Xylo arrives to rescue the UK from its planning nightmares

0:00Dermot O’Riordan:Welcome to PathFounders, where we use journalism, not influence-ism, to unpack the code of tech entrepreneurs, the capital that backs them, and the consequences. Before we get into today's podcast, I'd like to preview the upcoming launch of the new PathFounders product called Network. Europe has a problem. Between performative LinkedIn and X posts, the real work of building tech companies has to be done. Who can founders really trust in such a chaotic world? Well, PathFounders is poised to address that problem. In 2026, we'll be launching a new and innovative service. PathFounders Network is the forthcoming curated membership platform built for trusted peer connection insight and signal over noise network is being built as an extension of path founders editorial ethos and will remain independent of editorial operations but seek to provide genuine service to founders navigating the choppy waters of building companies fundraising hiring and keeping up with the fast-paced changing world of ai and tech you can apply to join path Founders Network by going to pathfounders.com slash network.

1:18Dermot O’Riordan:So on to today's podcast. Today we're talking about one of the least glamorous but possibly most consequential bottlenecks in Britain, local planning. Across the UK, planning departments are under pressure. They have overloaded teams, long determination times, mountains of consultation, responses to deal with, you name it. And the problem, of course, applies across the whole world. Any country faces this issue. XYLO, we're about to find out how to pronounce that, is a UK startup trying to fix that with purpose built AI for planning offices. And here to talk about it is founder and CEO Habs Kim.

2:00Dermot O’Riordan:Thanks for joining PathFounders.

2:02Habs Kim:Hi, Mike. Thanks for having me. It's pronounced Zylo.

2:05Dermot O’Riordan:Zylo, thanks very much for enlightening us. Now, first of all, I mean, they have systems to deal with planning. There's computer systems that have been in place for many years. So what are you bringing to the party exactly?

2:22Habs Kim:So planning officers historically reviewed everything on paper and they had a secretary called Elaine. Elaine would handle all of their admin. And in the 2000s, as you said, many years ago, the software arrived. Elaine got cut and the officers themselves actually consumed the admin and handled it themselves. But officers now spend half their day actually processing the admin and they themselves call it feeding the beast. So our officers tell us that Zylo is like having Elaine back. So our agents do the legwork, they hand over the first draft report in minutes and the officer edits the report and then makes the determination faster.

2:55Dermot O’Riordan:Right, so it's very purpose built. I think it would be great Great for you to kind of give us a bit more background about how you came up with the idea. You know, why did you engage with this subject in the first place?

3:09Habs Kim:So this is a this is a mission that is really, really close to my heart. So actually from South Korea, when my family first moved to the UK, we tried to set up a takeaway shop. And my parents first interaction of the UK local government was from waiting for planning permission to change of use, which should have taken eight weeks. Instead, it took six months. And I found that was incredibly impactful on me in terms of the impacts that planning delays actually have on very ordinary people and ordinary families. It's always stuck with me. I met my co-founder, Dermot, many years after, after going on 54 co-founder dates in one month.

3:44Habs Kim:I'll tell you about that another time. And we immediately bonded. So he grew up in Dublin during the Celtic Tigers boom and bust. And we were sick and tired of waiting. I will tell you now, Mike, we have absolutely no business doing government. We're just two ordinary citizens that thought this has to be a problem that we could fix. So we spent about 300 hours hunting down and speaking to every planning officer that would speak to us across the UK before even writing a single line of code. And from there, that's how we started to get invited into local government offices, sat next to officers, shadowing them.

4:15Habs Kim:And that's how we uncovered the huge admin problem that is drowning planning officers.

4:20Dermot O’Riordan:Well, it certainly sounds like a worthwhile project and a great backstory as well, I might say. But, you know, one of the questions that anyone would ask you, and especially investors, is to what extent are you building a kind of a wrapper over an AI model? And tell us a little bit more about how it functions in terms of AI.

4:42Habs Kim:Yeah, of course. So we have an AI workspace that supports the full stretch of the planning application review. So we help with the research, all the assembly files and also the map research and then we get to the first version of the report. That's where officers spend the bulk of their time but actually loads of hours are wasted right now with officers assembling the right information before they even decide where the policies are applied or how the site context should be considered in the report. So we use a variety of all the kind of big models but on top of that accuracy is really really important because it is essentially legal AI that we're doing for government.

5:16Habs Kim:And so we apply evals, we also apply a lot of accuracy, measurements of regression testing. And what we do when the final report is actually published, we then ingest that back so that we can compare the version that Zylo produced compared to what the officer actually published in the end.

5:33Dermot O’Riordan:You mentioned, you know, that you're, you know, looking at a lot of the regulations. isn't there a danger that you know we've seen judicial reviews and you know in the court in law the legal sector going over things that are produced by AI are you subject to the same potential problems?

5:54Habs Kim:I mean it is a judicial process and it's a legal process to approve or refuse a planning application so we are under the similar risks but that's why we stop before the decision is actually made. So we help with everything up to the decision and we leave it up to the officer to apply their judgment, do the review and actually make the final decision. We are pretty strong in our opinion. I know everyone is talking about how models should make decisions and AI in government. We're pretty against that. We believe very strongly that the officer is the expert in this case and they need to be the person to actually approve and make the final decision.

6:29Dermot O’Riordan:And you're presumably furnishing them with accurate references, for instance.

6:35Habs Kim:we are we give them suggestions for everything from policies to consultation summaries all the way to certain parts of the appraisal that should be applied but they have full control to edit it

6:47Dermot O’Riordan:as they please i mean um obviously you're doing this in the uk um how old is the company so far

6:54Habs Kim:since when did you launch uh about a year ago now right and do you have any plans to kind of

7:01Dermot O’Riordan:expand internationally?

7:02Habs Kim:Yeah, so we are looking at the US actively right now. So the similar kind of problems apply when permitting in the US. The US is slightly different because everything is broken up into different departments, but the same admin burden and incumbent software kind of slowing everything down in the permitting space is the same in the US.

7:20Dermot O’Riordan:And there must have been some other people looking at this space. What's the sort of competitive environment?

7:26Habs Kim:So there are There are lots of people building, I'd say, AI chatbots and looking at the system facing side of improving customer services when it comes to submitting a planning application. We're the only ones in the UK actually looking at how decisions are currently being made. There are some government projects funded by central government happening right now to basically explore something quite similar to what we've done, but only focusing on household applications, which we're very certain that those projects will soon realise what we learn quite difficult in the process, which is that unless you support every single application type, officers will not adopt your tool because the varied day-to-day of an officer's single caseload is so different.

8:10Habs Kim:So today in Xylo, we have everything from householder extensions all the way to energy sites to, I don't know if you're a football fan, Mike, but the Ellum Road Leeds Stadium as well. Okay.

8:20Dermot O’Riordan:And presumably you think that the 300 hours you spent speaking to planning offices, it may give you a certain kind of moat. I mean, that's a reasonable amount of time for someone else to try and reproduce, I suppose.

8:36Habs Kim:I mean, I think 100%, that's the reason why we were able to actually spend very little time building a product that works on the first launch. So we actually only spent 10 weeks going from zero to enterprise ready. I think the 300 hours really helped us learn. I'd say a lot of the remote calls that we did as soon as we then sat next officers and shadowed them all of those hypotheses were completely flipped and we were pretty aggressive in terms of trying to meet officers where they are they're under incredible amounts of pressure so the very last thing they want to do is to speak to me uh so we turned up unwelcome to planning committees you know we cornered planning officers until they would speak to us it was really really important because we are outsiders to government that we really understood the problem and heard from the people actually making decisions and part of i think why we will definitely win in this space is because we're not just selling a product of government we had a design partnership with lee city council which is similar size to san francisco and because we built it with them for the planning officers it helps us deeply understand the domain and what actually incentivizes them and makes them feel empowered to use modern software let alone ai well it sounds

9:41Dermot O’Riordan:encouraging but um as we all know dealing with government can be quite a slow process especially for tech entrepreneurs. So aren't you facing quite a sort of an uphill struggle to have to bash down doors and get the product into these planning departments?

9:59Habs Kim:Everyone loves talking about how government is slow. I think that is definitely the status quo, but that is not good enough to not be working at a startup speed in government. I mean, Leeds City Council themselves have spoken on multiple talks about how they move uncomfortably fast when working with Xylo. I think we target ICPs where they are the hundred busiest local governments in the UK. They usually have a staffing crisis. They've also already tried some kind of digital transformation or TriLink co-pilot, which allows us to come in. And usually at that point, they're willing to move very, very quickly.

10:31Habs Kim:And it would surprise you how quickly a lot of local governments are able to operate.

10:37Dermot O’Riordan:Right. Now, finally, tell us a little about your funding. I know Common Magic has invested. What's going on with the funding side of things right now?

10:47Habs Kim:Yeah, we've just closed a£2.8 million pre-seed led by Capital T. We also have Common Magic, Shore Valley Ventures, TinyBC. So we have the announcement coming out next week. But really, the raise isn't the story. I'd say this is a strong story because it's all about whether AI belongs anywhere near a public decision. And I think with what we've done in a very short space of the first year we've shown exactly the right way to do it.

11:13Dermot O’Riordan:Well, very interesting to talk to you. Tackling quite an intractable, normally intractable problem there and something that probably a lot of tech companies haven't really thought about. That was CEO Habs Kim from Zylo. Thanks very much for joining us on Path Founders. Thanks, Mike.