Simon: So thanks Dan for joining us for the Kimber Conversations podcast. I just hoping if if you could introduce yourself and maybe explain what content catalyst your your company does and and who it's for. Daniel Lord: Well, Simon, thank you very much for inviting me to join the podcast. So Content Catalyst provides the content intelligence layer to analyst research firms. So that includes the client portal, includes MCP gateways, API access. We provide an AI chat-based discovery tool. And we also help analyst firms to create new research products from their data. And by analyst research firms, mean companies such as 451 Research, IQVIA, Freedonia, Science and Medicine Group, Trend Type, Sperix Global Insights, so companies of all sizes. Simon: some some some startups in there and some very, very big blue chip names that that I think everyone everyone would have heard of as well. and you you founded content catalyst back in two thousand and three. What what made you want to build the technology behind analyst research rather than do something else? Daniel Lord: Yeah, well, was working as an analyst at the time. And I was conscious that I had information in all of these different formats. So I was making a data model. I had my narrative. I was illustrating the points and so on. And all of that would get packaged into a PDF and shipped to the customer. And back in those days, the customer might pay $5,000 for a report. And as Conscious, they probably wanted to do a whole bunch of the same things I'd been doing, for which the PDF format was a bit of a roadblock. So the original idea was to imagine the workflow they had, you know, making presentations and their own models and so on, and to provide a tool that kind of automated that. So 20 years ago, that was a really innovative thing to be thinking about. Today, obviously, AI opens up all sorts of new opportunities for how we do that. But the core ethos remains the same. We're looking for how to enable customers to get much more value from the research in a way that works for the research firm. Simon: Great. and when we talk about finding more value from the research, is that just in findability or or or how it's presented and and and delivered or or do we mean something else? Daniel Lord: Yeah, mean, findability is definitely part of it. mean, that's often surprising to think that even today discoverability is an issue, but some customers still have difficulty finding what they've bought. So yes, we can definitely help with that. But I think also there's a big opportunity for research and data to be used much more extensively. So I think as it can get packaged and reused in other places, there's an opportunity that's a bit analogous to the digitization of maps. So Simon, you're probably too young to remember the days when an evening out would mean taking a copy of the A to Z of London. Simon: Just barely. I rem I I'm young enough for multimap, I remember that. Daniel Lord: Exactly, multi-mapping the idea of printing off a page was a genuinely Simon: Yeah. Daniel Lord: innovative thing to do. And now that maps have become ubiquitous, you've got them embedded in your iPhone, you'll share a place according to a map link that the CarSatNev has it and so on, it just means that map data is used in far more places and creates a lot more value. So I think that there's something similar going on as well with proprietary data and research. Simon: Yeah, very interesting analogy. I haven't heard it put like that before, but I may well reuse it. and twenty odd years in, w what what do you think's changed about what analyst firms need from a technology part? Daniel Lord: Well, think technology needs have become a lot more sophisticated. So we're looking at 20 years of all sorts of different projects layered upon each other, of which, you know, sort of nothing ever quite dies. So all of this lives on in tech debt and so on. So I think that's one of the really interesting things about this is we're now looking at a whole new set of initiatives and ideas and the pace of that is accelerating. And so I think the key question is how do you help a firm to manage that, to stay relevant, to evolve, have the technology they need whilst their business might be dependent on maintaining all sorts of kind of core parts of their business for which there's existing technology. So Yeah, I think that's the key challenge that firms have is how do they focus on their key activities when the technology piece is getting more more complicated. And that's what we provide to our customers, that we have a business model that's aligned with their longer term success. we're providing technology solutions that we think are fit for purpose for them growing their businesses. Simon: Great. And I guess by that you mean you provide the platform so so they can focus on, I guess, insight and client value. Is is that fair a fair understanding? Daniel Lord: Yeah, that's probably how I'd see it. The successful analyst research firm has certain key activities and those should be talking to their clients. So sales, client success, like as much of that commercial activity as possible. And it should be as much of the creating the source of value, whether that's like how the proprietary data is created or whatever is getting out there and talking to people to shaping what it is. Basically those are the things you want to be putting all of your resource into and everything else I would see as overhead. Simon: that makes sense. and you you set up Deep Rethink, another another venture. what how does that differ from content catalyst and and what's the thinking behind that? Daniel Lord: So Content Catalyst supports real customers with a whole variety of the different solutions they need. So that goes through from the client portals to some of the kind of the more recent AI capabilities that their clients require. So we're supporting the full range of requirements that a successful analyst firm has. With Deep Rethink, I wanted the opportunity to just focus on some of the kind of AI first opportunities. So it's a kind of an agentic offering where we have a look specifically at, you know, what I think is going to be the future of, of getting kind of proprietary data, what kind of research products you can create from that, and then how you can ensure that gets impact with clients in the AI era. So it's I'm really enjoying the opportunity to work on that. But it's complementary to Content Catalyst in that Content Catalyst provides a much broader platform and it also provides a lot of the supporting infrastructure upon which AI initiatives can be built. Simon: Great. So content catalyst is kind of the broad, robust infrastructure that's tried and tested and I guess deep rethink it sounds like that's almost your incub incubator for the really innovative thinking and and and what might create the products of the future for for analyst firms. Is that is that a fair summary? Daniel Lord: Yes, exactly right. Simon: Brilliant. and you've talked a little bit about MCP. It's a topic that some of our listeners will be incredibly familiar with and and and others less so. But for anyone listening who isn't that deep in it yet, what what what do we mean by MCP and and and how does it apply here? Daniel Lord: Yeah, great. think, well, I think MCP is tremendously exciting. And I think it, you know, part of it is kind of analogous to the shipping container in that it's a way of standardizing something in which, you know, creates so much more opportunity. I think taken from the perspective of the of the customer, like the the enterprise user of AI tools, MCP is a way of grounding their questions and their workflow tools, it's a way of grounding that in high quality data. From the perspective of the data provider, MCP is a way of enabling what their customers want to do, but allowing it to happen in a kind of governed way, which means that they can have a sustainable business model to deliver it. Simon: Great. and I think it it will sound very familiar to people who are listening who who maybe were familiar with things like APIs in the past and we we we remember some vendors would just create an API and actually the the the closer you get to it, it it that doesn't necessarily mean an awful lot. It's it's kind of what what can you actually do with an API that was the important bit. But but what's what's really different this time around with MCP? Daniel Lord: well, I do think that thinking of MCP is a, a kind of evolution of API for AI use cases is, think it's, it's broadly a good, you know, it's a good analogy. I think one of the obvious things that's different is that the, the entity consuming it is more intelligent and might use it in unexpected ways. So, to some extent it could do that with the APIs as well. but a lot of the MCP aspect is it's kind of self-describing of what you, of what it does. So it opens up a whole new dimension in terms of like, do you correctly describe everything? How should it be navigating? How should it be using it? How should, you know, as bots in the future choose between different research vendors and so on, how do you ensure that it can make good choices? So I think it's got all that to it. And also there's a MCP is not just being used as a way of providing access to data. you can also have kind of functionality and capabilities that can be triggered by it as well, as I guess you can with APIs. But yeah, I think it does come back to that shipping container. It's a standardization, but you can get really creative with it, and there's a lot you can imagine doing with it down the tracks. Simon: and and what's the most common misunderstanding you hear from clients or the industry about what MCP actually gives them? Daniel Lord: I think the most difficult thing really is the evolution of different AI kind of modes of access that have evolved since that kind of chat GPT moment. So I think if you remember when chat GPT suddenly became, know, it was everywhere. And I think those who'd spend their lives, right. Simon: Looked like magic, didn't it? Daniel Lord: Yes, but if you've invested your working life in developing proprietary content that you look on in absolute horror as something appears to hover it all up and then just answer from it. So I think there's a bit of big shell-shocked from that worry that, God, all my customers are going to start using these tools and I'm not going to be able to enforce it and the future editions are going to be trained on it and answering from that. Now what's happened since then is there's been a set of solutions to that problem. So the first thing is the AI companies undertook to own business plans, not train the model, which is obviously great if your customers 100 % use the business plan with no train. And also there's a level of trust required with the AI company. And then it became about well, some people want to make their own models. Are you okay about training that model? So I think I heard the private equity firms, for example, are quite likely to have like a private model that they wanted to train. And so that had a different set of risks and a different commercial model implied by it. And then it sort of moved on to the being the inference and the rag usage being sort of separately defined. And then Now we have MCP, I think gives like, I mean, all of them have their place and depending on the, like, if you've got a, an appropriate licensing mechanism and so on that everyone's happy with, but it obviously needs to be negotiated by parties that both understand the implications of everything. So I think MCP overall is, is my more clearly aligned with the existing business model of most analyst firms. So it's a step forward, but to go back to your question of like, what's the misunderstanding? I think it's cumulative from these different technologies and just knowing the difference between them. then I think obviously there's varieties of different levels of knowledge within an analyst firm and presumably amongst their customers. And it may be that two years ago they asked could they use it for AI for which they thought that means train a model. And today they assume they can't, whereas the analyst firm might be very happy if you asked again and they'd say, yes, we've got an MCP gateway. We're very happy to do that. So I think it's how quickly we've gone through these different forms of accessing the content with completely different exposure for the analyst firm's intellectual property. I think that creates a lot of confusion. Simon: Yeah. And I I I guess there's so many different touch points that a an end client might have with an analyst firm. It's it's it's picking up on those conversations and keeping the the the clients up to date with with where the policy sits in terms of of AI use with content and and everything else. you mentioned a couple Daniel Lord: Yeah. Simon: of words there and I I I think a lot of our audience will be familiar with them, but for those that that aren't, would you be able to just at a very high level explain rag and inference? Daniel Lord: Yes, the inference is the opposite of training. one is that you can train the model and then once you've trained a model, it's expensive to build a model and once you've trained it, you can never take things out. So that model will in perpetuity repeat answers from what have been put in the training data. Now, an actual model might not have a very long lifespan. It may or may not be a problem, but that's something you probably want to have handled in the licensing. Whereas inference and RAG are all about taking... So that's using the AI in a non-trained way. you might take search results, for example, and then you use the AI to write up the answers to it, but it's in a way... that doesn't go into, it's kind of through the whole mechanism of it, it cannot train the model. Simon: and w wh where does the the real complexity sit with MCP? Is it is it technical, is it commercial, is it governance or or all three? Daniel Lord: It's very easy to get started with making an MCP gateway. It's sort of trivially easy with, say, code. You could take your APIs and make an MCP gateway. The difficulty comes when you want it to be governed, when you want it to enforced usage rights, licensing, ensure that it fits fair use of the content. And then you also, you're doing it for a reason. want the customer to be getting as much value in a world where they're probably going to have MCP access to lots of different vendors. So you need to be really clear to the AI that's consuming it, exactly how to get the most value from it. And also the way in which it gets the value will show up in your... usage statistics and your understanding of what's valuable about your research product to kind of make an improved research product in the future. So it's one of those. I think it's a bit like playing paddle. it's very easy to get started. It's very hard to get good at it. And I think MCP is one of those that the deeper you go into it, you realize there's quite a lot to it. Simon: Absolutely. and and in in terms of good governance of MCP, you touched on that, but for an analyst firm, how how I mean, I guess they they they really have to think about some sort of policy for good governance of MCP and and AI in general. But what do you think that looks like? Daniel Lord: So the most analyst firms will want to have a subscription business model. And so for a subscription model to work, you need to have a concept of what's been licensed. need to plan for the renewal conversation. So at renewal, you need to be able to show that you have had value from it, idea that you can show it's... you've had impact from it, but that requires understanding the usage data and so on. So the governance really follows from what's required to ensure that customers get as much value from the content and that you can implement the kind of licensing that goes around that. But you can also provide the usage information that's required to support the renewal. and also to guide the analyst firm in terms of the research and data they create in the future to satisfy evolving customer needs. Simon: So it's it's complex. and and and what what's what happens if a firm exposes an MCP without proper governance? What what's the realistic worst case there? Daniel Lord: Yeah, so that's an interesting question. It's obviously one as a supplier and a partner to to analyst research firms, that's not a risk I'm at liberty to take. But I think it looks like... Simon: Well, I th I I think, you know, you y you're obviously providing a a kind of best in class type solution, but I think there are lots doing kind of looking at building their own platforms, which is also very common in this space. But I I I think when you have all these new concepts pop up I I I you know, I I think it's it's it's kinda quite interesting to think about, you know, without without without bashing anyone in particular, but you know, just I 'cause I I I think I think governance around MCP, certainly some of the conversations I've had, there's a there's a wide spectrum of understanding there. But yeah, it would be really interesting to Daniel Lord: So Simon: hear your view. Daniel Lord: I think the really basic bit would be just ensuring that it's got access to the same data and content as the customer's licensed. And it's ensuring that you have kind of competently tracked the usage of that. So I'd say that that's the very foundation. I would imagine that any any analyst firm would get that in place. The next level becomes more around, yes, so are you actually getting the, are you capturing the real usage? Are you getting that in a way in which is going to give you the insight you need for this being a viable thing in the future? And then I think the thing that is going to be really hard for most companies to think through is that avoiding commoditization in a world in which everyone else has got an MCP gateway as well. And we're thinking at the moment in terms of like, know, who's buying your research, you've negotiated with them usage, but this could quite quickly turn into what we'd like token based access if, you know, so different business model coming in possibly, maybe you Maybe you can push back against that. And the thing that happens next is that the buyer isn't even a person. It's a bot that's following an algorithm of how it wants to source the research and data for what might be quite a opaque set of reasoning for how it makes that purchasing decisions. I think there's obviously there's a There's some really basic fundamentals to get right now, but it's sort of opening up, is opening a kind of worms to some extent in that you need to have some, have put some thoughts into what's happening later. So we're, we're very interested in, for example, what, how would that future bot make the decision about which data advice? Cause we want to support our customer and being, you know, being able to make the best case at that point. Simon: Very interesting. and and who inside an analyst firm should actually own the governance? IT or research leadership or legal or I I guess it depends maybe, but what do you think? Daniel Lord: Yeah, it's a really interesting one. I think this is so core to what an analyst firm is that it's very unlikely to reside in one of those individually because the whole business model will have that threaded through it. So if you want to change to something which changes the way it's licensed or something like that, that's something which needs the CEO and the management team to be aligned on. So I think each of these roles will certainly have like a strong opinion about it, but I think what's the kind of challenge for analyst firms to move quickly to client requests is that they've got different stakeholders who are involved in thinking these things through. Simon: And are there any client trust or or IP risks specific to MCP that are different to the risks around a normal API or or data feed? Daniel Lord: I'm not sure really, think, the MCP, I mean, it depends what you get your MCP gateway to be doing. I mean, it could in theory as an API could be doing is, is it, it could be using the AI to draft things. I think most, most of the MCP use we see is very straightforward retrieval of data and content. so there's obviously there's, there's a bias and hallucination and what have you issued. that comes with what the client subsequently does with it. But it's transparently the client doing that, using their own AI tools. I think what's happening on the kind of customer side, so on the analyst firm side of the MCP gateway, to me looks a lot like being on the side of an API gateway as well. And I think that... I definitely don't want to minimize that there's a lot to think about there, including when these things are going to be consumed by agents. They might use these in repeated ways or in unexpected ways, but I don't necessarily see MCP as being significantly different to APIs on that level. Simon: and and we're seeing firms, particularly those building, making real sort of make or break investments in their systems right now. what what do you think that pressure does actually does to decision making inside a firm? Daniel Lord: yeah, I mean... I think a lot of it is just continuing down a track that they've sort of become committed to or locked into. I know quite a few firms who've kind of taken that path feel that they can't easily choose a different path now. would say that's not the case. We could definitely help you, but I think... that path becomes more more difficult in the sense that I don't really think we know what the future looks like. So. where somebody in a firm would have had a vision for how something could work, how I want to visualize my data or something like that. That might have been a viable thing to do five years, 10 years ago. I think now the chance of that vision being kind of the world not turning out that way is very high. And the lifespan of your current vision might not be five years, might be two years or 18 months or something because there's a whole set of other things at play here. So you've obviously got, you know, what you want to do to provide more value for your customers, but your customers are also thinking and being equally creative in terms of like how they want to organize themselves and what their vision for their use of AI is and where you fit into that. And then I'm sure the AI industry is going to continue to throw grenades into this in terms of a, you know, a fast pace of innovation and you. tools and things coming along for us to make use of. So I think that's one of the most difficult things is, you know, we're all tempted to think that we can see what the future looks like. But I think overconfidence in that and under appreciation of all of the other things you're going to have to build if you do that leads to people being more more locked into a set of technical debt that they've built over the years. Simon: and is is is is is is there a pattern that you see early on with with customers or potential customers that you you talk to that tells you a project is heading toward being a big expensive thing that that that nobody really needed or or or doesn't deliver ROI? Daniel Lord: I think the key thing really is to do things because your customers want it, or it solves a customer frustration. So I think if, if somebody's got a vision for something where they're not necessarily that connected with the customer, I think that's a real warning sign or, or if it's a sort of compound idea where there's it's strong enough because you've combined a few ideas into one. thing you're going to do, that often is a symptom of a bit of sloppy thinking in terms of what the core thing you're solving is. So I think really good projects come from a very good understanding of your customer and solving that frustration and keeping it really lean and precise. Simon: Interesting. and I I I I it makes me wonder, so within a an analyst firm, who's whose job should it be or or or does this person tend to exist to to keep pulling a project back to does this actually help our clients in terms of technology investments? I g I guess it depends on the scale of firm, but I'd be interested in your take on that. Daniel Lord: I think the most obvious thing is that the sales team should love it. I think if the sales team can imagine how they will sell it, they can imagine that that will go down really well in the customer discussions they have. I think that's a pretty good sign. I think the thing that that doesn't capture, and so obviously the whole product management piece is the thing that there are many books written on it and it's a very specialized role. And I think one of the things that makes it most difficult is that customers will be nice to you and they will lie to you, you know, to save your feelings. So you've got a vision of something that you think would be amazing and they may well say, you know, that sounds lovely. And you take that as a kind of like, great, let's build it. So there's a book I would really strongly recommend, which is called the mom test. It's my favorite, but it's basically how to ask questions in a way in which people want to be nice to you. can't tell you what you want to hear. I would combine the people who really talk to the customers in your business. I think that's really important. But I would try and combine it with a bit of the rigor of that that comes from that book, The Mom Test. Simon: That's that's mum like as in mum for mother, right? Right. Daniel Lord: That's right, yes. Yeah, I think, yeah, the author starts with the idea that if your mum so wants what you do to be successful and you to feel good, she will tell you it's great. So how could you even trick her into telling you the truth? Simon: Very good. Interesting. I'll have to I'll have to look that up. and given what you see across your client base, w where where do you think that that clo analyst firms should be focusing right now to deliver genuine client value? We took touched on this before, but I'd I'd be interested in your take. Daniel Lord: Yeah, I think the first thing is to have an appreciation of how much is changing and to focus on what you do know and what you don't know. so if I was running an analyst firm at the moment, I think there'd be three pillars to this. So the first thing is I would get really close to my customers. do everything I can to make them feel loved, but also to understand their deeper needs, to understand, you're using AI, so great, you do that, but what challenges do you have after you've been doing that? Where do we still fit in and so on? So I'd have that really as deep as possible connection with customer and really understanding their needs, not at a superficial level. really deeply understand them. I think the other thing is that I would spend as little as I could on kind of anything else. So I wouldn't be doing any big technology projects for a start because I just don't think, like the whole ROI case is typically built on it being around for five to 10 years. We're not in that world. So I think I wouldn't. Yeah, I wouldn't be having the distraction of that, which is all predicated on the idea that I can see the future. And then the third bit is, Simon: Yeah. Daniel Lord: you know, that therefore, I think you've got a kind of a resilient kind of anti-fragile business if you take those first two approaches. And then the thing you want to be doing is experimenting, you know, talking to your customers, using the kind of off the shelf AI tools, you don't need to spend very much money, but just be imaginative and creative with them and work out how can we, how can we be creating more value for you? And I'd be doing a lot of that. And then there'll be a point, you know, I don't know when it will be, and it might be uncomfortable waiting for it, but there'll be a point over the next few years where you'll start to think, actually, you know, things are settling down a bit and we we now have a view that that experiment is something we need to scale up because that's got legs and deserves some investment now. So I think I would take those three approaches to having a business that is successful really in navigating a period that's going to see a lot of change. Simon: Okay. and i if if a firm came to you tomorrow with with you know what you'd call a limited budget and asked where to spend it first, what would you tell them? Daniel Lord: I would say fly to wherever your best customer is and take them out for lunch. Simon: and and where's the biggest gap in understanding you see right now among analyst firms around AI more broadly? Daniel Lord: I think misunderstanding probably comes to conflating the different access technologies we talked about earlier, the idea of training and trust and enabling usage. So I think there's a conflating of those. also, there's a difficult conversation to be had within the analyst firms themselves where you reconcile different stakeholders who are looking at the business. obviously from different perspectives, also in different timeframes. And, you know, trying to work out what the customers want in all of this. So, yeah, think that that's, you know, I think it's a genuinely difficult piece to unpick, but yeah. Simon: Great. and how's AI changing the research process itself? Do you do you see much of that? Is is is there a risk that firms adopt AI very much at the delivery layer without reinventing how they how they operate? Daniel Lord: Yeah, I mean, think it's of everything's everything's up for for innovation, isn't it? There's there's the I mean, we often think about it in terms of the innovating how the how you get from the data through to the client use of it. So, you know, that might mean packaging the data in novel ways. I assisted to kind of create different research packages that might be suited to to AI consumption and so on. So there's a lot there. There's also, there's probably the aspect of like how you get the data in the first place. There'll be ways to either to make that more automated and more efficient or maybe to obtain it in different ways or to obtain different data in a different way. So I think that there's lots of innovation on all of that to come. But I think the other... thing that's worth noting is there's a lack of trust around things that incorporate some AI. So I think there is still a case to have some of your research products explicitly not being made by AI, not having an AI part, that's something known that people feel they can fall back on. So I think we're in a kind of period of a lot of change. But part of what makes it complex is it's not just about predicting the future. It's also about continuing to do the things that people need, you know, that they've depended on for some time. We need to do both of those things through this period of change. So yeah, I think some research products may continue to be made exactly as they are, or there might be slight AI augmentation in ways in which you can still satisfy clients that it's something that they can rely on. Simon: Yeah. and i i if you could get one thing through to analyst firm leadership at the moment, what what before they particularly before they sign off on their next big investment, what would it be? Daniel Lord: I would, I'd be very keen for them to appreciate that that isn't their specialty. And there's a sunk cost fallacy. I appreciate it's not ideal where they've got to, but there are ways of extricating themselves from it. But there's always the temptation to think it's just one more thing. And really it isn't just one more thing at the moment. So. I think I would be finding ways to focus on what your core competencies are. And I think it's quite illustrative to look at the kind of parallel world of investment research. So we're not, content catalyst isn't really an investment research tool. But when you look at research firms there, my understanding is that nobody builds their own, it's a real edge case that anyone builds their own technology of the type that in enterprise research, industry research, market intelligence and so on, it seems to be much more common that people would build their own technology. So yeah, I would suggest not being a software company. There's a lot of things that will keep you up at night with that model. Focus instead on what your core competencies should be. Simon: Brilliant, thank you. and is there anything else that you think nobody's talking about or it's not being talked about enough that should be? Daniel Lord: I was quite interested in the idea, the Harvard Business Review had just an interesting article I thought in terms of looking at the different biases of LLMs. It turns out that all of the LLMs are biased towards, like very strongly biased towards any question you ask about the future of AI and people is that they give you an augmentation answer, not an automation answer. And, you know, whenever, whenever we're toying with our ideas now, we always run them past the AI, don't we? And, you know, I always enjoy reading like a variety of thought leaders. on this space and on our space. But I'm conscious that the augmentation argument always comes through, which I don't, you know, we all want to be true. I want it to be true. I think everyone wants it to be true. I just, the bit that makes me feel a bit uneasy is that maybe even as we say, like the human should have judgment, maybe we're already letting the AI judge that one for us and we should be thinking of bit more about how some of this looks in a kind of automated world, I say that very much wanting the augmentation world. Simon: Yeah, and like most things probably it'll be somewhere in the middle or something completely different with the way things are going. Daniel Lord: Yes. Simon: great, well thank you so much.