AI isn’t making its way into deal sourcing now — it’s been here, behind the scenes, for a long while. But with the rise of LLMs, AI agents, and more popping up in origination workflows, firms that overlook what they put into these tools and how they put them to work risk getting swept up in the hype. Sourcescrub CEO Prescott Nasser and Filament Syfter Tech Co-Founder Martin Pomeroy are fintech veterans who’ve seen the industry go from tedious, manual processes to the data-driven and automated practices that power it today. In this conversation, they break down those insights, sharing where data was, where it is now, where we’re headed, and how to get the best results.
Both Martin and Prescott emphasize the importance of not leaving the heavy lifting to AI tools. It starts with the data. “It's quality and quality out,” Prescott states. “You need to have a quality data set. The value of market data providers is to do the cleaning and the aggregating for whatever their data set is, and then provide it to others to then wrap into your other workflows and processes.
Martin adds, “Sometimes you'll go through your process and you find the accuracy was not quite where [it needed to be]. And that's always down to the input data. So, rubbish in, rubbish out.”
Powerful outputs require powerful inputs, making breadth of data a necessity. “You need the pond to go fishing in,” Martin points out. “What is the pool of businesses that you may want to do business in, and how do we use AI to start prioritizing and scoring those?” The value AI brings here is undeniable, but the human aspect is vital for producing results.
With more players, more platforms, and more data-driven processes in the market, standing out is paramount, and that can’t be done without the human connection. "Getting in person with someone — [AI is] never going to replace that. And as someone who's gone through the process, it's hard to discount the value of feeling good about the person on the other side. All things being equal, who made me feel the best, and who makes me feel like they're going to be the best partner? These are pieces that are harder to automate,” Prescott explains.
There’s a new edge that firms can gain by striking the right balance between data, AI, and humans, but it requires quality and an enhanced strategy. When what you put in is what you get out, dive into the insights shared in this video to know how to leverage this combination to achieve the best results.
Transcription:
Mike Waite (MW): Good morning, everybody. This is Mike Waite with Sourcescrub. Appreciate you joining us for our webinar today. We're going to give everybody who's trickling in just a couple more minutes to join, and then we'll get started. So sit tight for just a second. We'll get started.
Members are still clicking along here. But why don't we jump in and get started again? Thanks for joining us today. We've got a great topic. Data and AI and Deal Sourcing: Proven and Practical Applications. We've got a couple of really thoughtful speakers joining us today for the topic. The thing that led us to this topic of artificial intelligence and machine learning have been a really important part of deal sourcing technology, almost from the beginning.
But it was just a couple of years ago that ChatGPT and these chatbot-style LLMs emerged and really opened up an exciting new window for innovation and impact using these technologies. The two speakers, Prescott and Martin, joining us today, have had a front row seat through this whole process. Prescott literally helped to create the category.
Martin literally helped expand the category with the introduction of multi-source deal sourcing platform. And so they have an incredible perspective to talk about what they've seen, but also to share their thoughts on where things are going. So, why don't we very quickly introduce our speakers for today? Who, if I could get the slide to advance. There we go.
Martin and Prescott, why don't I invite you guys to just give a quick introduction to your companies and technologies before we move on and talk a little bit more about what we have in store for today? Prescott, why don't you jump in first?
Prescott Nasser (PN): Prescott Nasser, Co-founder, CEO, Sourcescrub. What do we do these days?
Private. Private company intelligence, private market intelligence deal sourcing platform.
MW: There you go.
PN: Yeah, there you go.
Martin Pomeroy (MP): I'm Martin Pomeroy, one of the tech co-founders at Filament Syfter. We provide the backbone architecture along with the AI know-how to help firms build proprietary internal systems that bring together the data points for deal sourcing and monitoring, mainly.
MW: And just to be clear, for everybody in the audience, Filament and Sourcescrub have been teamed up for what, Martin?
A little over two years now, providing the data. And as Martin said, the backbone and the technology aggregation that allows companies to take advantage of multiple data sources and drive all kinds of interesting sourcing use cases, running through the topics that we have for today, a couple of quick housekeeping notes. And, I want to invite Prescott since he literally helped start the category to talk a little bit about history and context for how we've arrived at where we are today to sort of set the scene for the conversation, and then we'll and then we'll get into our conversation.
We're recording the session. I think everybody heard as you joined the session, that's so that we can share the recording out with you guys and everybody else who registered, but hasn't been able to attend. I think everyone's familiar with Zoom and how to share comments and questions inside the platform. We will be taking and responding to questions throughout.
And then save time at the end to focus a little bit more on audience questions. So please feel free to submit your questions along the way. And so with that, why don't we why don't we jump in? Prescott, why don't I ask you to give your take on the history of digital sourcing technology in the category and how we got to where we are today?
PN: Yeah. Thanks, Mike. I don't know how brief it will be. I could pontificate for quite a while. So, you know, the the, this fall is kind of just a general trend of, as we the, the other chart that we're not sharing and, but a lot of people have seen this kind of the maturity curve for, technology in deal sourcing, and private equity firms and the so this is kind of similar-ish kind of curve or history.
And so when we started Sourcescrub in 2015, you know, it was a pretty manual process. I think we had a lot of customers, that were, were still, well, by not combing newspapers in 2015, but started by combing newspapers with, classified section or the new business section, but but very manual processes, particularly around conference intelligence, pulling the lists, figuring out who the companies who are finding the contacts.
And then, so that's kind of where it started. And then you kind of moved into, when we started actually in CRMs, even in 2015, weren't necessarily the de facto kind of, you know, not every customer had a CRM. Probably not even a majority of customers had a CRM, even in 2015. And then, you know, that started becoming a standard part of the playbook, a standard part of the tech stack.
And you get a little bit bigger, you grow a little bit more, people get a little more sophisticated, and you start getting into the the, the best of breed kind of data solutions, the warehouses where you're combining multiple data sources together, you realize your CRM, your data sourcing platform Sourcescrub was not enough. You needed more. So you get an, I don't know, credit card data to that or sensor tower data, or if you're B2C, and so, Filament's kind of they've got a good story there.
And Martin can talk about that. And that was probably, you know, for the larger firms, it was a very dollar-intensive exercise, typically built in-house. So you had engineers in-house, you had infrastructure in-house. And for anybody who hires or pays engineers, you know, that, you know, get a couple of those, add some infrastructure, add some data cost to it.
And you're in the low seven figures pretty quick. And so only the larger firms kind of had those kinds of budgets. And so that was, 2016-17-18 timeframe, where they were building internal warehouses and then, I kind of want to say, Martin, what is it, 21-ish, maybe 22-ish was really around the time where, the warehouses, kind of the Filament, the Syfter platform and those like them came out where. It's okay. You don't have to necessarily build it in-house. You don't have to figure it out yourself. You can work with others who have offerings or platforms. And then, of course, probably the topic everybody's really here to talk about is in the last two years or so, ChatGPT, OpenAI, what they've done with LLMs, a lot of the agent frameworks, as probably the last couple of years.
And so we went from a very manual process to, more technology-based process, kind of a big data process. And LLMs are kind of a big data, but maybe in a different, a different kind of vein, which is where we're at today, and I'd say it's pretty early. And so, I think that's what we'll spend the hour talking about.
MW: Yeah, it's been an action-packed decade. Hey, Martin, Prescott kind of opened the window, but what would you add to that? Obviously Filament Syfter's emerged more recently than Sourcescrub, but what what did you what did you guys see? And what was your vision for the platform that you wanted to create at the time?
MP: Yeah. Great question. I think it's actually around the data warehouse days. I think this is when Snowflake and a few other big providers, you know, you know, hit the market. One of the things that we have seen was a lot of people, I'm gonna say, experimenting themselves internally, with I'm going to say self-built small solutions. So this is powered by someone in the business who's a creative, power user, some of these tools, and just really starting to kind of show some of the, the benefits of, well, we don't need to be copying and pasting all the data back and forth and have it static.
We can do some lightweight integrations. And I think, you know, at the same time, a big push on data directly into CRM, which is great and showed the benefit of like joining the dots and having a single view of companies. But that's obviously at a smaller scale for the known entities. So our journey really began, you know, I'm sure you've all heard of Mother Brain and similar ones, saying, well, actually, how do we take what we've proven internally?
I'm going to say, quote unquote, a bit of a hacky way, and scale that, and actually take this idea further. So that's really where it began. But as we'd discuss and go through the kind of strategic thinking, you know, it evolved to, well, with LLMs coming, we need to start building that fuel regardless of the use cases we're catering for today.
Technology is moving so quickly when you know what changes it comes up next week. We want to take advantage of it very quickly. Rubbish in, rubbish out. So it's evolved from small use cases to now big strategic drives to saying, well, we need to own the data and also increase their own enterprise value of the firm. We're building up IP that's going to differentiate them going forward.
MW: Excellent. That's really good history and context. So why don't I drop the screen share? Bring you guys front and center, and we can get into the conversation. I want to start with a simple sort of a prediction idea to set the stage for the conversation. So, if you were to describe in a sentence the way that you see the future of private equity, dealmakin,g and deal sourcing, what is it?
What does it look like? Martin, why don't we why don't we start with you on this one?
MP: I think it's going to be a partnership between humans and machines. Ultimately, to win deals is about relationships, and it's not going to replace any of that aspect of it, but what it is going to do. And as we know, an AI machine that is giving an entity an assistant doing some of the legwork in the background.
So when you head down on a deal, it's constantly being available, looking for the right signals. What do you need to know at this point in time? And pushing that data? So definitely a theme from moving from pulling data to pushing data I-E it, knowing what you need at the right time. But I will summarize it as a partnership that can play a big part, but only a part of it.
MW: That's, it's a nice, optimistic look ahead, right? I mean, breeds plenty of darker predictions that we can read about in this world. Prescott, how ‘bout same, same question to you. How do you briefly sum up the way you see the future of deal sourcing tech?
PN: I've got to think about that. Martin's got a pretty good answer.
I think for the next couple of years, it's going to be tumultuous. And, it's going to be interesting. And I think this conversation'll will head on that a lot. But I think there's a lot of platforms trying a lot of things and a lot of tools trying a lot of things. And, not really sure it's clear what's going to be the best solution.
I'm not really sure it's clear, but I think Martin's right on. It will be more of a push than a pull, in that, you know, notifying you that something has interesting has happened, and helping you reach decision makers, founders, executives, kind of more timely, will probably be the model as we continue to evolve.
MW: Yeah, yeah. Well, you guys have hit on a popular notion. Is this partnership between the technology and people, and people taking advantage of what they can uniquely do versus these new technologies and what they can you uniquely do to, you know, expand the potential. Right. So that's a good topic for us to return to. But, for now, right.
We've talked about LLMs and these chatbot-style interfaces. Obviously, that's probably the most disruptive force right now. But what or what other trends and technologies do you feel are having the greatest impact? Right now, as we're thinking about what the future holds. Prescott, can I put that one to you?
PN: Other than LLMs?
MW: Yeah. There are other things out there.
PN: Oh, man. I mean, yeah, but I think that, I think they're less new, I suppose. I mean, so like, Mike, you obviously know, you've been with Sourcescrub for a while, but we use ML behind the scenes, and we've used it for years behind the scenes. And, some of some of the AI wrapper hype today is, AI is a large bucket, and under AI and in a lot of folks, the way they use it is ML and other things.
And so I wouldn't say they're necessarily new. They've been around a while. They might just have a greater visibility today than they did historically. I think, you know, Martin could probably talk greater about some of the data warehouse technologies that are out there, because that's kind of more their wheelhouse. And so, again, I think a lot of those have been around a while.
You know, Martin hit on, like, Snowflake, Data Bricks, all the big cloud providers have their own kind of offering, if you will, for that. And so I think those are maturing. And so that kind of adds a layer there. But I still think those are not quite as new as LLMs.
MW: Yeah, yeah, yeah, yeah, it seems like, and Martin, I think one of the things that that's Filament Syfter saw, as a lot of these technologies that were accessible only to great big firms with loads of resources and technical expertise.
Right. But they're always the ones that are going to be sort of in the vanguard and trying things because they can afford them. And certainly a lot of what's exciting is that some of these technologies, powerful technologies, are now becoming affordable and available. And accessible to, you know, more, the middle pack, I guess, is the way I would describe it.
But without sort of as an intro. Like, let me turn that question over to you. Like what? What technologies do you see really shaping the possibilities out there?
MP: So Filament's been going for about eight years now and we've been doing, I'm going to say, similar approaches five years ago that we're doing today. So what LLMs have done have actually just made it quicker, cheaper, and easier, and expanded the art of the possible.
But the technology built before LLMs, when you look at more traditional data science. So a lot of the things we would start off with, the kind of warehousing is using more traditional data science techniques, regression, classifications and so forth to try and achieve the same outcome, which is how do we take a firm's investment pieces and bottle that into a set of, say, more traditional rules or traditional machine learning, some IP classifications, and that technology's been around a long time, as you say.
Then ChatGPT came along, and that's got people thinking. But actually, you know, people look at ChatGPT as a chat interface. That's quite a place. That's great. But the technologies what we are seeing, is it being used in the background for enhancing some of those examples I gave. So you know, a practical example that clients are doing is they are automatically preparing pre-meeting briefs.
Here's a meeting going to meet this new client. What do we know about them? And that's then summarizing if they're in CRM. It's going to Sourcescrub. It's getting your data. Have they got any conference lists? Was the headcount growth looking at SharePoint, looking at other financial sources? Summarize that into one email and then sending that out to the dealmaker.
So that example has been, you know, something that's been done in the past as well. But the I'm going to say the quality and depth and the nuance that you can get in those results now is just changed overnight with the with the LLMs. So, same stuff with a new technology. Makes it quicker, cheaper, expands the what we can do.
And just a point in that actually, what we are seeing at the moment is using it to automate background workflow, versus just a chat interface that you all refer to when looking at these types of tools.
MW: Yeah. A big part of what I've heard you describe as the vision is the importance of owning your data using proprietary data.
Why is it so important to, quote unquote, own your data? And where does data quality come in?
MP: So I think, as I said at the beginning, the very first thing you know stands out to me is if you don't have the fuel, the data starts building up. Now, baby steps, building foundation, then you're going to be catching up when, this technology, which seems to change, as you know, weekly, daily, sometimes, but essentially that's how, you know, if you look at how you get alpha, it's all the internal knowledge that you've got internally.
And how that connects to the external one? So if you've got the silos of information, you're not really applying all that knowledge that the firm's built up over, over the years and coming up with a unique insight or data point that the other competitors can't come up with because, you know, you, you can only do that by owning the data, bringing that in-house, combining it, and then getting creative on how do we get our unique insights out of that.
So, you know, summarize it saying that, that if you remain in silos, you don't have the opportunity to, I'm going to say, surface some of those signals that would never be spotted until it's all combined. And with that, you know, it's AI technology to find those.
MW: Yeah. And Prescott is as, as more of a syndicated data syndicate is not the right word, but you know, data provider.
What would you add to that as far as the importance of data quality, and maybe more than that, like data history? Right. Because it's having history that goes back, that's been proven and allow you, I think, to build successful models.
PN: What I say about quality in history. Yeah. I mean, I think that, you know, ultimately it's kind of it's, it's quality and quality out, and that applies to the raw data, whether it's whether it's the data from a data provider like us or your CRM data. I'm sure a lot of folks on this call understand or have dealt with like CRM.
And I've certainly had conversations with plenty of our customers about the quality of their CRM data and the structure, and being able to pull things out of it correctly. And so I think that still holds true. That you're going to want to have, we're going to need to have a quality data set, kind of across the chain or kind of horizontally, if you will.
Right. And so, the value of, of market data providers, I think, yeah, their job is still the kind of do the cleansing, do the, the cleaning, do the aggregating for whatever their data set is and then providing it, to others to, to either via our platform in our case or to, Filament Syfter platform or, or even to the LLMs to be able to, to pull that kind of cleanse to data, to then wrap into your other workflows, your other processes.
MW: And, and not to put you on the spot, Martin, no need to name specific names, but can you think of particular instances where proprietary data and data quality have led directly to better outcomes?
MP: Good question. I think when Prescott was speaking, the first thing crossed my mind straight away as well. And I'm going back a little bit in the history of Filament where when we've worked with firms to try and bottle their investment thesis, we come up with certain scores that were or, signals that we're looking for.
And sometimes you'll go through that process and you find, well, actually the accuracy was not quite where it's at. And that's always down to the input data. So that, you know, rubbish in, rubbish out is first thing that came to mind. The second thing is, what if we look at versus a real example of trying to find net new deals, the first thing you got to do is define your total investable universe.
You need the pond to go fishing in. And I think, you know, that's certainly why we love partnering with Sourcescrub and data warehouse offering, because we've got that raw data in a format that we can really work out. You know, what is the pool of businesses that you may want to do business one day?
And then how do we use AI to start prioritizing those and helping score those? And monitor those, and so forth? So one client comes to mind the moment. So I've just recently gone with them. We've gone through some KPIs and in their case across their teams. We found or recommended around 200 new companies over the last two, three months, and about 63 of those were deemed as really interesting, pushed into the CRM, then followed up and and now further down the pipeline.
So that's 63 new businesses that they weren't looking for. But by training it on the investment thesis the importantly helping the right data set that's got the right data points to capture that and accurate to make sure that we're not sending ones that when they dig in further, it's, you know, the waste of time looking at the business because it's disqualified.
It's due to an incorrect data point. You know, that's the direct KPI value that they get instantly. So yeah, to have accurate and say timely data is what causes good outcomes. If any of that input data is wrong, it kind of goes down the line. And when you get to the end of the kind of process, you will find the outputs that you don't expect because of that faulty input, if that makes sense.
MW: Yeah. Yeah, absolutely. You guys both kind of opened with rubbish in, rubbish out, which is interesting because with our, there's a new kind of rubbish ,which is hallucination. So maybe we'll, maybe we'll come back around to that. Of course, there's a correlation between the quality of the data and whether or not an alarm is going to make stuff up.
Right. But that's an interesting new kind of rubbish that I think maybe we want to get back to you and talk about it a little bit. In the meantime, we were talking about, and I think you introduced it, the, you know, there will be roles for humans in the process. There will be roles for technology in the process.
In today's situation, what what are what are the areas of the deal sourcing processes that are most affected by technology and AI? Let's start to start to make this human.
MP: Yeah. I'll go with two kinds of core use cases obviousl,y that we will help lots of firms with. So that initial screening or disqualifying.
PN: So even if you know your advisor led or if you look upon that new being sent, a list of companies, current processes, you know, have a spreadsheet, copy paste data, and evaluate them and disqualify them or pass them on at the right time. That using AI and models to just do that initial evaluation of companies at scale is the obvious quick win that we're seeing all the firms do.
MP: So, having every company scoring that sure investment thesis every day is new data points coming in, never being stale, and then surfacing the right ones is kind of the use case. A lot of them start on. The second bit is actually around shadow monitoring. So all the companies that are in the CRM are great. We found these new ones.
They're all in, but how do we make sure that they're being correctly monitored? No one's pick up the phone and also be within the deal because you added more value and build a better relationship than others. So one of the bits of technology that came out over the last few years is around classifications or signals, should I say, not very obvious.
One is a blog being published on a company's website, and realizing that it's about a new non-executive, or that's about a new office, or a new X, y, z. So that ability to monitor websites and Sourcescrub and other sources connect that with really what's in the CRM and just be there to alert you when there's something that you feel like, it might be a reason to get in touch.
And so those are the net new or scoring at scale, and then monitoring. So I set up companies. We want to keep an eye on these. And we want to look for certain signals. Certainly the kind of two ones that were traditionally been solving. And that's really evolving now into more advanced workflows. As you can imagine, with you know, LLMs, when it can start generating content versus just understanding content.
MW: Yeah. Prescott, I'm curious what you would add to that as far as where AI, as having the biggest impact, and maybe even focus more into where are humans, you know, assuming it's a limited, precious resource. You've got these expensive MBAs in your firm that are grinding out long days. How are they in this, in this new sort of hybrid environment, you know, how are they best deployed? How are they best used?
PN: Well, let me shift that maybe that question a little bit. You know, I think what we're what I'm most excited at seeing, a lot of our customers experiment with our other parts of the workflow. So, you know, deal sourcing, finding the companies that's important. The follow-up, the timely reminders of the timely notices, everything Martin just said.
Right. That's totally true. You know, won't dispute any of that I just want to shift a little bit. I think there's a lot of other stuff that those freshly minted MBAs are working on. The IC memos, the decks, and those kinds of things, which for me is probably some of the more interesting things that I'm seeing, some kind of all the platforms come out with, which ultimately play into having to have, kind of a bunch of different data sets.
Right. So, so for those that, that, I for like an IC memo, you've got things like what is the market, what do we think the market outlook is for the sector. You might talk specifically about your target company, from PDFs or PowerPoints, or pitch decks, or financials they may have sent you. And so pulling all of that together in kind of a first pass, IC memo, which a lot of it is, I mean, a lot of those are templatized, right?
A lot of firms have kind of this is the format they want the IC memos, and it's the same stuff. It's pulling all that data together, and it's just a time-consuming effort. And so that I, that to me is pretty interesting, just to see how that, that plays out, and see how that works. And there's a lot of different platforms trying, a theme on that, but ultimately needs all the data in there to be able to pull that together correctly.
The human part. So really it's interesting and I don't I would this would be a kind of survey question, I suppose, to others or those listening. But so we when we started in 2015, it was just a, it was a different world in terms of the number of emails a founder or CEO was getting from the private equity space.
And, and, you know, as we started to grow up in the field, started to grow up and the, you know, the sector started to mature, you get a lot more platforms, a lot more players, a lot more data-driven processes. And so they get a lot more emails. And so now it's how do you differentiate so that you can be heard because you know, any of the companies, generally speaking, that are interesting are probably getting many emails, many intro emails.
And so how do you how do you make yourself stand out? You know, some of it will be with the contextualized emails that you're, you know, LLMs are, are pretty decent at and again, you'll need decent data to be able to, to contextualize and reach out to Martin and say, Martin just saw you you guys just did a round.
Congratulations. We want to talk to you. And you know, or saw you at a conference, for Sourcescrub data or whatever the news article was that was published, or something that kind of imply that you're paying closer attention to them than maybe you really are. And so that's kind of, I think, still a challenge.
And I'm not sure anybody's fully got the kind of customized emails with LLMs, but that will probably come soon. And I'm sure some folks here have got trials of that working on, so I'll be curious to see how that plays out. And then I mean, for us still, I mean, you know, Mike, conference intelligence, is one of our core tenets.
And so getting in person with someone, they're never going to replace that. And as someone who's gone through the process, you know, I it's hard to it's hard to discount the value of feeling good about the person on the other side. You know, it's, you know, sometimes it comes down to who's going to pay more.
But, you know, all things being equal, who made me feel the best, who makes me feel like they're going to be the best partner? Which are just pieces that are harder to. Yeah, never going to automate that.
MW: Yeah, can't do that with technology. Yeah, we did a recent founder, we did a recent founder interview and talked about this very topic.
You know, how do you make these choices about who you're going to work with for financing, and his top of the list was, you know, the airport test, right? Is this a person I want to spend a couple of hours with, stuck in an airport? So, of course, money is important, right? But it's interesting how important that relation and comfort is with your partner as well.
Sorry, Martin, I interrupted.
MP: Yeah, no, I just wanted to give a, against a practical example of what Prescott was saying around those outreach emails. Because it's certainly something that we're exploring deeply with a lot of firms at the moment. It's really interesting on a few different approaches. So one particularly is we we refer to it as joining the dots.
And the question really is here. And you could look at a news article, and you can generate an email based on that. And you can use, take some data from Sourcescrub, some from CRM, and kind of personalize it based on the person, and send it. One of the questions we looked at was, how do we not just use an event, but also use the context locked up in the CRM?
So, an example I'll give that within a firm is, I think the best one to explain it is, there's a note in the CRM saying we met that company. They're not quite ready for transaction at the moment. CFO mentioned are looking to launch in the US next year. We'll get back in touch then we can understand that.
And then in the background, and maybe they mentioned they're too small as well at the moment, we might also connect to Sourcescrub in the background. We then look for those signals. So maybe we could take an article saying they've made the first hire in the US. Maybe we start seeing them appearing in conferences in the US, based on the sources in Sourcescrub
And then kind of join those dots and then send an email to the dealmaker saying, well, we feel like now's the time to be engaged because what you've noted has now happened. But then also summarizing that into an email template, they can always top and tail it and send it. And I think that helps a little bit of answer Prescott's question of how do you get to the top of the inbox?
And it's making sure when you're adding value and getting in touch at the right time for the right reason versus the pay praying spray. So interesting how that's using LLMs to do a number of steps to connect different data is but for the same output.
MW: Yeah. Yeah. So let's flip it around a little bit. As firms are preparing for the future, they're making technology investments.
They're thinking about how to reshape their organization and maybe even rethink the talent that they're looking for and how they put all these pieces together. Like what? What are the biggest mistakes that you see people making? And, you know, what's the what's the real sort of downside to making those mistakes? I'll let that one dangle out there for whoever wants to unmute their mic and jump on it.
MP: I can maybe, maybe have a stab at it.
So one thing I always worry about, I guess in either firms, is looking at it too short term and maybe looking at point solutions versus a strategy. There's so much out there now. That is a tool that will solve this particular task. But I think, as Prescott said at the beginning of the call, we don't know who the winners are going to be.
And I suspect that tomorrow is going to be a new technology that didn't exist today. And not wanting to, I'm gonna say, tie yourself in with one particular solution. So I feel like that's a risk in coming up with an architecture that means you've got the fuel. It's logical, the data. You can switch out what's running on that data.
So flexibility and making sure that it's not looking at a point solution is probably, you know, something that I would flag is being a signal. It might not be thinking about it quite in the right way or saying that you do need to start somewhere. You need to get your hands dirty. You need to get used to this type of technology.
But just making sure that, you know, whatever is coming in to the architecture or looking at there's a a way to bring those together, versus new silos. Now, just silos of AI.
MW: Prescott, what would what would you add?
PN: Yeah, it's interesting to think about it as, like, mistakes being made, because I'm not sure. I don't know that I would call anything a mistake.
If you're jumping right into one of the new AI tools, and you don't have anything, like Martin said, you kind of got to get used to these tools. You've got to understand what they can and can't do. It might not be as powerful as you want it to be because you don't necessarily, as Martin says, to have the feel.
You might not have the clean data or the data aggregation, quite right or solved, but you still are going to need to experiment with the tools and play with them. And so you know, I would agree as a data person, you know, I agree on the underlying kind of clean data, data architecture, and other things as an engineer.
But, you know, that's always going to be a useful kind of task. Do you have to start there now? You know, if you haven't started anywhere, though, you could absolutely go start with one of those tools and start playing with it. And I think, you know, we talked to a lot of customers and a lot of them are playing with those tools.
Some of them have kind of a data core set up. A lot of them don't quite have too much of a data core, or maybe their data core is just simply their CRM today, which is, again, like a it's a super common place for a lot of folks to be. And it lets them kind of play around and toy with it and either start to see some value or, you know, not see value based on what those platforms are specifically trying to target and do.
But it at least gets you comfortable and understands what the edges are, and will get you comfortable to understand what you have to do to make it more useful to you? I think ultimately it'll end up with, you've got to have a better data strategy under the hood. But again, if you don't have anything today, like, I don't think it's wrong to go start playing explore, exploring, experimenting with those tool sets.
You know, from my perspective, I think it's, it's, it's really early. It's really early. And most of these platforms, I mean, some of them have raised hundreds of millions of dollars, you know, in the last couple of months. And I don't think most of them have been around for more than a year or so.
Certainly not in any material way from a revenue standpoint. And so, you know, it's going to be really interesting to see what happens in year two and three of renewals and how many people are sticking with it or using it versus how many people are jumping in because they feel they have to jump in? I will, I mean, I think, I think, a lot of folks feel like they have to they have to do something, you have to.
Yeah, I have to be knowledgeable about it. You have to be able to talk about it. And so sometimes it just means signing a contract to get a tool to play with something. And I think there's probably a fair amount of that going on. And, you know, like any tool, they'll come to a year, you know, the year mark or whatever, and evaluate it and see how it does.
And, you know, it's going to be interesting. I mean, it's talk to a bunch of folks, it feels like it's a VC play right now because it's go do a land grab. You know, try to get big, try to get lift off, but there's going to be just a lot of players. And, it's not totally like a lot of them are built off of the big models off of kind of, you know, names.
And so the question will be is who's building the better interface, or which interface works or finds traction at the end of the day? You know, with, in our case, private equity or investment banks. So it'd be, it'd be interesting to watch I don't think there's a wrong thing to do. I mean, if there's a wrong thing to do, maybe it's going from 0 to 100 miles an hour and, you know, saying I don't have anything.
And now I'm going to spend several million dollars building a data internal data warehouse versus, you know, partnering with someone like Filament to kind of get you going. Right. And Martin don't... Close your ears. Right. Like we, you can always go with a Filament and then decide later that you've outgrown Filament's capabilities to help you and then kind of grow up past them, or outgrow them.
And I know, Martin, that's not your plan. Your plan is to grow as fast as anybody will. But, you know, that's still probably a smarter stepping stone than jumping all the way in the deep end and hiring engineering talent and trying to figure out how to build it on your own. And then for anybody who's aggregated data, like there's not a key that just magically this, this this record is the same as this record and the other data system.
And so it's just a yeah, it's just a mess of a process.
MW: Yeah.
MP: I kind of agree more on, you know, starting at the right place. So one of the things that we do with the firms is try to make it really clear that it's a journey. It's a, it's a, it's a roadmap. And, you know, we ensure that they've got the, you know, the right vision and product manager internally to do it in stages, you know, get the foundations right, get the data right.
And experiment quickly with use cases that are going to solve real problems. And then do it over, you know, a 12-month or two-year plan of what you're going to bring in to solve which tool, and how are we going to do this in a phased way? So we're not just, you know, overwhelming everyone. And yeah, as you know, if anything, it's all about change management and adoption, less about technology.
It's how do we actually safely and sensibly and
MW: Take advantage of it. Yeah. So I mean, I hear what you're saying, big mistake could be jumping too far, too fast, and getting ahead of yourself. It feels like also one of the biggest mistakes you can make a standing pat and not experimenting and innovating. I think one of the interesting things about the customers that we serve is that these technologies have important relevance to their businesses and the way they operate, but they also have potential for a huge impact on the potential acquisitions that they're looking at. Right?
So I think on either side of that equation, you want to make sure you're familiar with the technology and the possibilities and getting involved with it, experimenting with it, getting familiar from a, you know, actual use case, perspective seems like it would be really valuable.
MP: Yeah, it's quite interesting. I read a report from Bain saying, I think it's 2022 and 2024 were all about deploying AI import codes, and 2025 is about bringing that in-house.
So when we speak to firms, they've got a data and AI operator and team who are experts at helping their port codes adopt this technology. So the question is, how do you adopt that for yourself?
MW: I'm sure that's becoming a really important operating role for a lot of these firms. I'm sure it is. We're down to probably our last five minutes.
We've got a few questions in the queue. Why don't we turn to some audience questions and, as we do that, I invite anybody who's got a question that they want to throw in here based on the conversation so far, please just throw it into the Q&A panel and we'll we'll get to it. The first one that came through is, how do you view the value of market data platform providers in a world where all data is, in theory, available through commercial LLMs?
PN: In theory, it's doing some heavy lifting in that statement.
MW: Yeah.
PN: Yeah. I mean, it's a good question. And the reality is, we deal with it. On every renewal, we deal with our every new customer who says, why can't I just use ChatGPT to do the job? And, you know, I think we hit on a little bit. I think so far, the answer still stands on having a clean, clean set of data that's been curated and kind of cleansed at scale for, you know, we have a specific focus on what we focus on.
Others focus on other things. I think that's not going away anytime soon. You know, will LLMs get there? Maybe. But I think again, the “in theory†is doing a fair amount of heavy, heavy lifting in that statement.
MP: Go in kind of personal view on this because I have kind of questioned around, well, you know, like yourselves, the firms know the value of their data.
So rather than that data being divided into, you know, a ChatGPT, what I predict is things like Sourcescrub and others providing MPC servers at a cost for these models to have access using it, simply adds an interface to bring that premium data into the LLMs. Technology today, as well, is the yeah, it's moving and there's information out there, and LLMs could do it.
I think it'll be good at doing it on request. I tell me about this company, what's not going to be scalable or is going to, you know, cost so much is having doing that scale in the background across millions. But if it's already got the data and the core bits, you need to only rely on the LLMs for those initial bits.
It, you know, it becomes scalable. So we'll spend a lot of time looking, delving into LLMs and so forth. But you need to combine it with the known data. Again, back to the rubbish in, rubbish out until it becomes accurate, you know, going to end up. So I feel like it's going to end up with lots of mini LLMs, speaking together like API's versus one master one that's going to have access and replicate all the data itself.
PN: And, and I mean, just that's kind of what we're seeing at the moment. We've got a lot of partners, a lot of the AI platforms that everybody will know their names. But, you know, that's what they're trying to do. They're trying to connect with us. We've got the MCP server, we've got the APIs. And, they're trying to do that on demand, more enrichment rather than the universe scanning discovery.
MW: Yeah.
PN: As it seems to be what they're doing at the moment. Anyway, that's where they're focused. So.
MW: Okay, I have time for we're stretching our time out a little bit here, but I'd like to get to one more thing. It's a really interesting question. With regard to how you think about data signals in the age of AI.
So, for example, if AI streamlines workflows for some roles, there are less FTE being added to, tracking things like headcount. It may not necessarily be necessary to understand revenue growth. So curious how you're thinking about things like FTE signals or other signals, in this evolving landscape. So this question again is, is it necessary to track all these things as models get smarter?
PN: It's funny because we're at revenue of all things and such a funny one for, I mean, in Europe it's easy, right? Like it's it's reported and generally reported. It's easy. It's no problem in North America. It's a pain in the butt. So I mean, like, does it change anything? Yeah. I mean, sure. But the reality is today, when you look at how you do a revenue estimation or model, you know, the, the you already put a lot of modifiers on it, right?
If I'm B2C, like, take what was the, was it what was it? WhatsApp or instant? I think it's WhatsApp, right? It was like 20 people, you know, hundreds of millions of users. And so you already kind of apply that, that model or that scaling, depending on the industry or the business. Those on the call that might be experts in, industrials, for example.
And you generally have a sense of, you know, how many, how many folks does it take in a machine shop to generate how much tax revenue, and why revenue, or what are their margins look like? That kind of comes with the territory of being knowledgeable about that sector or that space. And so, you know, I, I guess the idea is that, well, I don't need as many people to do the job.
Sure. Right. Like, yeah, totally. And so, how do you apply that? I mean, just for, you know, let's play with that example for a second. I look at industrials. So I can do a lot for industrials today. Probably not. Right, I don't think the AI and the machining have changed at all materially. But in the B2C world, where maybe I'm a I'm web app or something, do I need less engineers or less support people?
Yeah, definitely. And I think that will take some time to percolate through, you know, and as more gets known out there, models will get better. And, and generally speaking for again, probably North America not not EMEA, but the models are the, the I'd say the best the best data are generally those that are following a sector, another sector pretty well versus a generic model that says, you know, this is kind of how we do things.
Because again, you talk to a lot of folks, you kind of get a sense of the margins. You kind of get a sense of, size of teams and other things. And so those tend to be the analysts tend to have the best kind of sense of what a model is versus a generic platform. Why like model.
MP: All right, I want to add size, okay. The. Yeah. Again, I got a good opinion on this. I think this is why it's actually becoming more important on having to combine the premium kind of more standardized data, and I'm sure of of social good and self bringing more secret sauce to the unstructured data. So I was like an example the other day of, Loveable, a company that went to 17 million ARR with less than 30 people in less than a year, would that have been picked up in, you know, maybe for the low growth rates?
Would that have been picked up? So the question becomes, if headcount over time is going to become less of an indicator of growth. And for US companies, we don't have revenue. How do you find those businesses at all? I feel like firms will start specializing in certain types of data. Maybe we become very good at analyzing job descriptions that have been put online to understand the business strategy.
Maybe we become experts in understanding GitHub comments and Reddit posts or, you know, looking at social engagement, or indeed looking at news articles and other signals. But that's got to be combined with knowing the businesses. But the scores I spoke about, I think that's going to change to try and do much more on subtle signals and unstructured data.
MW: So new signals may take on more, more relevance in the future as these models start to put the dots to connect the, put the dots together in new and interesting ways. I think we've got to leave it there. We've run a little bit overappreciative. Almost everybody has stuck with us though, so really appreciate everybody sticking with it's been a great, great conversation.
Feels like one we need to follow up on soon. Prescott and Martin really appreciate your time. Prescott from like one in the morning from the Philippines, and Martin from the +8 British summertime zone. We're covering all the time zones. Really appreciate you guys joining the conversation today, and look forward to being able to do it again soon.
Thanks, everybody, for joining us.
PN: Thank you.
MP: Bye-bye.
