Al Cattell: Practical First Moves for Safe AI Adoption in Private Equity
Summary In this episode Steve Budd speaks with Al Cattell, founder of GiantKelp, about the practical applications of AI in private equity and business processes. They discuss the importance of using AI to automate workflows, the challenges of managing expectations, and the role of subject matter experts in successful AI deployment. Al shares insights from his experience in helping firms identify use cases for AI and the significance of starting small with controlled projects to build momentu...
Summary
In this episode Steve Budd speaks with Al Cattell, founder of GiantKelp, about the practical applications of AI in private equity and business processes. They discuss the importance of using AI to automate workflows, the challenges of managing expectations, and the role of subject matter experts in successful AI deployment. Al shares insights from his experience in helping firms identify use cases for AI and the significance of starting small with controlled projects to build momentum for future initiatives.
Takeaways
- AI is not a magic solution; it requires careful implementation
- Successful AI projects often start with internal-facing tools
- Managing expectations is crucial when introducing AI
- Subject matter experts play a vital role in AI success
- Small, controlled use cases can lead to significant improvements.
- Building momentum is essential for broader AI adoption.
- The landscape of AI is rapidly evolving, requiring agile strategies.
Chapters
00:00 Introduction to AI in Private Equity
01:14 The Birth of Giant Kelp and AI Exploration
03:58 Identifying Workflow Automation Opportunities
06:05 Successful AI Deployment Criteria
09:12 Building Minimum Viable AI Products
13:41 Expectation Management in AI Projects
16:01 The Role of Subject Matter Experts
16:09 Trust and Branding in AI-Driven Legal Work
19:50 Case Study: AI Implementation in a Law Firm
28:32 Advice for Scaling AI Strategies
Links
Al's LinkedIn
GiantKelp
AI Pathfinder Private Equity AI Strategy Meet-ups:
https://www.aipathfinder.co/ai-strategy-day
https://www.aipathfinder.co/breakfast-briefings
Have a question or would like to suggest improvements, please contact steve@aipathfinder.co
AI Pathfinder Private Equity AI Strategy Meet-ups:
https://www.aipathfinder.co/private-equity-ai-events
Have a question or would like to suggest improvements, please contact steve@aipathfinder.co
Hello and welcome to the AI Pathfinder for Private Equity podcast. This series is about helping PE firms move from AI curiosity to confident capability. And one way we do that is by speaking with founders who are building focused solutions at services to solve real operational challenges. Today we're tackling a common frustration. Many companies are talking about AI, but too few are actually putting it to work. My guest, Al Cattell, founder of Giant Kelp, a specialist consultancy that believes the only way to leverage AI is to use it. Every engagement they deliver ends with a working automation, not just recommendations. Al, welcome to the podcast. Lovely, Steve.
SPEAKER_00Very happy to be here. Lovely to be here.
SPEAKER_01Thank you. Great stuff. Well, let's start off, if you don't mind, a little bit of background on yourself and of course, uh Giant Kelp's story. What gap did you see and what made you launch the business?
SPEAKER_00So I set up Giant Kelp in late 2020, and the original plan, to be fair, with the business was around optimizing digital products. So increasing conversion rates, reducing churn, and that could be done through a whole range of ways through UX, data, you know, analyzing, engagement, that sort of thing. And at some point, not too long after the business set up, I actually got early access to GPT. So this was OpenAI's earlier model that you that they gave access to for people to play around with effectively. And I realized that you could use AI to write blog content. So to help promote the business, I was, I was, I started tinkering around with using generative AI to write content. And it was incredible. Like you could just write, write me a blog post about XYZ and it would write it. And yet at the same time, it just wasn't quite good enough. And so there was this, there was this moment where it was like, ah, there's something very interesting here, but it's it's actually when you actually read the output that it's produced, it just, you know, it didn't pass the sniff test of being good quality writing sort of thing. For fast forward six, seven, eight months, and ChatGPT got released. And at that point, I kind of my sphere was blowing up with lots of people talking about ChatGPT. And I'd sort of seen this coming because I'd been playing with a similar type tool earlier. And so I just I thought oh, I'd missed, I thought I'd missed my opportunity to be ahead of the curve. And so I down tools, pivoted, and just spent most of 2023 not earning any money and just learning about what's possible with AI. That original experience I had, I think a lot of people have. It's like this amazing moment where they see it's relevant, but actually it's just the quality isn't quite good enough. And what I realized was that to get that quality good enough, you probably need to go through a series of steps. So to write that blog post, you probably need some idea about the business, who their audience is, what their pain points are, and so on. So I built a workflow that would be a multi-step AI workflow that effectively would stitch all that together. And the idea of that is that you could get reliable, consistent, high-quality outputs without having to spend two hours yelling at Chat GPT to give you the right sort of content. And it made me realize that actually a lot of the value from AI as I started using it was very much on individuals. So, you know, it was all about individual efficiency and it wasn't really about business processes and systems. And so if you think about it, lots of people are talking about amazing prompts. Well, that's about making me as an individual really capable. But it doesn't really significantly change the way your business is operating. And so the opportunity isn't, I think there's a lot of value to individuals having access to AI and being able to use it. But actually, how do you rethink some of the processes that you run day to day? That's where the real value will come over the next few years. So for Giant Kelp, I effectively set up as a agency, as a consultancy to help automate internal processes with AI as part of the steps along that journey.
SPEAKER_01Are you getting people coming to you and saying there's a particular workflow that I'd like to uh automate, or or are you having to go in and sort of um sort of handhold them and and pimp point towards the sort of workflows that would be most most uh I don't know, most uh likely to work with AI?
SPEAKER_00Yeah, I think the in all honesty, most people have a good idea about where they want to try and use this technology. You know, it might be there's a property broker we're working with who every time someone asks him to help him source a property, today he has to go through all of his emails and dig out all those things and remember stuff and that didn't and getting the information out of the emails into a searchable database is a perfect use case for AI because we can just funnel the the emails through AI, extract data, put it in a database, and he gets a little web GUI, a little website he can go and use so it massively makes his life a lot easier. But he very much already knew that was a pain point in his business. And and I you know I've spoken with other people who have done sessions. Now it's been, you know, AI has been a thing for a couple of years. Businesses are starting to to start to try and work out how to adopt it. Um, but most businesses have got a fairly good idea for the first couple of use cases, I would say. Um you're you're ultimately for areas which are gonna save some time. Typically, people are focused on efficiency savings or that sort of thing. But yeah, I would say you could you could spend a lot of time thinking about what's possible, but at the end of the day, people have a fairly good idea about what the first kind of couple of use cases are.
SPEAKER_01What what does a successful first AI deployment look like? Because you know, we were both at an event I hosted last week. There's there's a there's more than just the technology here, isn't there, to to try and move from uh I don't know, a pilot, um, a concept to actually it's it flowing and and working as a as a you know changing a workflow that could have been there for decades.
SPEAKER_00Yeah, I think so. What does successful look like? I think I think the most successful pilots probably have a couple of criteria. They're probably internal facing. So I think certain if you work in retail or um, you know, customer services is a use case that uh that is highly spoken of. But I think for the purposes of the private equity market, their portfolio companies, I think the the most successful sort of initial use cases will probably be internal facing tools, things that your own team use to inform some of their activity, be that analysis, research, day-to-day administration stuff. And because of that, you should be able to, from a um return point of view, you should be able to quantify it from a time-saved or money-saved point of view. So what I I I met a company not too long ago, a sort of a large, let's call them a media company, and they want to spend two years and multi-millions of pounds building an AI platform. And in some ways that's probably the right thing, but I think it'll be very hard for that to be a success. I think it'll be very challenging because they're almost trying to do too much in a world which is moving very quickly. So, you know, an internal project where we can say, right, we think we can save three hours per report and we do, you know, 20 reports a month, and that's a very narrowly defined use case, but it it is, it's more likely to be successful. And then I think to the point you were alluding to, I think part of the challenge that that comes from these is actually getting people to use the technology. So I met a membership organization. What and when I met them, their AI strategy was, and this was written down almost, was let's just give everyone co-pilot. Their AI strategy was give everyone copilot, which I'm sure Microsoft thought was a fantastic idea. But it wasn't really an AI strategy, and it almost certainly was just going to end up being a big sunk cost with very little success. And so if you focus, if you're thinking about, okay, within your firm or your portcos, what is a good well, how can we how can we what is a successful use case looks like? It's probably quite a narrow, narrow test case. And it's one with a small group of people can actually use it and see that it's delivering value. There's a wider sort of change management thing, which again I think comes down to using small, narrowly defined use cases to build momentum and that sort of thing is very um very helpful. But yeah, I think that having that sort of that ability to to not try and do too much, choose something simple and make it work, that will be successful and that will engender and breed success for future sort of rollout. There's been a lot of learning over the last, you know, since 2000 about the best way to approach digital projects. And when you're not a digital first company, that can be quite a difficult thing to do. But I think what we're seeing with AI is we're kind of seeing the digitization of services and knowledge work and that sort of thing. And so bringing the same sort of mindset, which is build something that's an MVP, like a minimum viable product, a minimum viable AI product, and then iterate off the top of that is the way to approach it. And, you know, a lot of the a lot of our focus is very much on get something into someone's into people's hands in a month, let them use it, and then let's work out how we need to improve the outputs or maybe the inputs or the UI potentially or whatever it might be. But the key thing is to get something in front of people's hands to use, because it's all well and good writing a PowerPoint that says we think we could do an AI project which would help in these areas, until you actually start using it, you don't know how good it's going to be and how effective it's gonna be.
SPEAKER_01Yeah, I I I yeah, I I agree. And and I think that there's some setting of expectations with those people as well, isn't there? That um, you know, you said at the start how how you felt that there's this amazing moment. And I, you know, I was the same, that amazing moment when you're trying something and you you're just like, you know, sit back and go, Well, that's it can do that. But yeah, if not, then it's only so far there, but you're impressed by it. Some people now going into it, they're expecting magic to happen. And as you say, getting something in the hands of people really quickly is important.
SPEAKER_00Yeah, yeah, yeah. And the reason it's important to get things in people's hands is like the AI isn't the answer. I think that so there's there's a couple of things, right? People, there's a lot of hype. There's a lot of hype that AI so when people log into LinkedIn or whatever they get sort of social or news from, everyone's talking about the latest AI, this, that, and the other, and there's a lot of people jumping on that bandwagon. And so it could be quite hard to understand what's real and what's hype. And then I think there is a misunderstanding that you can just chuck stuff at AI and it just magically works. So there's there's hype, there's misunderstanding. And then there's the thing of like, it's both really impressive and not quite good enough. So that that um that experience that we've had is true for everyone. I think they they both are like, I could use AI to do X, and then when they start using it, they end up spending two hours changing the output because it wasn't actually perfect. Or they're lazy and then they don't get a good output. And so, and so the the need to manage expectations is easier if you're able to take people who are subject matter experts and who know what a good output looks like and put it in their hands and say, right, so ultimately your first AI sort of projects should be validation projects. If you frame it, the expectation management is we believe AI could help us do this. And we're gonna we're gonna run a four-week pilot or an eight-week pilot, and we're gonna validate can AI actually do it. And the outcome of that then is not like it's a it's a learning experience for the business. It's a it's an ability to um to get closer with AI, but do it in a way where you're not promising the world, you what you're promising is is that you're going to to to test this new technology. And it means that if you're going to validate if it's good enough, you need subject matter experts to um to be engaged in the process. And I think if there's one area which I've seen consistently, it's that people already have day jobs. And you often need people to come in and can contribute their either their knowledge into the way the system is built or their evaluation of the outcomes of the outputs that it's producing. And that's that's probably quite important. Well, it is important. But yeah, so in that expectation management, I think it's it's about saying we're gonna we're gonna use AI to produce these outputs and we're gonna test if they're good enough. And I'm pretty sure most companies, the first step is to produce V1, and then you've got another period of X weeks to refine and iterate. But if if that's the journey that companies go on, I think by the end of month two or three, you will have an extremely good system. But you have to kind of the expect the the journey you're taking people on isn't this magic black box that you're just gonna chuck stuff into. It involves subject matter experts, it involves a test or build and iterate kind of approach. Um, but if you do that, then I think you will you will get the results. I I I firmly believe that.
SPEAKER_01Um not to labour too much on this, but just a final point, I suppose, on the individual, let's say that that so those subject matter experts you talk about. You know, I and there was uh someone that works uh uh in legal that was talking to me about how they were using Harvey, and because they've been in legal for 25 years, they they could very quickly understand you know be able to qualify the output and and know that it's done something very well, but for others that wouldn't be necessarily the case, or certainly in these early stages, it's really important to have that. What is their what's their incentive then?
SPEAKER_00Yeah, right. So I think that's a very good question. So I think the the AI isn't AI isn't gonna replace their job. It's probably, it should, it's likely if you're if you could be if your job is fully replaceable by AI, then there's that's probably not the best job in the world, you know. So I think if you're an experienced legal professional and your your ability to tell the AI what it is that you're looking to get out of it and sort of train it and evaluate it, is is extremely powerful and is IP. Now, whether you choose to give that to Harvey or Harvey's an amazing company, but if and an external provider to train their models, or you build your own system which channels that IP within your business and basically becomes the um the value add, I think regardless, someone's going to do it, right? And so I think the motivations for the very experienced legal professionals to do that, you can either look at it as an opportunity for your business to leverage the knowledge you have to be more efficient, more profitable, open doors to do different types of work, lower costs, so you could do more work for different people. But to do that, you have to kind of take knowledge that's in your head and put it into a more systemised approach. But it won't 100% replace you. So I think if people are very defensive about that and they've been in the industry for 25 years, they probably have got another 10 years of career ahead of them, which it doesn't matter massively. But along the way, other people are going to absolutely, you know give themselves the chance to spend more time doing more client work, like I don't know, high-value engagement stuff, rather than the in the in the weeds legal work sort of thing.
SPEAKER_01I think that, you know, but the incentive potentially, you know, certainly for the two of us and many people out there is to breaking processes that you've been working with for years and years and years, is really refreshing and energizing. You know, I don't think I was on the phone earlier to to someone thinking, I don't think I've ever been so excited in my career before. Because I'm doing things in every aspect, it's new, and that's what that's what I'm excited about. And I don't think for one moment I've sat there and gone, um, worried about my particular sort of ability to uh do the job I'm doing. I I suppose I'm rewriting the rules a little bit and and trying to find what that new role is. But so that you know that's easy for me to say in a one-person company rather than in a big business. But I think that at the same time, though, looking at a process you've done for so many years and thinking how you could disrupt that and become more competitive, differentiated as an individual, as well as adding to the value of the firm as well, I think's really interesting.
SPEAKER_00Massively so. And I think that's the thing where for you know, people who run one-person businesses or smaller businesses tend to be a bit more entrepreneurial and I think are uh are potentially more willing to just have a play around for want of a better expression. Like experiment with AI and and and it is good. When you're when you're within a much larger company, you have less time because you have more internal meetings, you also have less freedom to experiment, you kind of you have more worry about getting things wrong. And so I think there is that aspect where I think today if you if you have a portco where you you think there's AI opportunity within it, there are probably a couple of people within that company who also believe that they could do amazing things. And it's almost like how do you channel that energy and how do you enable them to, as you said, like reimagine this process or or or think about things in a different way. Almost all of the conversations I have are with business owners, business leaders, C-suites, rather than people at the coalface, because if the C and the reason for that is if the C suite, if the C-suite don't get AI as an opportunity, no one else has got a chance at all of doing anything. And so it's almost like the the most engaged companies are the ones where the CTO, the CFO, the COO, they know that there's something here and they want to kind of do it. And then they'll find a couple of people within the business who are also like, oh, wouldn't it be great if this, or who have gone off and done something on their own time in Chat GPT or co-pilot or whatever. Um yeah, so I think that's right. I think the it is an incredibly exciting time. Like it is really, really, really exciting. I absolutely, I absolutely love the opportunities that are happening at the moment. And I think people To the earlier point about the fact that we try and build something is like you can only learn to swim by getting in the water. And I think the only way to really adopt AI on any size or scale in any business is to start using it. And that will require like a plainful, experimental, test and learn kind of approach.
SPEAKER_01Yeah, yeah, absolutely. Um, I I wonder if you could talk through one of those um one of those firms you talked about that that had that sort of curiosity and in and intent from from the leadership that that um you went on and then worked with them to deliver an outcome. Is there something you could and I'm not asking you to name names, but is there something you could talk through there about you know what the process and and the outcome?
SPEAKER_00One of them, I think, um, given that we've been talking about lawyers and your audience being in the PE space, was a PE backed law firm uh who had was again, I think they realized that they needed to try and demonstrate that they had an AI, that they were adopting AI. So a lot of the motivation is is about can we can we improve this this process or or whatever it might be? But actually what they also wanted to just demonstrate that they they had moved the needle somewhere on AI. And so the place we explored, we explored one of the parts of their business, which involved contacting other law firms to do work internationally. So the of the 400 people in the firm, 50% of them are admin staff who just basically help information move around the business. They don't, they're not fee-earning staff, they don't, they don't generate revenue sort of thing. But when it came to, okay, we need to work with this law firm in New Zealand, the lawyer would email that their companion their colleague, a counterpart, and say, hey, we need a can you help us with this? They would then wait a few days. If they didn't reply, they had to remember to chase up. They then had to sort of so everything was very reactionary and very ad hoc. And so, but it was just generally to get a quote, right? We needed a quote so that we could give that back to the client and the client massive US multinational. And so what we ended up agreeing on was that wouldn't it be great if we could have a chat bot, we could just say, hey, I need a quote for this type of service in this region, and we could go and look at the data and we could just produce that back. So it's almost like a little very custom version of a chat GPT type interface where you've just got natural language and you say, Hey, I need a quote for this. And so that was a relatively easy thing to build. What was missing in that was there was no data because none of these, none of these emails that the lawyers had been sending over the years had ever been shared with each other, let alone put in a central database, and the back end IT infrastructure was, as people who work in the legal sector will know, relatively archaic. So we built the first version of that chatbot with synthetic data. We we effectively produced a ton of Facebook. Pricing information for made-up law firms that we could use to demonstrate the prototype to people. So again, about getting something into people's hands very quickly, what I think you will recognize, even probably the audience will recognize, is like it's all great until someone asks for some data, and then you hit the internal IT blockers and this and that and the other. And so what we did to circumvent that, A, we were, we were again, we were building an internal tool. It was relatively low risk. We weren't dealing with any client information. And we were able to produce synthetic data to fake so you could you could use the chatbot, we could get feedback, we could iterate, we could test. That sort of thing historically might have been possible, but it was very laborious. It's incredibly, incredibly accessible to do that today. So I think there's like an ability to instead of imagine what's possible, you can actually build a working version and just plug in some synthetic data and let people understand and and and and use what the tool is. And then to support that, we then built an email worker. So we then effectively plugged into an inbox and we emailed 559 companies, law firms around the world, and we said, Hey, it's this firm, we're updating our pricing. Can you let us know? And then we had AI monitor that inbox. And so whenever anybody replied to that inbox, the AI would pass it to check what the email said, was there pricing in there, extract the pricing, put it into a database. And we do that in sort of tranches, checked it worked, and then so again, we kind of were able to AI to effectively manage comms, extract content from emails, and then put it into the database. If the AI ever got confused, it would just flag it for a human to review. And we had some edge cases. Some people, the person had left the firm, some they sent us pricing for way more than we asked for, they sent us for multiple regions. And so there were some edge cases, but now they have a system where every six months they can an automated background worker will email all of their contacts, ask them for pricing updates, put them into the system, and we can add all their internal, like their markup, their currency transformation, you know, currency forex. And so now they're the the value to their firm is instead of it taking five to ten days to get back to a client, they can get back to their client in one or two hours. You know? There's an there's an audit trail for where this pricing came from because we've got the emails, we've got the attachments, we've got it all searchable. It's incredibly close to the team. And the lawyers don't have to worry about like thinking about this. So they're freed up to actually think about rather than the admin work of chasing people for quotes, they can think about actual legal issues, you know, sort of thing. So that was like I think that was an area which Again, it speaks to that thing of like there's a lot of internal admin type work, which is the low-hanging fruit, that is potentially actually quite valuable to improve or to either lower costs or to improve the the the sort of the client experience to to their customers.
SPEAKER_01How did you um yeah, so how was that then identified?
SPEAKER_00Because uh there could be so many other um So we had a we did have a workshop with um the team was uh spread across the UK and so we had an online session for uh about two hours, but it was identified that this particular team were the ones who were most open to exploring AI. So we kind of honed in on one section of the firm and said, right, this is the team. We had uh like a senior, a couple of senior fee earners, a couple of senior partners, a couple of people who were AI interested, and maybe a couple of sort of paralegals. So there was about half a dozen people on the call, and we just talked about different options, and we probably created a list of I know the short list was three, and I can't remember off the top of my head what the other two were. There was a long list, which was probably only six, seven potential areas. And through that call, we kind of were able to just discuss complexity and impact and you know appropriateness, right? And so there were a couple of use cases where it was like that will rely a bit on internal systems that that are currently being overhauled. So, but we we went through that process to to get to it.
SPEAKER_01And I assume from that point, you know, that that proof of value then means right, we we move on to the next or something that's that's I don't know, more scalable potentially.
SPEAKER_00Yeah, exactly. So then it's like so what you have at the end of that is you have like a tangible use case that you can show to the business. So so then we built so that uh system got extended a bit where we built a system part of it was it could go to the World Intellectual Property Office and fill out a form on behalf of the lawyer and then produce the results back. And there was a uh like uh who's the person I should talk to in this region, so we're able to sort of give them another way to find people for other uses. Um the but ultimately what they're able to do with that is to then really build momentum with the rest of the business to think about, oh, this is actually quite interesting, right? They had had a firm in a a consultancy and who they'd spent X six figures with on a on a document that was their strategy document. And whilst it was probably quite useful, it didn't really move the needle. It it basically just told them stuff they probably already knew or opened more questions. Whereas we were in month one, we built a prototype of the chatbot. By month three, we had the email workers sending out their data flowing in. By month four, we had the system ready to go. And that really enabled them to, as you say, start thinking about other use cases and rolling it out. And and from my perspective, I'm obviously quite keen to help with that, but they actually took on a bunch of they once they realized that they actually they were directed a couple of internal resources to things to just work on because they realized that they could build some of this stuff relatively easily and some of the things so that it basically changed a lot of the focus of some of the internal team to to really start investing time and effort on it. So, yeah, which I think was quite quite compelling.
SPEAKER_01Yeah, yeah, yeah. Look, um, half an hour is sped by. Really, really fascinating conversation. One final question. If we take the the round table that you were on at the AI strategy day and you had sort of five, six, seven people on that table with different maturities, hearing what they're they were saying, and and the roundtables around trying to move from sort of those pilots to something more more scalable. What sort of one piece of advice would would you have given them at the end of that sort of session um as they move into you know 2026 and sort of building upon their their AI strategy, whether that's been written down or not?
SPEAKER_00I yeah, I think you know it I think if you it if you want to learn how AI can impact your business, you you have to use it. That's that's the key takeaway. And then it's and then it's about how can we use it in a way that's controlled, relatively low risk, relatively low cost, to be fair. And I so I think start small with a very controlled use case. You don't have to know all the answers about all your data. You don't have to know whether you should be using Copilot or Chat GPT or Claude. You just kind of need to identify a small, very straightforward place where you think AI could help and just test, test, can it do that? And that ultimately is the first step to unlocking what comes next off the back of that. Because in every business there'll be different levels of um availability to technical talent, to, you know, just your infrastructure and how your business manages information. But ultimately, you can you can you can spend a a small focused amount of time to to start building the muscles up to actually what that that your business you know needs to use, I think is the is the way to go. I don't think there's a magic bullet. And I think what makes this time exciting is there are no real rules because we're trying to work it out. Everyone is trying to work it out, and everything moves very quickly. And so that's why I would lean towards small focused action rather than sitting back and writing a massive governance plan and you know, all because that's going to be out of date by the time you've you've finished writing it. So that that's how I that's what I would say is the way to go. And that's could be applicable. And then you just dial it to how to what's the resources you might have internally or if you want to work with external providers. Yeah, sound sound advice.
SPEAKER_01Al, thank you so much for speaking to me uh today. Thank you, Steve. It's been a pleasure. Um for for the listeners, I'll put in the notes a link to your LinkedIn now and also your your website, obviously. But uh yeah, it just leaves me to thank Thank you once again and for those that are listening. We'll be uh back very, very soon. And as some of you may know, I hold uh events in London and now soon in Manchester. So please check out the website for for those events uh for private equity. Okay, all right, see you next time. Cheers, bye bye.