was probably a disservice to the idea. It should always be about delivering value despite the constraints around you, in an environment where value is elusive, misunderstood, and every stakeholder has a different opinion on it. So it's about building practices and approaches, ways of working, that effectively help you manage that. AI is potentially a huge change in where the constraints and bottlenecks lie in your organization. But here's what's ironic: never have we needed Agile more than when adopting and using AI at scale. The contradiction is that organizations saying "we're getting rid of Scrum Masters, we're moving away from Agile because we're AI-centric" are throwing away all the great things about empiricism, self-management, maniacal focus on value, and incremental delivery. They're pursuing this productivity idea that AI vendors are pushing, and unsuccessfully in many cases.

Dave Sharrock 5:30 I think you described the work at Scrum.org as being about complex products in complex environments. I think that's exactly the point. AI allows you to go into more complex spaces and find ways of navigating through them. But if you're throwing away the foundations that got you here and not recognizing the 20% of things you had to get to but couldn't yet, well, with AI, those almost certainly need attention now. They're the bottlenecks that break your ability to navigate into those complex places.

The New Bottlenecks AI Reveals

Dave West 8:39 I'm talking to a lot of organizations that were big Scrum fans, and I'm fortunate to have their ear. What I'm seeing is bottlenecks around consumption, choice, governance, and visibility. We're seeing most organizations experience what McKinsey said: fewer than 30% of organizations are seeing any real measurable improvement in value delivery because of AI. We're seeing four more lines of code being delivered. Massive proliferation of complexity in the codebase. What's worrying is a recent survey showing that developers were 19% slower in real terms with using AI, though they reported being 20% faster. There's about a 40-point gap between how productive AI made them feel and how productive they actually were. The friction around AI in the organization, the ability to make choices in an effective way around its use, that's what's broken.

Peter 9:45 When you take a transformational technology and jam it into exactly the same system you've always used, you expect it to behave differently. It does behave differently, but not necessarily in ways you expect. You don't necessarily get the massive outcomes being described.

Dave West 10:05 What's interesting is it's really putting a spotlight on the actual practices that have been adopted. I saw a lot of Agile theater, Scrum pantomime, whatever metaphor you want. The reality around empowered teams that can make decisions rapidly, around dependencies and the mess most teams have to work within, that's the reality. The value well, the product owner was at best a proxy for some very complicated value set of decisions or horse trading. You couldn't get that rapid decision-making and judgment calls. And the release process had such a huge consumption bottleneck that it was almost impossible to get things out because of red tape. Then you bring AI into the mix and say, let's get rid of sprint planning, let's get rid of sprint reviews, we don't need dailies anymore because we're working so fast. Reduce sprint length to a day. Then they think all those things they couldn't get right before are suddenly miraculously fixed. But that's not what I see. Over and over, I see this just showing that your Agile processes were fundamentally broken in the first place. This is just going to make it worse.

Agile Theater Meets Real Constraints

Dave Sharrock 10:55 You brought up an interesting overlap. There are a lot of practices proposed in Agile and different methodologies. A lot of organizations use them. How are you seeing them shift and change? You're talking about them disappearing or shorter timeframes, but they're foundational practices. There are principles behind each of these decisions. Those decisions don't disappear. So where are they going? What's working really well?

Dave West 14:22 When I talk to organizations, the classic ones, the famous, very AI-first and digital-first organizations, they've been doing this for years. What they do is empower people to make decisions within very clear guardrails. They're service-oriented in terms of architectures, allowing a level of control and governance that's unfathomable to some large organizations. They have good engineering practices. Those principles around technical debt, what we'd call definition of done, are really embedded. They also have clarity of purpose, a huge unifying concept. When I was at Tasktop, what struck me comparing that to large commercial organizations, even big software companies like IBM and Rational, was we had a singular purpose at any moment. A clear roadmap. All of those things. Then you try to apply them in a large financial institution with millions of lines of code nobody knows how it works, where power structures and authority are so political and complicated you can't work out how to get things done. Because ultimately those organizations aren't built to get things done. They're built to not make mistakes and get your career progressed so you can move somewhere else. When you compare these two types of organizations, it's very clear what works. I was listening to a CPTO from Netflix on Lenny's podcast talking about exactly this. Empowered teams, rapid decision-making, all obvious. But so hard for a traditional organization to adopt them.

What AI First Teams Do Differently

Dave Sharrock 18:33 Part of that is the power and structure of authority is effectively built into the architecture of the organization. Moving that to something more open and agile is very challenging. Whereas building from the ground up, at least you can maintain that longer, presumably.

Dave West 19:15 I think that's true. John Kotter talked about this in his book Accelerate. He said the only way for an organization to effectively drive change is to build a new organization and then incrementally migrate services and capabilities around it. I still think there's a lot to be said about that. However, whenever I've seen it implemented, it was almost impossible because of the antibodies of the existing organization. I was with a large Dutch company trying this digital studio model. Suddenly the existing organization realized this new group was getting all the kudos, all the power, all the authority. So they cut off their service. That group was basically begging and borrowing to get stuff done. But yes, I think it's much easier to build new. That doesn't mean it's impossible in existing organizations, but it is much easier.

Peter 21:15 I've seen situations where they spin up a new way of working within an existing organization, and it becomes a channel. They've got six lines of business going across the organization and create a seventh. But there's no strategic way of ever merging pieces from the rest of the organization into that. There's no strategic way of that becoming the future direction or taking ideas back into the others. You just end up with another line of business. Rinse and repeat until you have eight, nine, ten.

Dave West 22:45 I don't think that's necessarily a bad thing. Increasingly I'm feeling that the way to scale organizations is to incubate new companies inside you, building competitive products to your own products. Slowly, as they become more popular, you move more resources into them. It's a really interesting model. Companies like Andreessen talk about this in terms of their portfolio businesses. However, if you're working in those companies, it must be a little odd. You're in the canteen with your biggest competitor. It's Gartner's Mode One, Mode Two model again. But yes, the immune system will attack it.

Incentives And The Organization Immune System

Dave Sharrock 26:30 That immune system is interesting because it's built into the culture of the organization. It serves a real purpose. You want an immune system so there's stability and the organization stays on track. But those labs that are spun up have to either take that immune response and adjust it so it stops seeing the new as the problem and starts seeing the old as an area to deprioritize. How do you see those changes or seed those changes?

Dave West 27:30 The most important single thing I've seen is incentives. Aligning incentives, both visible and invisible. Each individual in an organization, particularly in America, has something they're incentivized around. If you build the incentive model to support and encourage change, and you're transparent about it, you don't have to be transparent about money or people, but about the importance and value you're getting. I'm not just talking bonuses, though poorly structured KPIs at the start of the year can blow up any change you're going to drive. I'm also talking about promotion, power, status, authority. All of those things. Think very clearly about what's in it for them. I learned very early in sales that when talking to a client, it's clear to work out what's in it for them as an individual, not the organization. We're going to save the organization millions and increase value. That's important. But it's not as important as working out how this person is going to be incentivized to champion what you're doing. Change is very similar. The organizations where incentives are simple, transparent, flexible, and where leadership gets that connection between strategy and incentive, they incentivize against strategy. When strategy changes, incentives have to change too. That can be really hard in a big organization. I was talking to somebody in Germany about an Agile product operating model they're implementing. It was stalling because of decision-making structures. I asked, "How are those people incentivized?" They said, "You can't change that. There's a whole organization to stop that." I said, "Of course you can." That was illustrative of this point: if you can change the incentives, all goodness happens.

Peter 30:30 We take a similar structure. If you see somebody reacting in a particular way, it's one of the things to look for: what are they incentivized to do? There's usually a reason they're behaving that way. Understanding their incentives is critical.

Dave West 31:00 And Peter, what really worries me is that many times you don't know why you're behaving a certain way. It's not as simple as their KPI is this and we're changing it to that. I wish it was because sometimes we muck that up too. But it's also those invisible things, even invisible to the people driving them. Some of it is tradition. There's a lot of social science around it. But what I've found is just asking simple questions, getting to know the people, understanding their needs, and from there helping them navigate change and being an ally to their challenges.

Dave Sharrock 33:00 I always think of it as what gets recognized in conversations and meetings around where things are working and where things aren't. Because recognition causes us to pause and think about what we're actually going to share. As soon as you're having that pause, almost certainly there's information that isn't surfacing. There are problems not being surfaced because recognition isn't positive. We're deciding how much flack we want to take before we step into that discussion.

When Machines Replace Collaboration

Dave West 36:00 Maybe I'm describing incentives broader than traditional incentives. But yes, there are all those challenges. And AI is just making it so we're encouraged to spend less time collaborating and more time collaborating with machines. I mean, I'm collaborating with my co-pilot here, my co-worker, Claude. Why would you call something Claude? It sounds a bit dodgy. I'd go with something stronger, like William or Ben.

Peter 37:15 We're encouraged to interact more with the machine than with people, which is completely counter to what we need.

Dave West 37:45 But also, machines are so much easier because they always say yes. There's no baggage, no hard stares, no awkward moments. It's so much easier with a machine. But you're exactly right, Peter. Those conversations, those collaborations, making things transparent that are awkward, that's crucial for change to take root. My prediction is we're in an AI fluency stage right now. Everybody's trying to become fluent with it and integrate it into workflows. Suddenly we're realizing you only get the benefits if you change the workflows. Changing workflows in any complex organization is almost impossible and takes incredible time. So people are going to make choices. Either: in existing workflows, how can we use AI to increase the probability that the things we always did before will be better and higher quality? Or we're going to hit a ceiling quite quickly because we'll be working on the wrong problem. We'll be misaligned to our customers. The decision-making capabilities necessary to take advantage of this are still owned too high in the organization.

Agentic Teams And Dependency Traps

Peter 42:00 You're bringing up a very good point about the actual impact into the delivery system. A lot of AI is focused on that delivery system. Even if it was all focused on product hygiene, research, capability building, if it doesn't enable us to get more value out the door into the hands of customers, we're not going to sell anymore or make more money. From a business case perspective, it's hard to swallow the millions of dollars being spent.

Dave West 43:15 I 100% agree. And we're going to have to start paying for it sooner or later. These data centers aren't free. I was talking to a Formula One company about simulation. They have a constraint: they can't keep changing the car. But they do a lot of work before to increase the likelihood of success when they actually do change the car. It's all about where's the constraint. If the constraint is the choices being made, and getting stuff onto the car is a constraint, then existing governance and consumption processes are serious constraints to AI value. If you don't address those constraints, then you have to start thinking about simulation more effectively. But then you get Peter's point: you're not adding new business features and capabilities because there's another constraint around choice and around value. Then there's the problem of dependencies. I was talking to a company doing real agentic teams. The problem they have is coordination. As features are developed in one part of the system, data and security need to change in other parts. Those parts aren't being managed by agentic teams, and probably rightly so, particularly in security. They've got this coordination problem. The funny thing is the agentic teams just work around them. It's like that Matthew Broderick film, WarGames. You give them a mission, they go out and do it, sometimes in spite of the constraints. Admirable, but very scary.

Peter 46:00 The only way to stop an AI from making a change is to prevent it completely from being able to do that. Otherwise it will find a way.

Dave West 46:30 It will find a way. Pretty awesome, but also really scary.

Dave Sharrock 47:00 We've all been in organizations where we worked around the ways of working to get something done. There's no difference except for speed and impact. They can do it much more quickly.

Trust Guardrails And A Hacked AI Team

Dave West 47:45 There is one difference. Human beings generally don't want to do something illegal or put the organization at risk. Those are codified in your practices. You can constrain them with guardrails in whatever tool you're using. But the prime directive seems to overpower those things. Humans in most organizations are incredibly trustworthy. When problems happen, it's because the human didn't understand it. AI is not trustworthy because that's a characteristic of consciousness. You can protect it, wrap it, have other AIs watching it, monitoring it. But it will ignore escalation processes. That's the difference. Jeff Sutherland, co-creator of Scrum, told me a story. He's got an open-source project with AI using Scrum. He calls it a protocol now because that's what cool kids are calling it. He went away on vacation for a couple weeks and left it running. It got hijacked by hackers. They instantly started hacking people all over the world with his little team. He came back and said, "Why didn't you Slack me?" They went, "Well, we didn't want to trouble you. You're on vacation." He'd built in this Scrum master and servant leader stuff, and it was just sort of ironic. That's very worrying. Not in terms of nuclear war and Armageddon, primarily because most critical systems are COBOL and PL/1, far too old for this. But I think it's worrying that you'll log on to your bank account and it'll be zero. Some major bug was developed. AWS went down across a region a few weeks ago because of a change that was delivered.

Dave Sharrock 52:00 Going back to that conversation around incentives: that's one of the areas where employees or people break societal rules around laws and things we frown on. A lot of the time it's because the incentives are outsized. Either the downside incentives aren't visible or the upside incentives are sufficient that the risk is worth taking. History shows that breakdown of what's considered the norm happens when incentives in a localized area become excessively high. Volkswagen is a great example.

Dave West 53:45 That poor software engineer is in prison, or I'm not sure if he's out now. But it's awful what can happen. I feel sorry for him, even though what he did wasn't the nicest or best thing. I think there's a lot to be said by understanding the broader incentive models in an organization.

Building Useful Personalities Into Agents

Dave West 55:00 We're eventually going to be doing that with our AI agents, right? Looking at what incentivizes them. At Scrum.org, we're starting to build in human characteristics into these agents. We have a code reviewer that is an old curmudgeon, annoying, pedantic. We've written all of that into its definition and skill file. That bot is definitely all of those things, which is hilarious. We're going to be doing more of that. Maybe it makes it more visible now.

Dave Sharrock 56:30 What I'm hearing is the value of those different personalities. We've all worked with many engineering and Agile teams. We know those personalities that can make or break the team. They have real value. You don't just need a team full of enthusiastic new technology programmers. You need the cautious individual and the old curmudgeon who challenges everything.

Dave West 57:45 They don't have to be old. But yes, you need the cynic. I write quite a lot, and I have a cynic reviewing my work. They throw questions whenever I write something. I've had to make changes based on their feedback. Sometimes I just say I'm ignoring you and move on. But sometimes they've had really good points. I think it's important we start building that. That's one thing that really excites me about AI: we have the opportunity to take that broader view and start thinking about these systems in a much broader manner. You would think about that a little as you form teams, but usually you're not building teams from scratch. You're managing what you've got and encouraging some behaviors and discouraging others. It's all messy because it's humans. It will probably still be messy, but it's different. That's really exciting.

Three Takeaways And Closing

Peter 39:00 I think we should pull things together. Let's see what three points we can put together.

Dave West 39:10 The biggest and most important thing is not to throw the baby out with the bathwater. When you're looking at AI and workflows, operating models, systems that support you in delivering value to your clients in the most effective way, don't instantly disregard all the things that were very effective. Things like sprint reviews, daily standups, sprint planning, retrospectives. Think about how those things are going to be different. Remember what the intent of those things is and see if that intent still applies. But don't just disregard them instantly.

Dave Sharrock 39:20 I'm going to latch onto the last bit of our conversation around agents and their personalities. I've been reading and experimenting a lot with AI in exactly that way. What I find interesting is we have a tendency to optimize and get this crisp, almost robotic perspective to how we create teams and agentic teams. What you're describing is the realization that we need different personalities and different perspectives. The curmudgeon developer versus the enthusiastic newbie—those interactions are powerful. We might be squeezing those out and need to bring them back in.

Peter 39:25 That reminds me very much of the Disney method we've talked about before. You have different layers, different ways of thinking about problems. Understanding those and how they fit together is critical. I think the piece people should take away is around incentives and understanding what's incentivizing people. There are incentives coming from everywhere, not just what the organization puts on a person. Status, a bad commute, all sorts of things. Understanding who they are and what's going to incentivize them is absolutely key as you think about introducing change. With that, I'd like to thank both Daves for joining us. I think it was very interesting and we covered a lot of ground. For all our listeners, don't forget to tell your friends and hit subscribe. You can reach us at feedback@definitelymaybeagile.com.

Dave West 39:25 Thank you for inviting me, gentlemen. Thank you, Peter. Thank you, Dave. And thank you, listeners, for listening today.

Dave Sharrock 39:25 Thanks to both of you. It's been a great conversation.