How Urban Sports Club Operationalizes AI with Clear Ownership – with Artur Yatsenko, Urban Sports Club

00:00:00: So the platform stability, quality has all to play their role.

00:00:03: But the pinnacle is the trust that you build with the stakeholders and the data consumers And it's very important because we can lose them easily.

00:00:23: Thank you

00:00:25: so much for being here.

00:00:26: I'm very excited to talk about the data at The Data Culture Podcast.

00:00:31: You're with the Urban Sports Club, maybe not everyone is familiar with your business but it can give us a rough overview also of what data do we have?

00:00:39: Yes absolutely!

00:00:40: So, Urban Sports club has an interesting set up.

00:00:44: It's a three-sided marketplace.

00:00:46: We offer one flexible membership for our users and they work out in different sports locations particular type of sports, but it could be anything from going to gym your yoga martial arts studio whatnot.

00:01:00: So that gives the flexibility and there's three-sided marketplace.

00:01:04: why I mentioned that?

00:01:05: because we have... From one side are consumers.

00:01:08: We call them members.

00:01:09: We have the partners who is our studios Who basically supply parts for business.

00:01:16: And also they're corporates that are using Urbans POS Cloud to offer to the employees for well-being generally, so as benefits.

00:01:27: That's our business model.

00:01:29: we unify all of those sites together.

00:01:32: in terms of data is also quite exciting because if you can imagine there a lot behavioral data users produce from interacting with products.

00:01:42: We collect a lot information on behavioural signals our partners, also giving them insights related to the members and how they interact with their particular product.

00:01:55: And same-site corporates as well right?

00:01:58: So have effectively an aggregator form like employees are using.

00:02:03: the offer that is given for them... ...and gives us a lot of really opportunity to work a lot with this.

00:02:10: data building predicting models.. ..building things that can help us predict churn or like grow the member base, generally do a lot of personalization to our members.

00:02:22: Can you give us an impression about size?

00:02:25: So how big is that company?

00:02:27: Yeah so we are at this scale up stage.

00:02:30: so far I mean it's around four hundred people like at all once post-club and We're present in eight different countries in Europe.

00:02:40: And as saying from terms of numbers what i can mention Is we have around thirteen thousand partners that are part of our network, so it's quite flexible for our members to have a great choice Of things they can do.

00:02:53: And out of those four hundred how many are involved with data AI analytics?

00:02:58: Oh That's the good point.

00:02:59: I think i wish there would be more

00:03:02: of course.

00:03:02: uh...I

00:03:03: mean Our tech department is not that huge.

00:03:06: I mean There still some good percentage around like eighty people.

00:03:10: Eighty five people And from the data and AI perspective, there are maybe twenty individuals who work with data on a day-to-day basis.

00:03:21: So yeah I think they're quite efficient in terms of their size and like the outdoor that they produce.

00:03:26: Yeah!

00:03:27: And i know you do a lot in AI.

00:03:29: Maybe if we give us an overview as well The strategy also but also the AI initiatives your running.

00:03:35: Yes so... We've been talking recently how can increase more adoption of AI And I think there's also been the problem for a lot of companies, you know hitting the wall to move from The TOC purgatory towards like what is actually productionalized and we face this same problem.

00:03:53: So with definitely not alone in this.

00:03:55: I can say What?

00:03:56: We tried and what didn't work and would actually worked in the end so far as we We started like doing a lot more innovation hackathons together was the teams solve.

00:04:07: How can we bring some of those topics that could be Really innovative on like any I included and that was really good efforts by saying our challenge Was we asked people to innovate every six weeks continuously, okay?

00:04:20: And you can't reinforce the innovation to be on a continuous basis.

00:04:24: um at The same time.

00:04:26: We also try it.

00:04:27: approaches like AI governance board and things like Assessing every initiative from the AI well Every.

00:04:35: I initially from my perspective of the business impact and what kind of In fact, that would have on the business in a long term.

00:04:42: But I realized we introduced more complexity bureaucratically into this.

00:04:47: so what really worked for us is basically integrating within our existing processes and these are the objective key results which lot of companies adopt to want some hall-key basis.

00:04:57: We have initiatives every two weeks All of the iProtest initiatives that we plan live.

00:05:05: they go through validation, business impact then assessment.

00:05:09: Who is building this, who has the clear ownership on the end product?

00:05:13: Who's gonna maintain that.

00:05:14: And most importantly was impact of their business in terms of metrics and revenue and what not?

00:05:20: okay.

00:05:22: so What maybe had the biggest impact In terms of AI adoption as you mentioned like they hackathons didn't really work if your do them too often.

00:05:31: Yeah You mentioned linking to a business ochre airs which makes a lot sense.

00:05:38: So may be What do you think drives adoption the most?

00:05:42: Yes, I think it's an interesting question.

00:05:43: And generally if we open up AI to everyone who is in organization and without controls, adoption could be a hundred percent but can definitely harm us in the long term.

00:05:55: so really need to think about this sustainably from perspective of business impact that we calculate or give examples.

00:06:01: maybe two projects A lot of companies have challenged how to really measure the impact of AI within their business.

00:06:09: What worked for us was automating some of the flows.

00:06:14: that will take, let's say a lot effort from human standpoints and... Let's say we've got this project which is testing right now to optimize content in our partner network.

00:06:25: As I mentioned there are about thirteen thousand partners.

00:06:27: not all of them are uniformed.

00:06:30: We want it to be according to guidelines.

00:06:34: So we use some LLAMPs and workflows to validate the content, find information that is relevant for our members in terms of amenities.

00:06:42: And what they can expect when you go into a physical studio.

00:06:46: Now there's this team onboarding on-board partners who might take them forty minutes to find all their information online, research the opening hours input it into text validates translate different languages and post it within our platform.

00:07:02: But what we are testing right now is AI workflows, which also takes human in a loop to validate this content externally.

00:07:11: Infuse it with our database information about the partners and all of their loops and system.

00:07:18: take seven minutes out And I think that's very clear signal for business.

00:07:22: What used to take forty-minutes like for professional now four or seven minute time.

00:07:29: It definitely improvement.

00:07:29: you can multiply really turn into monetary value, what does it mean from the business in terms of productivity?

00:07:37: So that's one.

00:07:39: That's a clearly measured productivity gain just on time but... Yes!

00:07:45: Do you mention two projects?

00:07:47: What is the second one?

00:07:49: The second one I'm proud to say was able build this recommendation engine system.

00:07:54: we took a course within our product strategy for personalization And one thing that we built was the recommendation engine, which offers our members a mix of things that are relevant to their particular workout strategy or generally what would they like but also balances this out in terms of user's sports cost and having business engines on top.

00:08:20: So it is an example not in productivity game But you can measure its impact on your member base.

00:08:27: The challenge is, of course if you are an established member and having your routine there's less likely for you to be swayed in completely change your sports routine.

00:08:40: If I'm going through this one gym i am not gonna go into a different one that has maybe fifty percent or fifty cents cheaper.

00:08:47: but we'll find out how we can impact the users help them navigate as they register within our offer discovering and building those habits.

00:08:58: And it's way easier to show them, you know what might be relevant for them or they don't know where to start.

00:09:03: so we all have some good impact that can actually take from these examples as well in a lot of the business metrics too.

00:09:12: Okay

00:09:14: do also I have big topic around getting data ready for something?

00:09:21: Absolutely!

00:09:22: If we are not able to get the data ready, like a lot of our projects could fail.

00:09:26: And that was one example where we also tried to build our AI agents for analytics.

00:09:34: they did failed because adoption wasn't great but also We had to connect every new dataset that would be curated with more context To enable an agent run proper SQL queries in this data.

00:09:50: It is very important like to have really clear metadata, clear lineage and validated information.

00:09:57: To factually plug in AI out of the box.

00:10:01: Regardless of AI.

00:10:01: I think this still is The objective over data team and data platform as a general because you have to run the business.

00:10:09: so Having these like for AI it's just that is a bonus anyway.

00:10:13: But this is a core objective off-the-data Team to have data an accurate assets or together.

00:10:19: That

00:10:19: makes no sense.

00:10:20: If you recommend people running data teams or involved with that.

00:10:25: to focus on one topic, what would it be?

00:10:27: What has the biggest impact in terms of data?

00:10:30: because... You mentioned a lot about things which all have an impact like quality and lineage.

00:10:37: From your experience this is the most important thing for me.

00:10:40: It's

00:10:40: very challenging I think!

00:10:45: There are technical things but there also institutional things And I think like as a data leader talking from the data leadership perspective is definitely convincing.

00:10:54: You know people that data could be trusted and we build those trust gradually, right?

00:10:59: It is easy to lose this trust overnight But it's very hard to bill.

00:11:03: This he takes a lot of time.

00:11:04: if your pipelines are not ready before ten AM you Know on like five consecutive days People will stop looking into those dashboards and they're right.

00:11:14: so in the platform stability Quality has all to play the role, but the pinnacle is The trust that you built with the stakeholders and data consumers.

00:11:24: And it's very important because You can lose it very easily.

00:11:27: Okay therefore here right?

00:11:28: That contains let say process reliability Yeah.

00:11:32: So being able To provide the data But also the content yeah quality of the condors okay Very good.

00:11:39: Let's get back to AI and You mentioned your team.

00:11:43: you mentioned some use cases.

00:11:47: Let's also think a little bit about data culture.

00:11:49: What were in your team maybe the biggest shifts and changes you have observed?

00:11:55: Yes, I think data culture is an extremely important topic.

00:12:00: Something that cannot be metadata from the top but it often something comes to bottom up You can adopt like a data policy or anything they will try to govern And create their kind of culture.

00:12:15: But the only thing you could do, let's say is management.

00:12:17: You can only lead by example to establish a good data culture and at same time we have really to rely on a lot of individuals who are embodying this culture.

00:12:27: And what I mean about that particularly people who are curious to explore no tools.

00:12:33: Who're curious ask questions why?

00:12:35: Who were curious like validates Data makes things better within your systems generally.

00:12:41: So that as for me it was for example the great data culture, but it starts with individuals.

00:12:46: It starts with teams and I don't know ability to learn from them using educational budgets And I'm not saying an amount of trainings can really change that because training is something new.

00:13:05: people will come.

00:13:06: then you have to do all over again.

00:13:08: That's a bit of challenge.

00:13:09: So yeah start from bottom up.

00:13:12: But still, you also said leadership is important.

00:13:15: And when you meant leading by example... So how does that go together?

00:13:22: You say it's bottom-ups and bottom-up.

00:13:23: I agree!

00:13:24: Uh..You said you cannot mandate it.

00:13:26: i also agrees but still leadership isn't important.

00:13:28: so How can leaders encourage or influence the data culture?

00:13:34: Yes ,I think.

00:13:35: well maybe one example Is like ..i always Like to bring new technology and tooling To discuss with my team.

00:13:41: I mean, let's have Friday off meetings and then we just do a huddle.

00:13:48: And bring new exciting use cases.

00:13:51: We try to fix things that did not work Right?

00:13:53: Then we used this time exclusively for the focus time.

00:13:56: There is no meeting.

00:13:57: It was just like there are one meeting where you discuss what can be done better.

00:14:02: But ultimately... This is an open discussion You know!

00:14:05: A new thing which we can explore What could apply in our setup or architecture.

00:14:11: And as a leader, I feel this is necessary for me to show this example and then kind of come up with something new that's existing.

00:14:21: As the leader, my goal also is to encourage individuals who support them right?

00:14:27: So yeah... That's the only thing we can do so far!

00:14:29: Very good You mentioned education.

00:14:33: How did you see these skillsets changing now with AI coming into play?

00:14:39: I see a couple of trends for sure.

00:14:41: A lot of talk is coming about changing or modifying the titles as well, so now we're not talking about software engineering but AI engineer, AI architect and AI product manager right?

00:14:54: Also for data teams.

00:14:56: ultimately writing code easier that's true.

00:15:02: at same time we need to cultivate like ownership and quality, was in the software engineering as a whole.

00:15:09: So there's one thing to write AI-generated code but it is important that you make sure this production ready compliant security oriented first.

00:15:18: Security has the mindset so we'll be able to deploy those.

00:15:21: So I think even moreso important than now for data engineers or professionals.

00:15:36: Also a lot of that is AI generated right now, so yeah.

00:15:41: That's mostly what changes within the space I think.

00:15:46: Do you also experience here in let us say overall population and your company?

00:15:52: The overall employees might impact them for some of their jobs?

00:15:57: I think it's legitimate fear If there are recent news coming from other companies I'm not gonna name here, but in some of US-based companies that are letting people go because of this particular reason.

00:16:09: And i think all in Germany we have a good protection from employees and We need to set it up for the start.

00:16:18: To mention right now amount like any new automated let's say workflows or use of AI should be People out of work Right?

00:16:28: Like ultimately what you shouldn't expect More out of individuals if we take a subset of their you know, what do this from AI?

00:16:35: I think and that's time should be used more creative.

00:16:39: You know sinking an expansion in generally improving technology and the stack and architecture whatever.

00:16:44: We have it right?

00:16:46: So again nothing ever will have there own concerns And we cannot of course You know fix them with the magic once.

00:16:53: The only thing we can do as leaders is address those fears address those concerns and then basically mention Well, I mean like you know we're not gonna do any shortages because purely of the AI taking a subset off your work.

00:17:08: but as let's say an engineer that they are professionally having.

00:17:12: AI is their companion and then working with them... You have definitely more competitive advantage now on the market And generally in workplace.

00:17:19: so i think people should act out of fear But actually they should act.

00:17:23: our curiosity In such

00:17:26: The next other sense Maybe look into the future to end this discussion.

00:17:32: What changes can you foresee?

00:17:34: Or what interesting projects are your working on, so... ...what will happen let's say in the next twelve or twenty-four months

00:17:40: For us.

00:17:41: I think we're also looking to expand a lot of our offering at Audubon Sports Club and then as i mentioned like looking into integrating more partners that would be available for members.

00:17:53: So they'll come with challenges the data quality, you know bigger volumes of data that we need to make sure we have good standards for.

00:18:02: And one of the trends that we are looking into is also building more hyper personalization for our members.

00:18:10: and from a product perspective it's looking in communities because we say community as a driver for like good engagement For everybody who wants to do sports.

00:18:20: It's way easier be productive when you have peer pressure or community in general.

00:18:26: So I think that's something will help us and a lot of data products could also based on some of those product strategy features because we really want to see the world where people need active and household life, so this is our mission statement.

00:18:42: Excellent!

00:18:43: Very nice.

00:18:44: good luck with it for all these endeavors.

00:18:49: thankyou very much for your time.