Building an Agentic AI C-Suite for Fintech & SMEs

Doing well, and thanks for having me. And this is me, it's not a replica. I am here in person with you, so there you have it.

It would be more made up if it was the virtual one. More hair, a bit more muscular, maybe.

Yeah, thank you. Maybe if I start off with the genesis of Beyla, it gives more of the context. So in folklore, or mythology, Beyla was the goddess of agriculture, fertility and pollination, the goddess of bees and the hive. And what we set out to build was a digital hive of collective intelligence. So rather than collecting pollen from nature and growing that, we wanted to collect data that was relevant for the world we live in today. So Beyla is a digital hive of that intelligence. And what we're using it for is business, business and finance is effectively the area we believe remains ripe for transformation. Beyla is setting out to build a new category that is not only a business platform, not only a finance platform, but merges the two into a proactive technology that works for you, the customer.

Yeah. So Beyla has set out, and this will improve as technology improves and as the data improves; if we're more confident about the data, then Beyla will be able to act, and it's been built to act for you, but under your guidance, your approvals, your authority. So Beyla has an autonomous element to it, but it works for you, the business. We're setting up Beyla as a B2B platform, so we're not looking at consumers at this stage. This is small businesses, primarily, the people that have a huge opportunity and huge aspirations, but don't necessarily have access to the resources of a Fortune 500 or a FTSE 100 company, where you have that internal talent and third parties coming in to speak to you. Because in reality, if you're a small business, you don't have the time, you don't have the connections, and often you don't have the budget to pay for that.

Yeah, and focus on the things that add value to you. Typically that's your customers, growing your business. Part of the challenge with technology, certainly over the last few decades, is that it's been presented in a way where you drive it yourself. So I have to log on to the system, I have to work out what it is, I have to figure that out. And if I need somebody to do it for me, I have to pay for their time, full-time or fractional, and then I have to tell them what to do. So the challenge with technology systems is that yes, you're giving me information I can use at my fingertips, but it only works if I know how to use it. And the reality is, unless you're an accountant, things like Xero, QuickBooks and Sage are quite complex. Unless you're a banker, the banking systems and treasury systems look quite complex. So it's around how, if you're starting out or you've been in business for a number of years but you have an idea and you want to grow, you have to have all those caps on. And that's a hard thing to manage for many businesses, so you do the bare minimum. We want Beyla to change that. We want Beyla to be there as your ally and your advocate, seeing how we can support you with the tools that, if you were in a Fortune 500 company, people would be queuing up to give you.

It's how we've evolved. We, as in the human race, have evolved around working with other humans, and we've become successful because we collaborate. We build teams, tribes, businesses, cultures, and that's with other humans. We never evolved to type into a small box, or press buttons and type in "I want this action" with text. And it's interesting, in terms of how we all message each other on a daily basis, WhatsApp, SMS, messaging happens all the time. Often you find that actually something's been miscommunicated: I've sent you a message, you've not understood what my context was, you've received it and acted on it, because you're limited to those parameters. We wanted to build something new. If I thought about what I'd do if I had a literally clean, blue-sky piece of paper, it was around how I could take the digital experience, or the human experience, but digitalise it, based on where I believe the focus of AI would get us to. And that, for me, was digital humans. So how can we have something you can work with like a team member? If I was to hire a member of my team, they'd come in, sit in my office or in my working environment, and I'd speak to them. I'd have a Teams call with them. I'd have a face-to-face meeting with them. We'd say a lot more than the one instruction I wanted them to undertake, and through that context they're learning, they're providing me feedback. It's around how our brains work, in a way, to really understand complex tasks. So, for digital humans, talking is great, so having a voice, but also having the other sense: if Beyla's talking at me and I'm the customer, and I've got to look at their face when I don't know what they're talking about, that's body language, and it actually gives useful information.

Yeah, and it's not easy, I'll be honest. And we're not quite there today. It's not perfect, in the sense of it being a real person, it's not. But we're building this for the next 10 years. We're not building this for today and tomorrow, and things will iterate.

But what we want to build, and what we're trying to build, is around it feeling positive. It feels like actually this is a team that's got my back and is there to help me. That's really the key thing we're looking to drive out. Something that is not working for us and trying to sell to you, but that we give to you as part of an ecosystem where you grow, you thrive, and through that we grow and we thrive as a community. And that, to me, is all about the way humans work in the best way together.

Yeah, certainly, and this is an iterating process. As I mentioned, we want Beyla to be a positive. It's not there to tell you everything's right and perfect. And it's interesting with AI, if you're using AI platforms, they'll have a different persona. So for us, building out that persona is important, and it's got to be Beyla's persona, our corporate brand, if you like. That's what we're working on continually. In terms of what that means, obviously data is key, consent is key, and the ability to look at where you are, because the reality is that today, Beyla is primarily used over a PC or a laptop. In the future, Beyla will be on a mobile. So if you're sitting on a train, Beyla needs to recognise that and not tell you all your financial details out loud in the carriage, in the same way you would be careful if you were saying that as a person. So we're looking to build in those rails to help protect. Some customers will say, "actually, I want that information, I've got my headphones in, give me the info." We're not there to police a user preference. We're there to give options, to create a framework where the business can set their own rules and structures in the way it works for them. And that's key for us. We want to move into the regulated space, very much from day one. If I roll back, we actually started out to build not a business system, we started out to build the first AI-native neobank. That was phase one of Beyla. The challenge with building a challenger bank is that it's hard and complex, so we've pivoted slightly. That said, it's very much on the roadmap, we've been working with the regulators. We want to be in a position where we have that regulatory oversight, and we're working as if we are a financial services business, with all the requirements of Consumer Duty and everything else that goes around with that. Phase one of Beyla is a non-regulated agent platform, effectively, but we're building the two to link together seamlessly. That's important for us, because, as I mentioned, we want this to be a net positive for our customers, in a way that we can help them grow and support their aspirations. And by being able to control the movement of money, that gives Beyla an advantage in terms of adding that to their support.

Yeah. Mainly it's around timing and the ability to forecast. That's one aspect, it's not a simple thing, but unfortunately this is the sort of statistic coming back from people at the Federation of Small Businesses and various other institutions who've been looking at this and how they can put tools in to improve it. If you think about running a business, one of the challenges is that the smaller you are, the more terms are dictated to you. So if you're working with an enterprise customer, often they may say, "yes, we've got to pay you, but we're going to pay you on 60- or 90-day terms." That creates a cash flow gap for you, you've got to fund that gap in advance of getting paid. And then, okay, I'm expecting money to come in from another small business, and maybe they haven't paid me, how do I manage that? For many companies where they're running finance internally, or have a small team, they're not able to get on top of following up with vendors and suppliers: "are you due to make the payment? If not, please tell us so we can balance things out." Understanding, if we want to bring on a new person, what that cost is going to do to the business, it's not only the salary, it's the National Insurance, the pension, everything else. How do I factor that into my runway? How do I ensure that runway is consistent? How do I ensure my sales are keeping pace with my forecast? All of that happens, and for many people, it's happening in their heads. You sit down to look at it and it's, "this is next week, I can't make payroll, I can't make that payment to that supplier, they're not going to give me the inventory," and things break down. That's a really unfortunate position for many companies to be in. What Beyla does is work 24/7 to try and remove some of those issues and surface those things earlier. So, in the future, Beyla's working with that small business and it's saying to you, "we've looked at some of these receivables due, and the risk factor of receiving those is quite high. Should we look to reach out proactively and have a conversation to ensure they are coming in? Should we look at getting some short-term finance to cover that gap, and what type of finance should we get?" That's where we've been working, one of the programmes with the FCA, the Smart Data Accelerator, was around looking at what open finance, as a new paradigm, gives us. Open finance is effectively the umbrella above open banking, where it opens up APIs to finance companies, to lenders, to pension companies, insurance companies. So what that enables in the future is that if we find there's an opportunity, and Ian the small business requires funding, Beyla can go out to funders that we know have appetite, provide information, and bring back quotes directly to Ian the small business in a practical manner. Whether you choose to accept those is down to the business, and we can then help facilitate that. But it's trying to surface information in a way that really helps you drive direct action, rather than saying "I've done a search on the internet and here are some lenders, go and speak to them."

I think so. Now is the time to think about it, rather than "let's iterate and change that 10% that we've had and improve." AI gives us this huge opportunity of where can we go, it means we can do lots of things. And this will be a journey, in reality. It looks like open finance will be approved by 2028 now. Things can happen before that, but that would be under more partnership agreements with different lenders, in a similar way to what you see around insurance today. There are firms out there that can provide a concierge service for insurance. We can do similar things, but that's harder and takes more time and effort. If you have something like the regulatory framework saying all financial services companies need to have open APIs which are consistent and allow information to move freely in a positive way for consumers, then companies come into that space and build products which serve the consumers.

Yeah, for sure. One of the reasons we set out to build the AI-native neobank was very much around not taking the banking technology stack that was available. There are a number of providers that provide full-service banking stacks to companies that we could have branded up as Beyla and said, "we're now a new challenger in this space." But those technologies were built very pre-AI, ultimately. Typically they're 10 to 15 years old. They've been built around APIs, microservices, and very human-centred organisational structures. So what we didn't want to do was take what was out there today and try to push AI into it, which is what everyone else is doing, because the challenge is that the AI comes in and says, "all this information is in lots of different places, I can't understand it." It doesn't really have a voice, but in the sense of getting the data from a thousand different microservices and providing outputs on that, it is really difficult. So we knew that, actually, I don't want to take a banking system and put AI into it, why don't we build an AI system and put banking into it? Because we can then control that architecture, that framework, and say, "okay, if we have this aspiration, we want to build these products, we know what the knowledge and intelligence layer is, how do we now connect in the right industry capabilities to it?"

Some people I speak to today still think I'm a mad scientist. And that's the fun of it. I think we're proving that it is working. You asked why we pivoted slightly, one of the challenges is investment. And actually, I can monetise the agent services as a technology company far faster than I can as a regulated payments business. The reality of building a regulated payments business is that you have to build the platform, get the approvals, and then you can launch. That takes time, capital and talent. We can create a technology company, initiate that in a consistent framework with our customers, who understand we're growing with them, and do that in parallel.

So you'd think that. I worked in banking for 25 years, 16 years at HSBC and Bank of America, and I started out my career at Chase Manhattan, which is now part of JP Morgan, primarily in operations, servicing, customer engagement. I used to run parts of their cash management business, so I know at a deep level how those platforms work. And the challenge with trying to build AI into that, or partnering with those institutions, fantastic institutions, great leaders, lots of technology, is that they're not built to surface information the way that agents really require, and certainly not on an external basis. One of the challenges with Beyla connecting to a bank is that as soon as you go in via a gateway, those agents get blocked out from everything. And rightly so, the bank's guardrails and security protocols aren't going to let a third-party agent come in and sniff around. I know Anthropic is trying to get in there and do some other things, and that's an impressive opportunity, but they're in a very different state to where we are, and the reality is they are locked down in many different ways. So for us, I can connect to banks, and we will use open banking to get balance information and to send payment requests to a connected bank. But if I try to go deeper than that, all the information the bank has that they don't surface to the company via the API, I can't see. So by building the bank, and getting access to the information that, let's call it HSBC, just because of my previous employer, has, but can't actually see itself when making decisions, that gives us an advantage over fraud, over risk, and an advantage in servicing customers with the right product that fits their needs. If I go to a different organisation, that information is available within the organisation, but not in a way the organisation can consume it, check it, use it and refine it.

Yeah, so we've paused pushing forward on it. We know that is the pathway we wish to take, so we wanted to spend time with the FCA in advance of actually building anything, because they are the UK regulator we knew we wanted to be associated with. It was good timing, in the sense that they were launching the first AI Supercharged Sandbox, and Beyla was accepted into cohort one of the Supercharged Sandbox. That, for us, was really important, because we could then start to work with the regulator around the ideas and concepts we had, to test, "actually, if we want to do this type of thing, what do you think about it?" And they've been really open. In terms of what they don't have the answer to, they know they don't have the answers. We don't have a definitive pattern, and we work around what we think is right. For us, it comes back to the ethos of "how do we serve the customer in the best way?" The FCA had moved, over the last few years, to being an outcomes-based organisation, and that fits really well with our aspirations. We're not trying to do anything that's seen as breaking the boundaries or pushing the boundaries to support ourselves. It's around how we can help our customers grow, and if our customers grow, then off the back of that we support and grow with them.

Honestly, and I'm not saying this just as a position, nothing's been pushed back on, in the sense that, as I mentioned, I had my career in financial services. I know what we can do and what the requirements of a regulated firm are. So I was never setting out to build something I knew was going to be on the fence, "can we, can't we?" It was very much, "I know what Consumer Duty is, I know how we can support needs-based selling, I know that all of those factors need to be part of what our initial framework is going to be." So no, they haven't pushed back on that specifically. That said, we aren't formally through the process, so I'm not counting my chickens before they hatch. We have to put the time, resourcing and results around guidance and documents, as well as people, to back up any chance of securing that approval. So that is the future, and that's what we're looking at raising money to support and build around, so that we work hand in glove, in a positive way, with the regulator and all forms of government around AI. It's interesting you mentioned the fear of AI, because there is a lot of misunderstanding in the industry, which isn't helped by certain labs putting out information that I would challenge, in the sense of how autonomous, how "living" the systems are. Today, we remain a technology software-services-based business. The software is incredibly complicated, incredibly rich, and it works around the prompt you give to it. It doesn't have its own ability to go, "I'm going to think, I'm going to do something else," that hasn't been part of the parameters. One of the things that's quite topical at the moment is looking at hacking into different systems, it's because somebody asked it to try its hardest to hack into a different system, and then we're surprised when it succeeds in doing so. So guardrails are a really important area for the future. And the reality is that, unfortunately, that's not somewhere Beyla can exactly influence, we're not OpenAI, we're not Anthropic, we're not Google DeepMind. That's an area where those labs need to take responsibility and come out, so that we're doing things in a positive way, being open about what we're building and what's available, and not saying "AI is taking everybody's jobs, you should be fearful of it," making most people in society see it as a conflict to humanity.

Well, indeed, we could spend all day on this, and have the pros and cons of how the labs work. Part of the challenge is the monetisation of it. And as a small UK business trying to build something new, getting capital is really tough in the UK. So that's a big area around how the government, and the broader ecosystem of financial services, can help. We've remained one of the world's centres of finance for the longest time; if the right backing isn't put in place, other markets will take that position away from us.

I think it's something we have to be cognisant about, and something we need to take into the framework. One of the key things around open finance is that it needs to be consent-based, and it needs to be living. What I mean by that is, one of the challenges, if you're a small business: I go to a bank and say "I want some sort of facility," and they say "give me the last three years of audited accounts." The last three years of audited accounts don't necessarily tell you, the lender, what the next three years will be. Those audited accounts are already probably at least six to 12 months old, so you're actually looking four years back. And you look at data which has gone through a process, it's not real cash flow data. What I believe is far more valuable, if I'm assessing a business, doing due diligence around a business, is what's happening now: who are the management team, what are the aspirations around where they're going, what's the customer base looking like? All those factors, if you were investing in a business, that's the way perhaps we should look at lending. This is an investment into the company to help them grow, not looking at what happened over the last four years, because, as we've seen, that isn't always a perfect indicator of future success.

Yeah. My background is finance, so we started out in finance, and that's where I think consistently, the movement of money is fundamental to every business. So for me, the quickest wins are around making sure that accounting, bookkeeping, forecasting and cash flow processes are as automated and efficient as possible. That's where the big wins directly hit, day one. And, as you mentioned when we were speaking around time management and administration, it's around what's happening in my day, what's happening in my day tomorrow. So building into Beyla, and this is coming in a release in a couple of weeks now, so by the end of August, is a daily calendar. Before you go to bed, or as you wake up in the morning, depending on your preferences: this is what's happening in your day, these are the meetings you have, these are the three important things you need to look at. And part of the memory platform is around trying to, as you tweak it, self-learn for you. So that allows you to think about it becoming a bit like a personal assistant. We're connecting into things such as Outlook and Gmail. I don't want it to necessarily be writing emails for you, I want it to be pulling out information from that and saying, "there's an opportunity here; these things aren't necessary, that's a bit of spam, you don't have to spend time on that." That, to me, as somebody who receives about 200 emails a day, plus LinkedIn and messages and everything else, is a huge time-suck. And actually, when I'm trying to figure out things, they get lost because too much comes in. So having that ability to focus gives me the time to spend on the important things. What we then move into is where we can add value around guidance. So, for example, if I receive a contract, is that contract fair and applicable to us? It's not giving you formal legal advice, but it can pull out some red flags you may want to look at. Is it under UK law, the Laws of England and Wales? Is it a US contract, because you're working with a US supplier? Is there an opportunity to go back and renegotiate some of these terms? That gives the information back to the business owner, so you can say, "maybe there are five things here, they're big, we're small, we're going to look at one of them." And that, to me, is a win. That's where we want to build to, it's not going to run your business for you fully autonomously, not yet anyway (maybe in 10 years' time it's a different story), but it's around how we can do those small wins that show incremental growth, and you trust the platform is working in your interest.

Yeah. So we wanted to build, from day one, a platform that was standalone for you, the business. Certainly all of those platforms you mentioned, Sage are looking at that, Xero are looking at that, HubSpot's looking at that, they all have their own platforms. That now means that, as a business, I have to go and speak to 10 different AI platforms from 10 different businesses, or more, depending on my circumstances, and figure out how they work. Typically they're chatbots, so I have to type a prompt into it. Maybe I can speak to it and say, "hey, tell me what's happening in Xero today," and it will tell me what's happening in Xero today. It won't tell me what's happening outside of that system without me having to rebuild all those capabilities around it.

Yeah, exactly. I see it as an ecosystem. I didn't set out to replace Xero, I didn't set out to replace HubSpot, I didn't set out to replace email. It's around wanting to be that ecosystem of connected opportunity, which surfaces insights, is trusted, and uses those insights in a way for good, to help business grow. And that is really important. People will say, "I can build systems, I don't need to pay for systems anymore, SaaS is dead." I would disagree, because the reality of building your own system, if you're an IT firm with 20 developers, maybe you can. If you're a small business trying to build in a non-IT area, vibe-coding platforms that are sustainable and reliable, good luck, is what I can say.

So, I think it's yet to be seen, in reality. Because how would that work today if I was, let's say, an advisory firm, maybe I'm PricewaterhouseCoopers, or KPMG, or Deloitte, and we have the same conversations? I'm pretty sure their terms and conditions would actually not make them fully liable for that. You're providing guidance and support around a project. We take it seriously, I'm not going to say "tough luck, you made the decision, ha ha," because that's not going to support our business.

Exactly. So we've got to take it seriously. Our agents need to have factual information. They need to, and you live and die by your reputation. Bankers love to say "people trust us." I don't trust my bank. They're a profit-making organisation that's paying the shareholders as much money as they can, and that's their business, they need to be very clear about that. That's how the business works; it's not a charitable organisation. I like working with banks and I understand the business, but "trust" is often an overused word. I want the positioning where the information we surface is trusted information, because we put the right tools, platforms, procedures and checks into it. I can't take the responsibility away from a person that uses that information and makes a decision that I'm not part of. And I think it will evolve. As we move forward, there will be people looking at different, whether you call them regulations or guidelines, around what that really means. I don't think we're there yet in terms of what that will look like. And I think it does come down to the labs as well. If we're using information, one is "this is factual information that is in Xero, for example, these are your positions," that is a statement of fact. If we're then putting best practice around that, that's a procedural understanding of how businesses work. If we look at other things, and this is where maybe the intelligence comes in and says, "you can do something no one's ever done before," that is a future opportunity. And there are going to be a parameter of different things that I'm sure will come up around this. But our primary guidance is around: we're surfacing information to you, we want to consult with you, and if you're not sure about anything, don't make a decision. Ask Beyla to repeat it in a way until you're comfortable with it. If you're not comfortable, don't make the decision.

Let's look at today, over the last decade: how many billions of pounds or dollars have banks paid in fines because humans made errors that affected other humans? That's a fact. We can't hold, if we say humans are 100% infallible and AI has to be the same, maybe you have the argument. Unfortunately, humans are fallible. We make wrong choices. Humans have human errors, and that's what happens. So I think actually having technology that's able to get in front of some of these things, and by nature of it being different, do a different approach, gives us a real advantage to try and minimise some of that. Not taking away the humanity from us, we are still the people driving the businesses, we are still the ones driving the technology, but this is helping us.

I know we're running over, so apologies, your audience will be dropping off.

Yeah, and it will iterate over time, certainly. And we are dependent currently on some of the partners we work with. Our platform is model-agnostic, and we have a cost of running and getting information from models that we need to cover. So the way we think is best to start out is a credit-based system. Depending on what you believe your usage is, you'd have a plan. Plan A, let's say, would come with X credits; Plan B would come with more; Plan C would come with more again. And it's around having the ability to show you what you're using. So we're not saying "this is a fixed fee, you pay it upfront and you're locked in." If you don't like Beyla, don't use it, and you won't be charged for it, and you can cancel any time. Your data is not stored by us, or kept, outside of the things we'd need to do from a regulatory-oversight point of view if we were regulated. So for us, we're not trying to harvest that information. What we're trying to do is provide the tools that allow you to get inputs, and we'd put a charge at that level. So you have a plan and you utilise that credit, and we show you how many you're using. If you're on the wrong plan and, at the end of the month, you've actually only used half of your credits, we'll put you on the lower plan, so it more aligns with your business, we're not impacting your cash flow, we're not stopping you growing. And if you come halfway through the month and you've used all your credits up, maybe you need the next plan, and actually that's a good thing, because what we're hopefully showing is that the cost and value of Beyla to you is far cheaper than going out and replacing our value with third-party systems and technologies that you're having to run yourself.

So we're currently working through it with some pre-selected customers. We do have an open waitlist. The way we're looking to roll Beyla out is to have customers come on board and make sure that we've got everything around the service we offer. So the best way is to look at the website, beyla.ai, sign up to the waitlist, and we're happy to contact you. As part of that, we give you the opportunity to say what you're interested in. That also allows us to say, "we've got those capabilities live now, come and try it out, get into the beta," and as we come out of that beta, we'll look to formally launch the platform and grow the future of finance from here.

My pleasure, Ian. Thank you very much. And we look forward to, maybe in the future, Beyla coming on herself and joining me in the conversation.