Rachel Curtis explains why banks can't build their own AI affordability tools and how Inicio AI achieves 100x ROI
Summary
Rachel Curtis, CEO of Inicio AI, argues that while building AI demos is easy, deploying them in regulated financial environments requires massive datasets and regulatory approval that incumbents struggle to replicate. She highlights that her platform, featuring the 'Budgie' virtual agent, achieves 90-95% form completion rates compared to 50-60% for manual methods, with clients reporting up to 100x ROI. Curtis notes that Inicio AI is the first and only FCA-authorised provider for this specific service, having navigated the FCA innovation pathway. The company is currently targeting Commonwealth markets like Australia and Canada for expansion, avoiding the US due to different collection cultures. Curtis advises founders to embrace non-linear growth through networking and emphasizes the importance of data privacy and strict AI guardrails.
Episode sections
- 0:07 – 2:06
Intro and Background
Host Tom Britton introduces Rachel Curtis, CEO of Inicio AI, highlighting the difficulty of deploying AI in regulated environments compared to building demos. Curtis discusses her 30-year career in financial services, starting at Halifax Building Society and ending as Chief Experience Officer at a challenger bank.
- 2:06 – 4:31
Origin Story: The Pregnant Woman
Curtis recounts meeting a pregnant woman in debt who was too embarrassed to speak to a human agent about completing an affordability form. This incident, combined with emerging AI capabilities, inspired Curtis to build a virtual agent that could assist consumers without human interaction.
- 4:31 – 7:04
Product Overview: Budgie and Affordability Assessments
Inicio AI offers a B2B SaaS platform featuring 'Budgie,' a virtual agent that guides consumers through complex income and expenditure forms. The system ensures accurate data collection for affordability assessments, enabling better payment plan structuring and debt advice.
- 7:04 – 8:27
Market Size and Use Cases
Curtis estimates millions of people in the UK require affordability assessments, ranging from those in pre-default to wealth management clients needing financial health checks. The market extends beyond debt into areas like targeted advice and simplified advice.
- 8:27 – 16:03
Performance Metrics and ROI
Clients report 90-95% completion rates for self-serve journeys, compared to 50-60% for manual or spreadsheet methods. Benefits include higher payment plan amounts (up 60% in one case), improved stick rates (3-5% increase), and reduced agent handling times (e.g., Hastings Direct from 50 to 8 minutes). ROI is cited as up to 100x.
- 16:03 – 17:51
Data Privacy and AI Guardrails
Curtis explains that personal data is redacted before entering the LLM, with security standards including ISO 27001 and ISO 7001. The AI is strictly constrained by prompt coding to stay on task, preventing hallucinations or irrelevant responses.
- 17:51 – 24:45
Barriers to Entry: Data and Regulation
While building a demo is easy, launching requires millions of real conversation examples to handle exceptions and sensitive customer behavior. Inicio AI is the first and only FCA-authorised provider for this specific service, having navigated the FCA innovation pathway to enable data passporting.
- 24:45 – 32:01
Fundraising Lessons and Future Strategy
Curtis discusses her fundraising journey, noting that underpromising hurt her initial pitches. She now pitches Inicio AI as an international data business leveraging rich financial data, targeting Commonwealth markets (Australia, NZ, Canada) rather than the US due to regulatory differences.
- 32:01 – 32:49
Market Expansion Strategy
Speaker 1 states that the US is not the first market for expansion due to a prevalent 'shout and pay' collection culture, preferring to target it later. The current strategy focuses on supporting consumers to determine affordable payment plans, which leads to better long-term outcomes despite potentially longer resolution times.
- 32:49 – 34:27
Five-Year Vision and Product Positioning
Speaker 1 defines the company's mission as providing the most reliable and accurate income and expenditure product. The five-year goal is to become the standard 'plug-in' for affordability assessments across debt and wealth sectors, analogous to Intel inside a laptop.
- 34:27 – 35:32
Rapid Fire: Reading and Advice
Speaker 1 discusses reading history books about royalty as a break from business. She advises founders to 'say yes' to every opportunity and introduction, noting that growth is often non-linear and relies on networking chains.
- 35:32 – 36:23
Brand Identity and Consumer Psychology
Speaker 1 explains that 'Budgie' is a pun on 'budget' and 'income and expenditure' (IE). The brand uses a non-judgmental budgie avatar to create a soft connection with consumers and agents, even distributing physical plush toys during client rollouts.
- 36:23 – 37:29
Contact Information and Outro
Speaker 1 directs listeners to inisio.ai and LinkedIn for more information. The host, Tom Britton, concludes the episode by promoting the podcast and his own venture, Syndicate Room.
Claims worth verifying
Getting an AI agent past a bank and regulator requires hundreds of real conversations, every exception handled, and proof of safety for vulnerable customers.
Unidentified speaker (UNKNOWN) ·0:07
How to verify: Primary source: Podcast transcript [0:07]. Verification: N/A (Speaker assertion of process requirements). For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Rachel Curtis was named KPMG's UK Tech Innovator of 2025.
Unidentified speaker (speaker_0) ·0:46
How to verify: Primary source: KPMG UK Innovation Awards archives or press releases for 2025. Verification: Search KPMG website for 'UK Tech Innovator 2025' winners. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Curtis's first job was on the counter of a Halifax Building Society.
Unidentified speaker (speaker_1) ·1:24
How to verify: Primary source: Rachel Curtis's LinkedIn profile or CV. Verification: Check 'Experience' section on LinkedIn for Halifax Building Society. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Inicio AI is a B2B SaaS company with a virtual agent called Budgie.
Unidentified speaker (speaker_1) ·4:41
How to verify: Primary source: Inicio AI website or pitch deck. Verification: Visit inicioai.com or check Crunchbase for company description. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Millions of people in the UK are struggling with affordability assessments, with stats ranging from 7 million to 17 million depending on the definition.
Unidentified speaker (speaker_1) ·7:10
How to verify: Primary source: UK financial inclusion reports (e.g., StepChange, Citizens Advice, or MoneyHelper). Verification: Search for 'UK debt statistics' or 'affordability assessment prevalence' in financial inclusion reports. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Clients see 90-95% completion rates for self-serve journeys using Budgie, compared to 50-60% for Excel/spreadsheets.
Unidentified speaker (speaker_1) ·9:27
How to verify: Primary source: Inicio AI case studies or client testimonials. Verification: Request case study data from Inicio AI or check published client success stories. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
One client saw payment plan amounts increase by 60% using the platform.
Unidentified speaker (speaker_1) ·10:56
How to verify: Primary source: Inicio AI case study. Verification: Check Inicio AI website for client results or request specific case study data. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Hastings Direct reduced agent handling time from 50 minutes to 8 minutes.
Unidentified speaker (speaker_1) ·12:06
How to verify: Primary source: Hastings Direct or Inicio AI press release/case study. Verification: Search for 'Hastings Direct Inicio AI case study' or 'Hastings Direct agent time reduction'. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Inicio AI is the first and only FCA-authorised provider for income and expenditure assessments.
Unidentified speaker (speaker_1) ·22:19
How to verify: Primary source: FCA Register. Verification: Search FCA Register for 'Inicio AI' or 'Budgie' under 'Authorised firms' and check 'Permissions' for income and expenditure services. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceCurtis won the 'One to Win' competition with a prize of £1.3 million.
Unidentified speaker (speaker_1) ·25:47
How to verify: Primary source: One to Win competition archives or press releases. Verification: Search 'One to Win competition winners 202X' or check One to Win website for past winners. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
The US is not the first market for expansion because the collection culture involves shouting until payment, whereas the current strategy supports consumers to find affordable plans.
Unidentified speaker (speaker_1) ·32:01
How to verify: Check Inicio AI's official website (inisio.ai) or press releases for stated market entry priorities and geographic roadmap. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
The company's five-year goal is to become the standard 'plug-in' for affordability assessments, similar to Intel inside a laptop.
Unidentified speaker (speaker_1) ·33:15
How to verify: Review Inicio AI's website mission statement or investor pitch decks for product positioning and long-term strategic goals. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
The brand name 'Budgie' is a pun on 'budget' and 'income and expenditure' (IE).
Unidentified speaker (UNKNOWN) ·35:32
How to verify: Check Inicio AI's branding materials, website about page, or marketing content for the etymology of the brand name. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
The company distributes physical plush budgie toys to agents during client rollouts to create a non-judgmental connection.
Unidentified speaker (speaker_1) ·35:57
How to verify: Search for social media posts from Inicio AI or client partners showing agent onboarding events or promotional materials featuring the budgie mascot. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Tom Britton is the host of Angel Insights and associated with Syndicate Room.
Unidentified speaker (speaker_0) ·37:05
How to verify: Check the Apple Podcasts or Spotify listing for 'Angel Insights' for host credits and the website syndicate room.com for company details. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Do your own research
- Verify FCA authorisation status for Inicio AI on the FCA Register.
- Check ISO 27001 certification validity for Inicio AI.
- Review Hastings Direct's public reports or press releases for claims regarding agent handling time reductions.
- Analyze competitor offerings in the UK affordability assessment space to assess the 'first and only' claim.
- Examine the FCA innovation pathway documentation to understand the regulatory hurdles cited.
- Research the market size for affordability assessments in the UK and Commonwealth countries.
- Review the data privacy policies and redaction processes described for LLM integration.
- Compare the completion rates claimed (90-95%) against industry benchmarks for digital form completion.
Full transcriptexpand
0:07Unidentified speaker (UNKNOWN): With
0:07Unidentified speaker (speaker_0): today's AI coding tools, you can build a demo of an AI agent in just a few hours, sometimes less. But getting one past a bank and a regulator, well that's a whole nother story. You need hundreds of real conversations, every exception handled, and proof that it's safe with vulnerable customers and their data. And once it's live, keeping it right is often harder than building it in the first place. Welcome to Angel Insights, the show that goes deep into the UK startup ecosystem. I'm your host, Tom Britton, co-founder of Syndicate Room, and on today's episode, I'm delighted to be joined by Rachel Curtis, co-founder and CEO of Inisio AI, KPMG's G's UK tech innovator of 2025. On this episode, we'll hear why people in debt tell an AI what they need to
0:50Unidentified speaker (UNKNOWN): do.
0:51Unidentified speaker (speaker_0): never tell a person in real life, why banks can't just build their own, and why underpromising was the wrong habit for pitching VCs. But that's enough for me. Let's welcome Rachel to the show. Rachel, thank you so much for joining us today. It's a pleasure to have you on the show.
1:05Unidentified speaker (speaker_1): Thanks for having me, Tom.
1:07Unidentified speaker (speaker_0): Before we dive into Budgie, which I'm sure people are excited to hear about, I want to talk about your background a little bit, because you spent 30 years in financial services before setting an issue
1:19Unidentified speaker (UNKNOWN): up.
1:19Unidentified speaker (speaker_0): Where did your career start and what were you doing by the end of
1:23Unidentified speaker (UNKNOWN): it?
1:24Unidentified speaker (speaker_1): Um so I have to cast my rhyme back a long way here, Tom. My actual my first job was on the counter of a Halifax Building Society. Um and I kind of think really fondly about those days because it taught me something really important. around consumers, how they behave, the challenges they're struggling with. So I think that customer facing experience has been super valuable in in my whole career actually, with everything I've then gone on to do.
1:51Unidentified speaker (speaker_0): And then before starting initio, what was the last role?
1:56Unidentified speaker (speaker_1): Yeah, so I was chief experience officer. So kind of running the operations and the marketing and and customer side of a challenger bank.
2:06Unidentified speaker (speaker_0): Now I find that every entrepreneur has some form of story that set him down the path of of creating a business. And before the show you kindly shared with me a story about a lady who was pregnant and in debt. And if you don't mind talking Walking us through that story and what it left you thinking, that'd be great.
2:25Unidentified speaker (speaker_1): Yeah. I I guess I didn't plan to be an entrepreneur. It was sort of I was faced with quite an emotional problem and a solution that I felt could fix it. And that kind of, you know, I thought I didn't have any option other than to try and solve that problem, I guess. So yeah, you're right. It was a very heavily pregnant lady who was it was one of my friends who was in London one night on a night out and came across this lady who was unfortunately. trying to take her own life and to try and keep her safe while I was waiting for the police. He was sort of chatting to her, and her story was she was in debt. She owed money to a number of companies, and all those companies were asking her to complete an an income and expenditure form
3:04Unidentified speaker (UNKNOWN): or
3:04Unidentified speaker (speaker_1): an affordability statement, like a budget form type thing, so that they could work out how they could help her financially, whether they could set a payment plan, that kind of thing. And her challenge was. she was not very good with her money. She was being asked to fill in a complicated financial form. She was struggling to do that on her own. But she was so embarrassed about a situation that getting the help of a a human, you know, an agent in a call centre to get a get help to do that was also a real barrier for her. So she just felt stuck and that she didn't have any options.
3:35Unidentified speaker (UNKNOWN): And
3:35Unidentified speaker (speaker_1): it was about that time I was working on a digital transformation project. and was starting to come across because this is sort of six odd years ago, was coming across the early stages of sort of the the small and large language models and and what we used to call chatbots that we all were very frustrated by six or so years ago. The new models were starting to show that they were capable of much more. And when I was looking at that technology, my view was well, I'm just not sure as humans how we're going to adopt that technology unless you find me a group of consumers that need help but won't get it from a human. And that's when I came across the story of the pregnant lady who was exactly an example of, I really need some help, but I'm just too embarrassed to get that help from a human. And it was that realization that maybe this new technology could create a solution that would support that consumer to complete the task they needed to do without them having to speak to a human.
4:31Unidentified speaker (UNKNOWN): So
4:31Unidentified speaker (speaker_0): we've hinted a bit about what the the product is and what the business is, but for those who haven't put all the pieces together, could you give us your elevator pitch for the business?
4:41Unidentified speaker (speaker_1): So Inissio is a B2B SaaS company and we've created a Especially specially trained virtual agent that guides consumers through completing an affordability assessment. So our virtual agent is called Budgie, although everything in the platform is configurable. So even Buggie can be a completely different avatar if our customers already got their own sort of chatbot or brand in there. And the platform basically guides agents and consumers through completing a detailed, verified, auditable income and expenditure statement. So you've got a really good understanding of the financial position of the end consumer. And then you can go on and set up payment plans or centre debt advice or whatever's appropriate for that consumer because you really understand the position that they're in.
5:28Unidentified speaker (speaker_0): For someone who's not had to do an affordability assessment, can you talk us through what's involved in the process?
5:36Unidentified speaker (speaker_1): Um, but the standard financial statement is quite long and complex. Um, and for good reason, for us to really assess somebody's affordability, you've got to understand it in detail. Because there's no point saying to somebody, oh, what's your household expenses? Because what they'll do is they'll add up sort of their mortgage, maybe their utility bills, their council tax. So everybody will include slightly different things in there and everybody rounds up figures because it it's like, well, actually, my my my rent is 760, but I can't add that number up. So I'll just go 800 and my electricity bills may be 164, but I'll just say 200. So you just don't get accuracy if you do this sort of short form or a more sort of what you think might be a faster version of doing it. But yeah, the income and expenditure statement has, depending on your circumstances. could be like 80 fields of everything broken down. Because without the complexity, you don't have accuracy. And therefore you might set up a payment plan, but it might not be the optimal amount or that payment plan might break because you haven't really got to the actual detail of it.
6:44Unidentified speaker (speaker_0): I'll be honest, I think I would fail if that was an exam. That sounds quite complicated. I'm I'm glad you're you're building this system. How many people need to go through the affordability assessments? You know, is is it half the population one in ten, or
7:04Unidentified speaker (UNKNOWN): what
7:04Unidentified speaker (speaker_0): what numbers are we talking about if we're looking just at the UK market that are having to do these?
7:10Unidentified speaker (speaker_1): Millions. I mean, you can get different sources of stats. You've got some that will say, there's seven million people that are struggling in this way. Some will pay say it's 17 million if you bring in others. Um, because I guess affordability assessments come in at different stages, so you can have it at priories. So it might be that a customer's saying, I'm going to struggle. I've just lost my job. All my payments are up to date. I'm not in debt, but my financial circumstances have changed, therefore, I need to. check my affordability and work out what to do. And then you've got all the way through the life cycle. We also do stuff sort of even earlier than that at origination stage. Um you've also then got a new market which is the wealth management market for the product because there's the new stuff around targeted advice simplified advice and targeted support where we're trying to get wealth management to sort of the masses. And again, you need to do a quick financial health check for those customers. If somebody says, well, I want to invest £500 a month, there should be a quick check of is that a good idea for that consumer? Are they are they is their income and expenditure robust enough to support that? So we don't just do affordability in that debt space. So yeah, uh slightly woolly answer, but millions and millions, Tom.
8:27Unidentified speaker (speaker_0): That's the number I was looking for. It's not just a a technology though. Like you've already got customers coming in. You've got people using using the agent. We'll call it an agent. And what with these early customers, what are what are you finding in the data? Talk me through what you're seeing from from those first customers.
8:47Unidentified speaker (speaker_1): Yeah, so the data's really powerful and it's it's it's really good from my perspective because it's helping us secure more and more clients because you kind of the the ROI for the benefits is is is really compelling. So all the way through what we do, so right up put at the front, more customers engage because we've now given them a way to get the support they need via the virtual agent at two o'clock in the morning. The call centre would normally be closed. So if if they do need some help so we can provide that support 24 7. So that gets more customers in at the top of the funnel to get help in the first place. We then get more customers through the journey. So we convert better. So we typically get sort of um I mean some of our clients are seeing ninety ninety five percent of customers that start self self-serve serve through all by themselves. Um
9:37Unidentified speaker (UNKNOWN): What would be the
9:38Unidentified speaker (speaker_0): dropout if it was an agent?
9:40Unidentified speaker (UNKNOWN): So
9:40Unidentified speaker (speaker_0): you're getting 90 to 95% completion through through your technology.
9:46Unidentified speaker (speaker_1): So the the agent dropout kind of really varies by client. I mean, they do see a lot of when you start the journey with an agent, some consumers will say, Look, I I this is too much information, I'm embarrassed. So you do get a call dropout. That'll really vary by client and depending on how how experienced their agents are. I guess one comparison figure is we've seen a lot of our clients that were using other forms of their own Excel spreadsheet or something they built themselves and typically conversion rates on that are nearer the 50%, 60% if they're really lucky and they built a really good one. But typically we're almost doubling the conversion of the throughput three put through the journey. So more in the top, more get through it. The other benefits come from better customer disclosure. So customers are really happy to disclose all sorts to the AI because it's not human, they're not being judged. So we get a lot more detail, a lot more colour to kind of the case, which is really helpful sort of assessing vulnerabilities and making sure people get the right support. And then at the end of the journey we get typically higher payment plans. So one of our clients saw the amount of their payment plans
10:56go up by 60% because we were given the support. And we we gently check and challenge the customer and prompt them within the form if we think they're making mistakes or you know, just sort of say, Well, that looks a little bit high or that looks a little bit low. So by getting that accuracy, we tend to get a higher payment plan. So more money. coming in. And then those payment plans tend to stick better. So our clients have seen, you know, some of the banks have seen their stick rates of payment plans go up by sort of three to five percent.
11:26Unidentified speaker (UNKNOWN): Now
11:26Unidentified speaker (speaker_1): that doesn't sound like a massive number, but if you talk of the millions of payment plans they're dealing with, you've got more people paying more money and that payment plan sticking. The ROI is really huge because you don't have to then go back. If a payment plan breaks, you have to go back and do the whole process again. So we're avoiding all that rework as well. Um, and then the other benefits you've got is it's faster for agents because we support agents through the process as well, and the way we've built our experience for agents. Um, some of our clients have seen agent time sort of half in terms of if they're still doing it on the phone. Um one of our clients in its at public case studies, Hastings Direct, fifty minutes they used to take, they're now taking eight minutes. So that transition and cost saving, and it's you know, the consumer was able to do it when they wanted to in their time, not just between say nine and five when the call centres centers open. So it's kind of better all round on lots of levels.
12:25Unidentified speaker (speaker_0): I just want to repeat that. So
12:27Unidentified speaker (UNKNOWN): you
12:27Unidentified speaker (speaker_0): went from almost an hour to under 10 minutes on a case or Hastings Directed.
12:33Unidentified speaker (speaker_1): Yeah. Yeah. Because the consumer was able to self-serve to do the painful part of the call, which they preferred doing on their own. And the agent was just picking it up at the end, just double checking everything and then setting up the payment plan with the consumer. I mean ultimately that whole end end-to-end to end process can be digitized. And the customer can set up the payment plan by themselves. Um, so you can save the whole call in in theory.
12:57Unidentified speaker (speaker_0): If we just talk about that call right there in that case, so they're seeing a reduction of you know 80% in the in the time that their agents are are spending. That's productivity. It can be reduction of cost because they don't need so many agents doing it. What's the cost to them? of using the service?
13:17Unidentified speaker (speaker_1): So we charge with sort of typical SAS. So there's an implementation fee to get set up. So we create a fully branded, tailored, configured version of the platform. And then they pay like a per they they buy like a tier of usage and they pay sort of per form. So the saving is sort of, you know, it could be 10 ten times. We've seen one client that have saved an ROI. of over a hundred times because of all of the benefits of more cases coming through, collecting higher payments on it and the stick rate being higher, and the AHT, the average handling time saving of the agents. So when you wrap them set them all up, it's, you know, like a hundred times ROI is not unusual for our clients. What what's been really nice for our clients though is it's not necessarily been about cutting hair. head cut headcount. It's been about freeing up resources to either deal with the growing number of consumers that are struggling. So there's more customers that need help. So whilst they haven't cut the headcount, they haven't had to grow the headcount to deal with that growth in demand, but also it frees up agents to help with the more complex, more vulnerable cases. And we I was sat in a call centre center on Friday, and you and can just hear some of the, you know, the cases that, you know, agents are having to deal with and the level of vulnerability for some consumers is is really quite shocking.
14:35Unidentified speaker (UNKNOWN): But
14:36Unidentified speaker (speaker_1): what we're doing is freeing up those agents so they have got the time to have those really detailed conversations and provide that support.
14:43Unidentified speaker (speaker_0): That's fantastic. And I want to go back to an a a number that you shared earlier. So you mentioned that for one of your clients, you're seeing a sixty percent increase in in payment amounts from the clients. Now, putting my skeptic hat on,
14:58Unidentified speaker (speaker_1): I
14:58Unidentified speaker (speaker_0): see that and I'm going, okay, this is technology designed to get more out of people, vulnerable people. What do you say back to that? Like how do you explain it that actually it's for the benefit of the client as well?
15:12Unidentified speaker (speaker_1): Yeah. So one of the things with so so some consumers will sort of complete an income and expenditure and they might game it because they know they only want to pay a pound off that debt. That's their end target. But actually, that's not in the best interest of that consumer because they're going to be paying that for a long time and they're never going to potentially clear it because they're going to be paying a lot of interest. So actually it is in the interest interests of the consumer to get debt cleared down as quickly but as safely as they can afford to. So the whole point of doing a really accurate affordability assessment is to get to the optimal payment plan that is optimal for the consumer and it's also optimal for the organisation. So we're in one of those sort of rare situations where it works the same way for both parties involved, really.
16:03Unidentified speaker (speaker_0): Obviously, consumer data is a big topic and keeping it nice and secure. And you've you've got AI models running on this, handling quite sensitive information. What are the guardrails that that you put in place to ensure
16:18Unidentified speaker (UNKNOWN): that
16:19Unidentified speaker (speaker_0): none of that is is getting out beyond where it needs to be?
16:23Unidentified speaker (speaker_1): So everything that goes to the actual AI. the LLM, is there's no personal data. So it's all redacted. So it's understanding intent and utterances, but it's it doesn't know whose they are. So that we keep very separate. Um the platform itself where the data is is all fully, you know, locked down secure I say 27,001, 7001, all of that good stuff to make sure that's the case. And then I guess the other guy guardrails are really about how Budgie is programmed and what Buggie can do and not do importantly. So Budgie's got, you know, tens of thousands of lines of prompt coding that are all about this is your task, you must stay on task, you cannot do this. You know, if a consumer says, Write me a poem about a tiger, you know, that will break through some some 'cause because it's quite an innocuous. request. But Budgie will be like, I'm only here to help you with your financial form. I'm sorry, I can't help that. And we do a lot of training and testing to make sure that Budgie stays within the limits of what she's she's supposed to do. And I think that for me is, you know, if we talk about AI more generally, I think that's where where AI
17:33has got really strong use cases on a really narrow, locked down, clear use case, I think it works really, really well. And it can also be safe.
17:44Unidentified speaker (speaker_0): I I think I agree with you with having these very locked down niches and specific use cases that you're building around.
17:51Unidentified speaker (UNKNOWN): Um
17:51Unidentified speaker (speaker_0): elaborating on that a little bit, because everyone who's got CloudMax or Astra Six going will We'll go, Oh, I could probably rebuild that. Why isn't one of these large commercial institutions or the bank building it for themselves?
18:05Unidentified speaker (speaker_1): Yeah, that is the most common question I get asked by investors because everybody's sort of like, Well, this sounds great and it's obviously working well, but technology's moved on so quick, surely I can just build this. And we've actually done an article recently where, you know, the barrier to demo is dead, but the barrier to launch is almost even higher now. So if you can't build a prototype of a version of Buggie in, I mean, I'd probably say hours these days. It used to be sort of weeks, it was then days, probably hours. You're not using the right tools. You could build a demo super, super fast. The problem is to make that work in what is a very sensitive, you know, vulnerable customer, heavily regulated type environment, you actually need millions of examples. of what consumers actually do, how they chat to Budgie, how it should work, and be able to deal efficiently and effectively with the reality of consumers and agents having these conversations, the exceptions, how you handle them. Um I mean, I I can give an example from very early on in Budgie's life. We were
19:11trying to get the cla calibration of the sensitivity on vulnerable behaviour right. Very early on, a consumer said, I'm done.
19:20Unidentified speaker (UNKNOWN): And
19:20Unidentified speaker (speaker_1): Budgie dove in with, Oh, I'm really sorry to hear you struggling. Here's why don't you try a call in the Samaritans? They're there to help. And the customer was like, No, no, I've just finished the form, you know, chill out. But it's it's that sort of thing that it's very easy when you go, Oh, this is what the customer meant, and therefore we're going to gonna do that. But it's never that black and white. One of the really lovely examples I use of very early on with RAI is a consumer said, My booze is fifty quid and I spend a pony on fags.
19:51Unidentified speaker (UNKNOWN): Uh
19:51Unidentified speaker (speaker_1): so the AI needs to be able to understand that they meant twenty five pounds on smoking products in the form. So it's that consumers speak in a very colourful, nonlinear way. They'll jump all over the place. and talk about all sorts of different things with their money. And it's we've got six years of that experience now. So yes, you could build a prototype. You would never get it past a bank or a regulator without having the data and experience to show that it works and it works safely. And even if you manage to do that bit, that's not the hard bit. The hard bit then is maintaining it. So so often benefits amounts change, TV licence figures change, council tax figures change. We have to keep all of that up to date because we're prompting the consumer in the form to say, Well, that doesn't look quite right. Have you got that figure right? That kind of stuff. So maintaining the system is harder than building it. And it is it is very niche, it is very specialist. And I guess we've also got data across all sorts of different clients. That means our our training data is much broader and richer versus if a cons if a client organization tried to build this themselves, they'd only really have their data and their consumers. So it's kind of another reason not to build a
21:07narrow version of within your own organization. Go and get one that's used much more broadly.
21:12Unidentified speaker (UNKNOWN): Well
21:12Unidentified speaker (speaker_0): I think on that point as well, you were the first firm with FCA permission to passport affordability data so that someone who does it once doesn't have to keep doing it again and again. I mean congrats on that. But regulation must be a barrier as well and permissions. What are the the hoops that someone who wanted to try and replicate this, what would they have to go through?
21:35Unidentified speaker (speaker_1): Yeah, I'm also a non-exec non exit director and senior independent of a regional building society. So I sit in those regulated roles and was sort of very conscious of you don't necessarily strictly have to be FCA authorised to provide B2B software to do income and expenditure forms. You know, there's a way round it you you don't have to have that regulation. But to my mind, the process and what we're doing for consumers and how we're supporting them is so critical to them getting a good outcome at the end of it that I felt we should be regulated. Um, we worked with the FCA on their innovation pathway because the authorisation didn't exist in our space. Yeah, we did become the first, and we're still the only FCA authorised provider that does exactly what we do in terms of the income and. expenditure space. It does help our clients as well. You know, we work with banks and they're working with another FCA authorised firm, so it does help. Um, where we specifically needed the FCA permissions was passporting. So when a
22:39consumer's done an income and expenditure, if they want to share it with other organizations, those are the permissions commissions we've specifically got to
22:47Unidentified speaker (UNKNOWN): be
22:47Unidentified speaker (speaker_1): able to share that on. And the the purpose for doing that is if
22:50Unidentified speaker (UNKNOWN): a
22:50Unidentified speaker (speaker_1): consumer can get all of their debts under control, they're more likely to maintain the payment plans on all of them. So again, it comes back to how do we help consumers have good outcomes. And that's why I sort of passionately believed in going down that status. But it is more time, more cost, more people. So it's not it wasn't a cheap decision, but I believe it was the right one.
23:15Unidentified speaker (speaker_0): Well, I also believe it's the right one, and it seems to be paying off. You've got BP Parabus, Hastings Direct, Tandem, and now Octopus Energy as clients, some incredible names for what is a scaling business. What were some of the key factors in them trusting you as as a smaller firm? Yeah, because it's in this space it is quite difficult to get an incumbent, a large institution to work with with a startup. What what convinced them?
23:48Unidentified speaker (speaker_1): Um I guess for some of them it was the fact that we'd already got case studies of working with organizations like them. Um for some of them, they're just genuinely the innovators in the space and they were comfortable. So BMP Parabar, for example, we were part of their innovation lab where we went through a programme of working with the business area inside the bank to really understand the problem and really design the solution that that solved their needs. So yeah, it's a mix. You've got the innovators. I think we're fortunately now past that early adoption. We've got sort of more than 20 clients. and, like you say, some really strong brands using the product. So we're no longer, you know, the the risk. We're not quite in the nobody got fired for buying AB IBM just yet, but we've definitely moved a long way up that curve from, you know, when we were trying to convince our very, very first customers.
24:45Unidentified speaker (speaker_0): Changing topic slightly, but you shared something with me before that a few years ago you didn't know what a cap table was. So I'm curious. Without knowing what a cap table was, what did your first fundraise look like and how did it come together?
25:03Unidentified speaker (speaker_1): Um so we've kind of done a few rounds. Initially it was sort of bootstrap, sweat equity, you know, everybody worked for nothing basically at the start. Um but you yeah, bootstrap, friends, and family kind of stuff was the very early sort of funding. We then did a pre seed round where we got some angel groups, tobaccos. We then went through a seed round where um actually, slightly before that, we managed to get on the Morgan Stanley Inclusive Ventures Lab, which is a fantastic programme. I'd recommend the founders to look at that. one. Not only did we get a quarter of a million funding, which was super helpful, it was a five month programme of really detailed support, guidance, training, education, connections, all that kind of stuff. And then as part of our seed round, we won the one to win competition. So we won a pitcher competition for one it was supposed to be a million prize, but we actually won one 1.3 point three million. They increased the prize. fund. Um, and that helped sort of close our our seed
26:07round.
26:08Unidentified speaker (UNKNOWN): And
26:08Unidentified speaker (speaker_1): I guess it was a very steep learning curve, um, not without bumps in the road. I got an awful lot of no's on that journey because I went to the wrong investors at the wrong time with the wrong message and had to learn from all of those kind of bruising experiences to get through to the point where, you know, yeah. we did have a successful seed round ultimately.
26:30Unidentified speaker (speaker_0): I hate to bring up the elephant in the rear, but we're we know the numbers for female founded businesses in raising capital are dire.
26:38Unidentified speaker (UNKNOWN): I
26:38Unidentified speaker (speaker_0): think it's it's less than 2% goes to all female founded teams. I think it's about 14% goes to mixed teams. Given your background, 30 years of experience in financial technology,
26:49Unidentified speaker (UNKNOWN): a
26:50Unidentified speaker (speaker_0): deep understanding. of the customers and all these case studies, why do you think it was still so hard to raise some of these rounds? And why do you think that female founders are still finding it so hard to raise?
27:01Unidentified speaker (speaker_1): I think there's so many reasons on that. And they're they're all sort of adding up to give that that really poor end result, I guess. From my perspective, my learning was I didn't go in with enough ambition. in terms of our numbers because you know I've I've struggled with imposter syndrome in my career and I've also built a really successful banking career on underpromise and over deliver, which worked really, really well in banking. You sit in front of an investor underpromising and going, look, I've presented you a case that says this business will get to, I don't know, 10 ten million pounds worth of they're not interested. It's not big enough. But for me to present one that's at a billion, which is what a lot of the VCs were looking for, I'm like, well I don't know. I've hardly got any revenue yet. How how can I, you know and what I needed to do was just go, yeah, fine. Here's an Excel spreadsheet. It's got a lot of logic behind it. This is what I'm forecasting. And I needed to be super brave.
28:03but I didn't have that experience or background to know. Like I say, I spent 30 years underpromise deliver. And here I was in a a scenario where I was supposed to overpromise because that's what you had to do to get through to the VC sort of conversations. I did my master's dissertation on imposter syndrome, and it isn't just something that's experienced by women, it's experienced by men and women equally, but it does tend to manifest slightly differently in women where they will underpromise on stuff. So so those sort of financial scenarios I think women can find harder to stand behind and create the kind of crazy financial models of getting to the, you know, the billion pound sort of size business. So I think that's one of the problems. I think the other one is we tend to back people that look like us. And I can confirm that
28:57Unidentified speaker (UNKNOWN): I
28:57Unidentified speaker (speaker_1): have never been in front of an all-female investor panel. I have been in front of many all male investor panels. I think it's better. We're certainly back, we're backed by three female angel groups, which are brilliant. We've got Angel Academy, Mint Ventures, and Hermesa, Hameesa, and they've been really great. not just with the capital, but also with supporting me as a female founder. So there's really good support networks there. And they're not just exclusively female investors. You know, there's there's men in there that champion women as well.
29:27Unidentified speaker (speaker_0): Well, I would love to read your thesis at some point. I think that'd be fascinating. Um, but talking about the the other learning that you that you mentioned before, so about being bolder, braver, and more ambitious, how are you pitching the business and at what scale are you pitching it? So you're likely gonna need to raise another round, I would guess? How are you gonna take those learnings and apply it to that round?
29:52Unidentified speaker (speaker_1): So our investor pitch doesn't really lead on budgie necessarily. It leads on an international data business. And that's the messaging that that is more powerful, I guess, because in creating our project product that helps consumers gather affordability data, what we gather is very, very, very rich, detailed, broken down data on people's actual finances. A lot of the data in the space is sort of model data, and it's directional rather than specifically accurate to. an individual circumstances. So we've got that more detailed data. The product we've got can be rolled out internationally, um, it works in different languages. Um, you know, you can use voice as well as texting, all that kind of stuff. So it can support all sorts of customers all over the world. And therefore, that that data becomes even richer with that sort of world view that that's the business we now pitch. and that's what investors have backed. Budgie is our product to get to that data, but the value of our business is is in that data really.
31:00Unidentified speaker (speaker_0): Given that pitch, you know, the US would seem like a great market to go after. It's a bit bigger than than the UK. Um, but your head is shaking, which other people can't say. And I'd love for you to to explain that. Like
31:13Unidentified speaker (speaker_1): staying.
31:14Unidentified speaker (UNKNOWN): Where are you going?
31:15Unidentified speaker (speaker_0): to go?
31:15Unidentified speaker (speaker_1): Yeah, so the international markets would be more of the sort of um X uh not X more of the sort of Commonwealth type businesses. So Australia, New Zealand, Canada, they collect debt in a very similar way to the UK, or they're regulated in a very similar way. US is very different. Obviously, you've got all the different states. and they approach it very, very differently. They tend to adopt a methodology of you need to just pay. I don't care whether you can afford it or not, you need to just pay. So for us to break the US would be a slightly different approach. We will break the US when we've got a number of countries and we can demonstrate in all of those countries that if you take the time to support the consumer to understand their affordability, your end result is better payment plans that stick that would drive a change in the US behavior from we're just gonna sh
32:13Unidentified speaker (UNKNOWN): we're
32:14Unidentified speaker (speaker_1): just gonna shout at you till you pay and some people won't pay, we're gonna support you to work out what you can afford to pay and it might take a little bit longer, but the end result is better. So that's why I won't be rocking up to the U US as my first market, but it will be one of our sort of later international targets.
32:32Unidentified speaker (speaker_0): And before we go into the rapid fire questions, I want to just do a quick exercise. So I'd love for you to project five years down the line, businesses growing and achieving what you want it to achieve. What does the business look like and has the mission changed at all?
32:49Unidentified speaker (UNKNOWN): So
32:49Unidentified speaker (speaker_1): our internal state. statement is to be the most reliable, the most trustworthy, the most accurate income and expenditure product on the market. So success for us is we help our clients get to the most detailed, clear position of affordability, because that has the best outcomes comes for everybody. Um in five years,
33:14Unidentified speaker (UNKNOWN): I
33:15Unidentified speaker (speaker_1): expect that to have translated to us being just the way that affordability is done. So we're that plug-in because we are a plug-in to a broader process. There's lots going on around sort of affordability assessments in the debt space, in the wealth space and everything else. Um but we are that plugin that just powers I guess you know you've got that Intel inside and a laptop. I kinda want us to be the equivalent of that. We're just the plug in that you use to do affordability because it's the strongest, most effective. It's just the way to do it.
33:53Unidentified speaker (speaker_0): You know, sim similar sounding. I
33:55Unidentified speaker (speaker_1): I need
33:55Unidentified speaker (UNKNOWN): to get a little
33:56Unidentified speaker (speaker_1): jingle. I need to create a jingle so that, you know, we don't that's our brand. You have to play the jingle.
34:02Unidentified speaker (speaker_0): Just make it different to the Intel one, because I think
34:04Unidentified speaker (UNKNOWN): that one
34:05Unidentified speaker (speaker_0): gives me nightmares. So
34:07Unidentified speaker (speaker_1): it could be a budgie tweeting, maybe. That won't be annoying at all, will it?
34:11Unidentified speaker (speaker_0): Uh honestly, it's it's been a pleasure chatting. I I do like to end every show with a few,
34:15Unidentified speaker (UNKNOWN): a few
34:16Unidentified speaker (speaker_0): quickfire questions. So nothing too hard. First thing that comes to your mind, just just let me know. And the first one is always, what's the best book that you've read recently from any genre?
34:27Unidentified speaker (speaker_1): So this is nothing business related at all. I I'm really fascinated with history and the royals. And I read a book that goes back right to the very first sort of kings and you trace through sort of how where it went and what happened and all the weird and wonderful stories of kind of history. So yeah, I don't know how that's supposed to help me in business, but sometimes I think you need a break from business. books and I did something completely different.
34:53Unidentified speaker (speaker_0): Second question, what's the best piece of advice you've received as a founder?
34:58Unidentified speaker (speaker_1): That's really hard. I've received so much good advice. I guess if I boiled it down, it would be say yes.
35:07Unidentified speaker (UNKNOWN): And
35:07Unidentified speaker (speaker_1): that's really dangerous, isn't it? But I think just say yes to every opportunity, say yes to every conversation. where we've been helped, it's not been linear. It's been, you know, you were at an event and you met somebody who introduced you to somebody else, who ended up with somebody else. And then do you know what I mean? It's it's that kind of thing. So say yes to every sort of opportunity and introduction you're offered.
35:32Unidentified speaker (UNKNOWN): And
35:32Unidentified speaker (speaker_0): the last Next one, why a budgie?
35:37Unidentified speaker (speaker_1): So it was a bit of a joke at the start, but budgie sounds like budget, and also it ends in in I and E, IE, so income and expenditure. So we thought we could do some sort of clever stuff with it. But we'd always plan to do some research and look at whether it was the right avatar. But it's been so popular with consumers and agents, even to the point we when we roll out with a client, we take little fluffy budgies in to give to the agents and it just creates that connection with a really non-judgmental, soft character, I guess.
36:12Unidentified speaker (speaker_0): Honestly, thank you so much for sharing all of that with us today. It's been absolutely a pleasure having you on the show. For those that want to find out more about what you're building, where should they go? What should they do?
36:23Unidentified speaker (UNKNOWN): Uh
36:23Unidentified speaker (speaker_1): we're inissio.ai online and also LinkedIn. We tend to use quite a lot as well.
36:29Unidentified speaker (speaker_0): Well, thank you again. And for now, I'm going to say goodbye. Hopefully we can revisit this in say 12 to 18 months, see how everything is going, see what countries you've moved into and and what other products you've come up with. But for now, have a lovely rest of your day. We'll speak again soon.
36:44Unidentified speaker (speaker_1): Thanks, Tom.
36:46Unidentified speaker (speaker_0): Well, that's it for this episode of Angel Insights, and I do hope you found it as interesting as I did. If you did enjoy this episode, could I ask that you follow Angel Insights on Apple Podcasts, Spotify, or any of the players that you do? Like, subscribe, tell a friend, and it would mean the world to us. To find out more about Inisio, head on over to inisio.ai or connect with Rachel Curtis on LinkedIn. And to find out more about what we're building at Syndicate that Room, it's www.syndicate room.com or follow me, Tom Britton, on LinkedIn. Until the next episode. I hope you have a wonderful rest of your day.
37:17Unidentified speaker (UNKNOWN): Bye.