00:23:45 · Luca Ferrari
[00:23:45] S3: more worried about our ability to find enough great people, but not so much about the scalability of the model per se. I think what we find consistently is that the best antidote to handle, let's say, complexity and scale is the combination of talent density and extreme ownership that beats rules, processes, centralization of decision making any day. I've heard people say that Bending Spoons is highly centralized and that's completely true or completely false, depending on how you define it. Centralization is a broad concept and the devil is in the details. If we talk about decision making, for example, we're extremely decentralized. At any point in time, each of our businesses is run, but also our platform teams are run by a dedicated team of Spooners and acquired team members, if it's a business. At any point in time, they have a complete mandate to do whatever is needed to help not that business, but Bending Spoons as a whole succeed, which generally is started by running that business well. I and others monitor, oversee progress by being immersed in all the different Slack channels and threads, and looking at dashboards and analysis. But no decision has to go through us to be, let's say, call it the executive team, whatever that means, to be approved other than a super small set of decisions, for example, related to pricing. But we're talking maybe decisions that are made once a year, 99% decisions. The teams are trusted to take them. We sometimes joke that we buy established companies and we bring them back to startup mode. And we really do. These small talent dense teams with massive mandates to just go and get it. So decision making is extremely decentralized. And I don't think we'd be able to scale otherwise, because centralized decision making has multiple flaws. Obviously it can be slower because you don't have time and people have to wait. You may not have enough context to decide well. And especially if you have an inflated sense of ego, then you're probably still going to think that you can make up for that lack of context out of your extraordinary intellectual abilities. But then guess what? Often you don't. But the most important flaw of centralized decision making is that it deprives people of that precious sense of ownership.
00:25:51 · Luca Ferrari
It's difficult to feel excited about the work and responsible for the work if ultimately all the important decisions someone else has to place a stamp of approval or even make them then destroys productivity on a daily basis. So we don't do that. Another way people define centralization is rules. So very specific rules for how things are done. And I think that's partly a way that management teams try to implement their decisions at scale. They believe they know best, they decide, and then rules are the way that those decisions are imposed throughout the... Rules become a bureaucracy, obviously could be more or less complex and more or less red tape, but sometimes rules are just an enabler for centralization, meaning if you strongly standardize how people operate and report, then you make it more feasible for this supposed central management team to make their decisions. If they had heterogeneous information, it would be more difficult. We don't do that either. We actually tell our teams, don't report in a way that makes me happy. Use Slack channels to talk to the team. I'll use the same info for the most part. If the info is bad, it's because it's bad for everybody. I'll tell you, but not because it's bad because I don't like it structured that way for the most part. What we massively, let's say, centralize, I don't know if centralization is the right word, is we standardize the toolkit. The underlying technological foundation is the same for everybody, but that's something we bring in on almost on day one after a new acquisition takes a few months and then it's done. And by the way, businesses still have control. We are like an open source community internally. So if you find that there is a bug or a missing feature, you build it and that benefits all the other businesses. So it's not that you're limited necessarily. And then we give you a set of principles, good ideas and lessons learned that hopefully you follow, but they're not prescriptive. They're not rules. We have a massive homogeneity. We make sure that the toolkit and the principles are homogenous across the group. But then operationally, in terms of decision making on a daily basis, teams are completely autonomous. And for us, that's worked pretty well so far.
00:27:55 · Paul Buser
[00:27:55] S1: How has the chatty pity moment influenced that? I know I have heard you say that you were doing AI back thirteen years ago when you launched this, even your previous company. Now these tools have become much more useful at scale. How does that figure into this equation of centralized versus decentralized and ownership by people?
00:28:14 · Luca Ferrari
[00:28:14] S3: AI is having a massive impact on productivity for us. It's enabling people to do more, faster, better work. It helps in the context of scalability. I think it helps in a bunch of ways. Again, if each person can do more work, you need fewer people to say, absorb the same number of companies or larger companies, which in turn makes a lot of challenges easier. For example, you don't need to hire as many amazing people or people capable of extreme ownership, which is a win check. Alternatively, you could do more acquisitions in parallel or bigger acquisitions, which is good for growth. It also helps keep the organization functional at scale because it makes producing documentation much faster. Now writing a memo describing, say an experiment becomes 10 times faster than before potentially. We're trying to automate it fully, by the way, it's mostly automated already at Banish Spoons. Similarly, if I need to follow progress across different businesses, I need other people less. I can be more autonomous. In the past, I had to ask to be granted access to the relevant database. I would need to query data or ask a data scientist to query for me, have them run an analysis, do it myself. Now I go on Slack, we have built a set of technologies and automations, the superficial layer of which we call alternative spooner. And I've got my own Alternative Spooner, I will ask it to run that analysis and then it will go and automatically request access to the relevant database. Access will be automatically granted for say five minutes, the amount of time that the agent needs to review the data. It will compile the analysis, it will share it with me whether it's a plot or in textual form. And so I can be autonomous and even though I don't have time to follow everything in detail, I can very quickly catch up indirectly through the agent. So it is a boon. It makes us even more bullish that a call it centralized model, I think it could be a slightly misleading word can scale massively much more so than what would have been possible in a more traditional, technologically speaking, side of boundary conditions.
00:30:16 · Paul Buser
[00:30:16] S1: If you take it to the next logical conclusion, all of this internal work for thirteen years, my mind as an outsider makes you a better owner of a lot of these companies than just about anybody else. Can you take me through why M and A is part of the strategy linked to that and then what the market gets wrong? I don't know. I think some people label used to be kind of an acquirer of apps, then it's private equity and software. I think there's probably other companies you mentioned Berkshire or Danaher Constellation, other companies that really are permanent holders of companies. But if you're aiming to be the best operating company, you get people who think like owners. You're able to scale them and have a large team and now 50 proof points of companies. How do you think about deserving to be the best owner of a whole swath of companies now?
00:31:05 · Luca Ferrari
[00:31:05] S3: I don't think we deserve anything. We just try to work our ass off to be that. It's not about deserving. It's just either you are or you're not. But you asked why M and A first of all, because we could also launch competing products from scratch. There are a couple of issues with that. First of all, we like to own businesses with some moats, with some staying power precisely because these businesses already exist in certain segments. If you launch something from scratch, they are very unlikely to greatly erode their market share. If it were so easy to do so, we probably wouldn't be that interested in participating in that market anyway. So that kind of rules out the follower strategy in existing segments. Other launch from scratch potential pathway, which would be radical innovation coming up with something completely new where the segment almost doesn't exist or you'd be disrupting it completely, which is how startups do gain market share when they succeed. The problem there is, if you look at the statistics of this, are absolutely dismal. Out of 1,000 startups, how many get to a level of revenue, just to pick the most obvious KPI, that would move the needle for us at this point? Assuming we close the acquisition of Miro, we're at about $4,000,000,000 run rate revenue. So we need to believe that this new thing we launch is going to get to at least 500,000,000. And by the way, if it gets there, so there's probably a one in a thousand probability that it does. If the broader markets statistics apply to us too, even if we were a little bit better, it would still be very small. And then generally these companies get to say, billion dollars in revenue in ten years, maybe today with AI and a better distribution and more efficient markets a bit faster. If it's in ten years, we're not interested. We would need earlier. Otherwise in ten years, probably want to have it at 5,000,000,000. It's just too statistically unappealing. I think there is a lot of hidden loss making investment in the world of startups in terms of the massive amount of time invested by very talented people, founders and early employees in projects that fail. And they do that for basically no money for a dream. Probably the expected value of that is awful. Naturally, there's selection bias.