Aaron Phethean (00:01.314)
Hello Phil, welcome to the show. Absolutely delighted to have you on. You're just leaving the world of corporate, head of data and entering the brave new domain of doing your own thing and finding your own clients. I'm super excited to hear about what you've done in the past and kind of what the future looks like. So welcome, thanks for coming on.

Phil Thirlwell (00:26.008)
Thanks for having me. Pleasure to be here. Yeah, it's just such an exciting time right now. With data and AI, really disrupting every industry, but at the same time, the opportunity is massive, Yeah, I've worked sort of the last nine years in FinTech, but...

Aaron Phethean (00:29.173)
That's cool, Lee.

Phil Thirlwell (00:49.028)
I've made the jump to interim and fractional. The reason there really is that most companies don't need a full-time CDAO on the payroll. They need specialist skills on demand and someone who can meet them where they are, solve the problems, and then without the overhead really of a perm hire.

Aaron Phethean (00:59.736)
That's right, yeah.

Aaron Phethean (01:12.814)
Yeah, exactly. Like you said, there are tons of companies at a size where having someone very experienced on the team all the time doesn't really make sense. That's going to be awesome to dive into that. And there's obviously a bit of a new experience and some people have considered it for themselves. So I'm quite keen to dig into that. think when we were chatting before in a couple of our pre-calls, we're thinking actually about what are the kind of main challenges for companies.

And so I think we'll dig into that forever. And this thing that kind of be at the first stop. Second area that we're kind of digging into is kind of technology and platforming and how people spend their time. And then I think there's a sort of still relevant AI question and what's happening that we intend to discuss. So let's dive into the kind of corporate world and what companies need. Philip, if you could give us your opinion on what are kind of typical mid-sized

company that perhaps doesn't get the most out of their data team looks like and what they could be doing better.

Phil Thirlwell (02:21.144)
Yeah, I think, I mean, it really comes back to getting the value of the data. Ultimately, I think our job as data leaders is getting the right information to the right people or process these days at the right time, right? To make a decision that's going to impact the business. That's ultimately what it boils down to. But, you know, there's a number of obstacles that get in the way of that.

And I guess that's kind of what we want to move on to as our first topic.

Aaron Phethean (02:57.23)
Yeah, and I think you're right. Like the, seems like such a simple task. I met with someone yesterday, actually, I was trying to sort of explain our proposition. And the sort of question came up, you know, they were very much on the analytics side of things. I was like, well, why is this even a challenge? Like getting the data in the right place to understand what's going on. It can feel very far away from people. So in your experience,

Why is it a challenge to bring data together and do these sort of you know, what what you know this day and age should be a solve a problem?

Phil Thirlwell (03:39.021)
Yeah, I think there's a few things. I think historically in data, we haven't done a great job of understanding what the outcome is, right? There's been a lot of investment to building pretty data models, funneling loads of data into a data lake, but not really thinking about,

what's the problem space here or where's the opportunity that we're gonna drive with that data? And in many ways, I think in data, kind of, we've followed suit of software engineering. Software engineering, I think is kind of set the standards with DevSecOps and agile ways of working. I think we've always been slightly behind that curve from a data perspective, if you know what I mean.

Aaron Phethean (04:09.709)
Mm-hmm.

Aaron Phethean (04:32.384)
Yeah, I agree. I definitely agree with that perspective.

Phil Thirlwell (04:36.392)
We've seen it, know, it's in data ops, ML ops, that's, you know, that came afterwards. And I still think a lot of teams aren't really putting it into practice in the best way. So it's really reframing what you think when you deliver a data project, right? Because often they, it's not clear what the problem is to begin with. You get a question,

or you get a problem statement that isn't, it's really a symptom rather than the problem. But taking that kind of structured approach, right, and really thinking, okay, if I'm gonna deliver a dashboard, deliver a model, then treating that as a product rather than kind of a

Aaron Phethean (05:10.83)
Hmm.

Phil Thirlwell (05:28.1)
projects that's unbounded, I think leads to better outcomes when it comes to delivering data.

Aaron Phethean (05:37.017)
That you've touched on quite a few things there that kind of need to think about the organization upfront. Definitely see as a positive for lots of organizations and that kind of product and scoping that you're just discussing again to me is like feels incredibly valuable because you're not responding anymore. You're kind of framing and figuring out what the organization needs with a lot more focus.

And that's relatively recent really. I wonder if like what advice you'd give people to take on that approach. Like how could they go from the kind of person turning up at the desk asking for a dashboard to a more structured conversation around what is the product for the organization or products for the organization.

Phil Thirlwell (06:28.516)
Yeah, it's a good question. You still see it a lot, right? That happens. it feels good to be answering those questions, because it feels like you're giving it immediate value, right? But you're not really producing scalable solutions that are maybe getting to the heart of the problem. For me, it's really around how you structure the

Aaron Phethean (06:44.973)
Yep.

Phil Thirlwell (06:56.472)
the data capability in terms of your operating model. I think it's a flip in that operating model in a flippant thinking. think what's worked for me in previous leading data and AI in FinTech is taking the lead from the software engineering business and looking at an agile delivery approach. So rather than potentially having

Aaron Phethean (07:21.678)
Right, yeah.

Phil Thirlwell (07:24.982)
a Kanban board where you just you kind of work and stuff through and okay you you may be limiting your web so you you don't have too much on but you you're essentially just whoever shouts loudest they get the next piece it's it's really thinking okay what we're going to deliver in the next you know three months and let's break that down into deliverables to it so you can kind of continuously be showing impact and it forces I think you to think things through

before you get started and work more closely with stakeholders on getting to the heart of the outcomes that they need. you can deliver things with less rework, I guess.

Aaron Phethean (08:02.933)
Mm-mm.

Aaron Phethean (08:07.541)
Yeah, I think that's that's a mistake of seeing time and again is that yeah, the adoption of an agile approach doesn't mean there's no plan or no overall direction. Yeah, in fact, it just helps you choose much more robustly what it is important to work on. And you kind of get away from that. Yeah, I think some people describe it as firefighting, but actually, it can feel quite good, like you said, to solve someone's problem.

Phil Thirlwell (08:35.352)
Yeah.

Aaron Phethean (08:35.692)
We get drawn to like fixing small things is actually quite easy and a bit of a dope I mean here at the same time

Phil Thirlwell (08:40.684)
Yeah, it is. That's exactly what it is.

Aaron Phethean (08:44.095)
Yeah, so that's a really, really good point of view. So then if you if you take the organizational life, so that's looked out a little bit as the data team, if you take the organizational point of view for a second, what do you think the organization wants from their data team that they're perhaps not getting?

Phil Thirlwell (09:03.428)
What they want and what they need is not always the same thing, In many ways, they want that, right? They want the instant response. They just want the answer. It's often the case, right? But what they need from the data team is really a partner that will challenge them and, you know, kind of help them get to the root cause and

Aaron Phethean (09:18.636)
Mm-mm.

Phil Thirlwell (09:31.844)
and get to the decisions that matter for the business. think those are two different things. But there are a level of standards that stakeholders need. They need to be able to trust in the data. I as soon as you lose that, you've lost. I guess you've lost the game, right? It's game over. And they need

Aaron Phethean (09:35.981)
Mmm.

Aaron Phethean (09:56.62)
Yeah.

Phil Thirlwell (10:01.732)
to be able to understand the products and use the products. And I think that comes to literacy. Not everyone's going to be able to interpret the same sort of outcomes, right? If it's going to a data science team, they want to get really well labeled sort of wide data set. But maybe you finance a HR, finance maybe they want some tabular data, HR, they just want some graphs that kind of

Aaron Phethean (10:17.441)
Yeah.

Phil Thirlwell (10:30.66)
tell them where they need to focus and being able to kind of understand that and deliver outcomes that.

Aaron Phethean (10:38.797)
you

Phil Thirlwell (10:44.812)
your stakeholders can actually use and make decisions off is a really important part of it.

Aaron Phethean (10:49.869)
Yeah, I really like that link to the kind of product view of the world and you know actually the the product might not be the same for everyone Yeah, thinking of data and where it comes from and it's freshness is certainly one aspect But you know like different teams and they said might need something more visual might need something more This is directional other teams might need something much more highly accurate and then you know feeds into their work in some way So yeah, that actually you're thinking about the people and the product is a you know, definitely definitely can see that

One of things that we chatted about actually was that, you know, it still seems a problem that organizations have gazillions of dashboards and that seems to be the main output from the data team. I wonder in your experience, how do you solve that? And do companies need hundreds of dashboards? What does kind of good look like in that kind of more visual space?

Phil Thirlwell (11:46.009)
Yeah, I think it's definitely more or less when it comes to dashboards and even KPIs, they cease to be key performance indicators if you've got too many of them, right? But I think particularly in larger businesses, that kind of, it does, you do get that sprawl and that's something that should be addressed in most cases. It's, I think, you

Aaron Phethean (11:54.573)
Hmm.

Aaron Phethean (11:58.382)
Yeah.

Phil Thirlwell (12:15.652)
when at one point at FIS there was 600 on Power BI dashboards. You know for a fact that they're not all getting updated, probably some of the people that have built them no longer with the business, but they're still out there. And people who've got the link can get that data, maybe misuse that data if it's not accurate and reliable. think having an end of life to

Aaron Phethean (12:30.697)
I'm done.

Phil Thirlwell (12:43.94)
to a dashboard is back to that product ideas. That's really important. And if a new product gets created that does the job or overlaps with another one, then it's a decision, is this still required? Is this still useful? I think one of the mistakes that I've made in the past is just pulling the plugs without looking at the user base though, because you might think, oh, well, this dashboard's over.

Aaron Phethean (12:49.685)
Yeah.

Aaron Phethean (13:09.837)
I'm sorry.

Phil Thirlwell (13:13.708)
If you're not monitoring who's actually looking at it, you may have hundreds of users still who are getting value from it. And OK, you've launched a new dashboard that does a job in a better way. But you've got to communicate that out, right? So that's an important piece of it.

Aaron Phethean (13:29.389)
Yeah, that sounds like a painful lesson learned and probably a good bit of advice for a friend on this thing. You have the like the product, you know, end of life concept has applied to that as well. And, know, I think that change is always really hard, like for anyone, they necessarily want it and they don't necessarily get that the new and better version is new and better for them. Yeah, I think there's a

quite a few aspects there that feel quite challenging for data teams. One thing I'd like to revisit is the kind of, I think you mentioned it briefly, the kind of consulting or trusted advisor or, so if you're heading up a data team, you've got to build a relationship with the organization to establish those KPIs. And I think I see quite often they would like that to be a kind of two-way conversation. Like, what are...

key, what are the metrics that are key to indicators for us? And perhaps the organization as a whole is not as literate about what's going on for them. I wonder where you could advise people to start. It seems like the dashboard sprawl is happening and they're lacking KPIs. Where can they start to establish what's important?

Phil Thirlwell (14:49.59)
Yeah, I think it's got to be, you've got to co-develop these things. If I look back to when I started at FIS, the challenge was around engineering metrics and having engineering metrics that made sense to the board, but also could be that made sense to the scrum teams and development teams on the ground. So that kind of

thread runs all the way from top to bottom. And really, there's obviously industry standards, so that's a good place to start, but every organization is different. And what we found really got this sort of engagement, it got people engaged in actually using the dashboards, using the KPIs to drive decisions was kind of when it came to the definition of

Aaron Phethean (15:31.507)
No.

Phil Thirlwell (15:45.985)
of how it would be calculated at a technical level from the tool, involving the development teams, the development leaders in that process. it was kind of co-owned by them. And I think why that was so successful was because by working with them, obviously, they felt invested. They felt they were part of the process. But also,

Aaron Phethean (16:11.127)
Yeah.

Phil Thirlwell (16:12.152)
from a governance point of view, they knew how it was calculated in the tool. So they knew we have to update this, you know, when we go through this part of the process. And they knew which fields mattered. And at that time we were using Raleigh. They knew as they were going through their development process, what they should be updating, what they should be filling out. And that just obviously just gets a, made sure the data quality was in a good spot. the KPIs that were coming out of it were.

even more useful. it was like, yeah.

Aaron Phethean (16:42.589)
I do see that as a pretty big part of defining that product or what an organization needs to do better. I think there's a lot of frustration I see in people are inputting things or creating things, but not necessarily recognizing that they are creating a data point. are creating a piece of information that's going to feed into some decision or something that's important.

You can get into these situations where there's capturing loads of things and most of them don't feed anything anymore. You know that why they want to catch you then capturing that rather. And so actually, if I think back to one of my first jobs, it worked in a supermarket and you that you're capturing things it feels like for the sake of it. And actually, as time goes on, I'm sort of more and more convinced.

that I was, you know, was literally just doing these things because it seemed like a good thing to measure once and it's just, just not anymore. So yeah, actually, you know, looking at that whole process with the individuals who actually create the data sounds, it sounds like sound advice. The other thing you mentioned was kind of the tools. I was sorry, go on, go on you.

Phil Thirlwell (17:35.329)
Thank you.

Phil Thirlwell (17:51.419)
Yeah, I think...

No, I was going to say like on that point, know, there's so many systems and so many fields and custom fields that you can have. Like you can't govern everything. So really, you should be focusing on the ones that matter, right? You don't need to worry about having every field inputted to a certain level of quality. It's just which ones are most important. And let's focus on those ones.

Aaron Phethean (18:09.058)
Yeah.

Aaron Phethean (18:22.187)
Mm-hmm.

And they probably turn into your tests and the ones that you monitor more closely and you're checking the values of, whereas all those dozens of other custom fields that capture something for a particular project maybe, don't worry about it.

Phil Thirlwell (18:39.734)
Yeah, yeah, don't worry about it. If they want to use, if that works for that Scrum team, great. But, you know, we're not going to, we're not going to try and set a standard across the enterprise.

Aaron Phethean (18:51.021)
And that's, I suppose that's the big difference. A data team for the enterprise is really about the whole picture. Whereas a team that needs an operational report and needs to track something for themselves and their own productivity. Well, that's great too. They can do that.

Phil Thirlwell (19:08.792)
Yeah, no, absolutely. Both valid.

Aaron Phethean (19:12.301)
So then, you touched on tools briefly and I suppose, yes, people see dashboards or they see the product that the dashboard is in, but there's a ton of stuff that they don't see. And one of the things that the experiences you had recently that we were thinking might be quite good to talk about is kind of Microsoft tool set and it's sort of suitability for an organization versus other tools and.

I think there's two big bits I'd love to touch on. It's like, do organizations decide what fits for them? And then also, what does good look like in the technology at the moment? Where are we headed with that? So, let me start with the first one. What, when you were experienced with fabric, why was it a good fit for you at FIS and the team there?

Phil Thirlwell (20:07.107)
Yeah, I think certainly for FIS Fabric made a great deal of sense. Already the enterprise was all in on Microsoft. you know, people knew Power BI, people already working in Azure, right, Synapse, et cetera. Fabric just kind of brought everything together in a kind of one-stop shop in many ways under the one leg. So,

Aaron Phethean (20:17.62)
Yeah.

Phil Thirlwell (20:36.695)
But for an enterprise like that, I think it's a really good option. Everything's covered. You've got that one relationship with Microsoft. And that can be a good way to go. There's other options, though, right? Depending on the business, lot of businesses may be.

Aaron Phethean (20:48.29)
Hmm.

Phil Thirlwell (21:02.153)
know, or GCP, like on the Google stack, it's less integrated than fabric, but they've pretty much got everything there if you want to. But, you know, what I see a lot of peers doing and a lot of maybe scale-up businesses, smaller businesses doing is kind of picking and choosing their vendors. And that can make sense there as well. It really depends on the context of the business. So, you you maybe you...

Aaron Phethean (21:21.549)
Yeah.

Phil Thirlwell (21:30.243)
you're choosing Databricks, maybe you're choosing Madotika for your ETL, and maybe using Tableau for your dashboards. That's a valid option. And probably if you're a smaller company, I think it gives you a little bit more control. And it allows you to select vendors that are going to really get to know your business and build a relationship. Whereas

Aaron Phethean (21:47.724)
Yeah.

Phil Thirlwell (21:57.988)
know, the Microsofts and Googles of the world, probably, you know, they're probably looking at the Fortune 100 and spending most of their effort there.

Aaron Phethean (22:04.723)
Yeah, they want to focus on the big case studies and the big extreme uses of their product. And I definitely would like to think as a vendor that, you know, as a small company, we're really like over delivering on support and engagement and, you know, a certain level, that's like a really, really valuable to a company. And yeah, the other thing I

Phil Thirlwell (22:09.549)
Yeah.

Aaron Phethean (22:31.181)
kind of fundamentally believe in is the kind of openness of the stack and that kind of technology choice. I guess the trade off of if I think about the kind of organization that's completely in Microsoft is that, know, in theory it should all work together better. It should all like, you know, kind of be more seamless. You might lack some of the opportunity to tweak and tune, but that can also be its own problem. You know, if the probably another little tangent of that thought is that if I look at too big

cloud data warehouses, Databricks has got lots and lots of tuning opportunities. Snowflake, equally good as a product, equally good in the market, has far fewer tuning opportunities. they're supposed to solve exactly the same problem. So yeah, you definitely, you've got different approaches there that work, whether the organization's engineering heavy or not. And I think, just coming back to Fabric, it's a relatively,

you you adopted at the time when it was a new product and as a new release and that definitely comes with some pain always does. wonder what it was like as the, as the time went on and sort of where you see it now as a product.

Phil Thirlwell (23:33.699)
Yeah, very new.

Phil Thirlwell (23:38.787)
Yeah.

Phil Thirlwell (23:45.188)
I'm definitely an advocate when it comes to Fabric. think there's a lot of good features there. It does offer a lot with the One Lake, opportunity to kind of share data products across the enterprise. So you don't have to have everything centralized. You can have different teams that own their own models, own their own data sets, but you can shortcut to them.

and you can control the security at rural level. you can then kind of rather than centralizing the governance, you can kind of have that as a federated governance model where the specialist teams, own and they're responsible for what comes out of it. But they're also the source of truth. There's only one fine answer. There's only one answer. know, one of the problems that we had at FIS was

Aaron Phethean (24:26.55)
Interesting.

Phil Thirlwell (24:44.387)
that we tried to address with Fabric really was like some basic questions were kind of you get different answers you ask how many people work at the business how many people work in this business area and you depending on the definition that finance has, HR has, that the business have you get wildly different answers and that's just you know not really acceptable at an executive level or any level really like you've got a

Aaron Phethean (24:55.703)
Thank you.

Aaron Phethean (25:00.822)
Yeah.

Phil Thirlwell (25:14.551)
be able to define these things and have a unified view of it. Yeah, more so than you think.

Aaron Phethean (25:18.183)
It is actually really challenging though, isn't it? I mean, it sounds maybe if you're exactly if you're someone listening, that's probably not in the data space. You think you know, how hard can it be to like know how many people work in your department? there are so many subtleties, like are they actually in that department now? Or they working in some for someone else?

Phil Thirlwell (25:38.413)
Yeah. Are they part-time? Are they a contractor? You know, like, all of these just layers of complexity that you add into it.

Aaron Phethean (25:42.559)
Yeah, yeah.

Aaron Phethean (25:48.492)
Yeah, definitely. that's, you know, I would say people and definition of roles is probably far, far simpler than revenue and products. And, you know, when, when revenue is recognizable, he's kind of like, it is actually quite challenging in any piece of data to really define it, but important and important work for a team. The other thing that probably second part is that we were telling you earlier about the

Platform itself and the working on the platform. Yeah, that's also not the end outcome, but it can demand a lot of time You know, you're trying to get a benefit as a data team and you can spend all this energy migrating and getting to a technology I wonder what advice you'd have for people who are perhaps facing a legacy you know a bunch of technology and need to move on for a reason or Find themselves perhaps spending too much time on the technology stack. What what?

What do you think teams should be doing in that area?

Phil Thirlwell (26:50.923)
Yeah, I think, you know, it's a mistake that has been made over and over again, I think with when it came to data warehousing, creating large enterprise data platforms without really connecting with the business. Just, okay, let's structure the data in a way that makes sense from an engineering point of view, without then doing all of that work upfront to ingest and transform the data.

before you then engage a business say, right, okay, how are we gonna use it? think it's asking the question of, if you are looking to change platform or modernize the platform, it's what problem are you solving? Why are you doing it? And if the answers are because our competitors are doing it, that's not a good answer, right? It has to be relevant to...

Aaron Phethean (27:24.449)
What do you

Aaron Phethean (27:44.77)
Yeah.

Phil Thirlwell (27:47.03)
a problem or an opportunity potentially that you want to exploit within the business context. It's not just because that's where the industry is going. It might still be the right decision, but if you don't understand that, why you're doing it from the outset, you're not going to get the commitment and you're probably not going to make the right architectural decisions as you go through that migration process.

Aaron Phethean (28:03.629)
Hmm.

Aaron Phethean (28:13.205)
Yeah, I definitely see that. Where I see projects that have a really clear pain, perhaps a pain, or really clear goal about what they want to do and that they can't do now, I kind of know that it's going to work, you know, and that you've got that focus on what is it that we're doing this for. And then you get some measurables at the end, you know, we achieved it or we didn't, might be quite binary, or it might be like we exceeded it or we've saved a bunch of time and

There definitely needs to be reasons to do these things and the more specific you can get, the better I see the outcomes.

Phil Thirlwell (28:50.407)
I yeah, the ROI kind of comes to mind. I think it's a difficult one, but it's important, right? And I think having a clear idea how you're going to measure that comes with having a clear idea of where you want to get to. But it's not just about the cost, right? It's not just about savings. There's two components, right?

Aaron Phethean (29:14.669)
Mm.

Phil Thirlwell (29:18.591)
Hopefully there may be some cost. There may actually not be a cost benefit, there might be, particularly if you've got a lot of bloat. But what is it that by building this new platform, migrating to this new platform, what's the additional value that you can create on top of it? And that's part of the equation as well, right?

Aaron Phethean (29:36.567)
Bye.

Aaron Phethean (29:42.157)
100 %

Phil Thirlwell (29:42.82)
That's harder to quantify and it's really hard to estimate in advance. But I think there's ways of doing it. Back to productize the data. To really be productized, you need to have a measure of success. And it's not always immediately like a financial one. It might be increasing NPS. It might be reducing time to market.

Aaron Phethean (30:02.955)
Mm-hmm.

Phil Thirlwell (30:11.267)
What I've sort of found is if you can, got to link that to, know, annual return. Let's just cut that, sorry. Yeah, yeah. I was trying to get to annual return revenue, but I was stumbling on my words.

Aaron Phethean (30:28.322)
We'll cut this one.

Phil Thirlwell (30:38.517)
I did want to make the point on ROI though. Where do we go back to?

Aaron Phethean (30:43.757)
You can just start from, you know, if you keep the focus on ARR or annual recurring revenue or start from, you know, you point about ROI and then I'll just find a good way to snip that in.

Phil Thirlwell (30:56.629)
Okay, I'll go again. So yeah, think ROI is, it's really important to have that in mind when you do any sort of make any platform decisions or going through any transformations. it's, it is tricky. And it's something that I think we struggle with in the industry. It's not just about the cost savings, you know, that might be an aspect, particularly in a

business that has kind of a lot of bloat, but it might actually, there might actually not be any cost savings in it. Yeah.

Aaron Phethean (31:28.62)
Mm.

There might be things that you can't measure so well that are definitely worth doing.

Phil Thirlwell (31:36.855)
Yeah, I think the other side of the equation that we often miss is what is the value that we're going to create from going through this migration or having this new, you know, fabric platform, for example. What do we enable and how do we measure that?

Aaron Phethean (31:44.415)
Mm-mm.

Aaron Phethean (31:56.406)
Yeah, let enabling for me is like my past binary things. It might be intangible, but you can either do it or you can't. And perhaps that that is a kind of challenge that a lot of people are facing now with AI and the demand or the the opportunity of AI. Like there are some foundational things that's probably been said quite a lot that of course you need your data available, that knowledge available to be used.

And one of the things I think that you were chatting about and sort of observed is that actually the interface to the data or the products, they might be changing and not optimized yet for the way AI will do it. And I wonder if you could share some of your thoughts on that. What do you think products look like or data looks like when everything's AI led?

Phil Thirlwell (32:50.423)
Yeah, I think, I mean, it's difficult to know, right? We've got to put our future looking hats on here, but I think we're kind of seeing software engineering leading the way here. A lot of the disruption that we're seeing is on the development side. So there's products like AMP, Cursor, even GitHub Copilot. mean, that's kind of ancient now, but you've got

Aaron Phethean (33:16.63)
Yeah, yeah.

Phil Thirlwell (33:18.721)
basically software engineers disrupting themselves. And as part of that, what they're doing is changing the way that their workflow, they're changing the way that they're structuring their code base so that the AI can better support that and it can do more. I think for a genetic AI in it, that's kind of the buzzword of the year, but for that to kind of

Aaron Phethean (33:36.802)
Yeah.

Aaron Phethean (33:45.121)
Mm-hmm.

Phil Thirlwell (33:48.965)
really come in and play a huge part in other domains, I think it's a case of rethinking the way that things are done. Just like the engineers are doing now, they're changing how they're working and their whole workflow. If going to do it in ops, if we're going to do it in finance, then it's the same thing.

Aaron Phethean (33:59.618)
Yeah, yeah.

Aaron Phethean (34:06.697)
in response to how they wanted to be. Yeah.

Phil Thirlwell (34:17.644)
the current workflow requires these different tools. That's great for a human, but that's difficult, right? Even though we have MCP now where we can connect agents to all these tools, that might not be optimal for a way that an agent would do it. And I think that the trade-off really then becomes, if you make it so optimized for AI, can we still...

Aaron Phethean (34:25.594)
Hmm.

Aaron Phethean (34:33.911)
Yeah, yeah, exactly.

Phil Thirlwell (34:44.356)
can we still see the process and understand the process in a way that it's auditable so we can keep an eye on it.

Aaron Phethean (34:48.748)
Yeah.

Definitely agree with that. It strikes me that one of the reasons that software development is iterating fast is because there a lot of people who are motivated to make their jobs simpler and understand it. So you've got a of a coupling of the two things. I think that's the point you're making that if you're to optimize operations, you sort of have that same problem that has been with us in software engineering or IT for a very long time that you might have people understand what the process is.

Phil Thirlwell (35:02.019)
Yes.

Aaron Phethean (35:17.953)
but don't necessarily understand the art of the possible of the technology to achieve process optimization. And AI to a degree is lowering that barrier and being more connected, but you still need people who are going to focus on doing that and optimizing it for their business, which I think is quite insightful.

Phil Thirlwell (35:36.184)
Yeah, yeah, definitely.

Aaron Phethean (35:38.35)
So then on the kind of the what if with AI and you know, perhaps where we end up with it. Yeah, you had some thoughts that, you know, the product interfaces will change the way people work will change. And when you mentioned that, you know, actually it might not be audible and visible what's happening anymore.

I suppose like one of my key things about technology has always been that it should there should be code like it should be coded behind it. There should be progression from user interface to code. And I feel like that is even more true in this space. If you're going to have an AI optimize something, there needs to be a representation of what it's done that you can understand.

And so these kind of like low code interfaces, I find them quite troubling that that's almost stopping the AI innovation, whereas if you had the code and the user interface, or you had the code and the AI, and I kind of like, see that in data and in business and maybe that becomes more and more code based is sort of one of my thoughts. I wonder if you have a sort of perception of that yourself, like what do you think?

AI would prefer or you know for an organization to describe its data. You mentioned a couple of tools earlier as well like the way ThoughtSpot manages things and the way they kind of.

Phil Thirlwell (37:06.965)
Yeah, I think it's kind of exciting, I think, particularly in specifically in the data analytics space, what ThoughtSpot are kind of aiming to achieve with it. It's going back to the idea of having too many dashboards. I think the future looks like less dashboards, certainly.

But I think even less reliance on dashboards that you now, particularly for a power user, maybe they've got a handful of dashboards with a whole lot of filters and they've got to turn all these dials to answer the questions that they have in their head. Having an interface that's more conversational, I think is kind of where we're going, where the AI understands

Aaron Phethean (37:45.005)
Hmm.

Aaron Phethean (37:54.177)
Yeah.

Phil Thirlwell (37:59.524)
The context, and I think the context is key, right? It understands this person, what their role is, and then obviously they can input a question, and it can produce a graph or a table or whatever makes sense to answer that specific question without the need for the user to go and change a bunch of filters on the dashboard. But, you

Aaron Phethean (38:02.978)
Yeah.

Aaron Phethean (38:22.923)
Yeah, exactly. More, time in the product. And actually we mentioned ThoughtSpot, but I recorded the episode two with Ollie from Count. And I really like what they're doing in their kind of BI space and thinking about the way AI integrates with it. And actually, you had a classic quote that, know, it's probably the product with the least ROI because it is so much energy and effort goes into that final mile, but that's probably not.

where the most value is generated. it's really interesting point.

Phil Thirlwell (38:54.315)
Yeah, think, I mean, there's a bunch of cool products there in use cases, but it really comes down to having the foundations, right? You've got to, for those models to get the right context and to get the right answer, you need the data model correctly. You need the right metadata. You need to really ensure that that's in a real clean, hygienic place before

You can layer those things on top and expect any good sort of output from it.

Aaron Phethean (39:27.179)
Yeah. Yeah. Well, I think that probably that's a good point to, to wrap things up and yeah, you're probably picking one of the most exciting times to enter this kind of fractional space and this, this opportunity to help lots and lots of teams. I wonder then if you'd, you'd like to share a few parting thoughts on, on what you think teams could be doing differently and how you think their focus could be over the next 12 months.

Maybe a little opportunity to sort of pitch how you might help someone if they were to introduce you into the business.

Phil Thirlwell (40:02.069)
Yeah, I think maybe just a kind of thought for leaders. It's reframing things, right? It's not which tool to buy or it's not to throw AI at the business. Do you know what decisions that you need to make to move the business forward? And let's build a solution with the technology, with the data that's going to help you deliver that.

Aaron Phethean (40:31.149)
Yeah, I'll guess that's a sound advice probably for the ages, not just even the next 12 months. So thanks, Phil. Thanks for coming on. I really, really enjoyed the conversation and touched on quite a few things that I think will help a lot of people there. So thanks for joining us.

Phil Thirlwell (40:46.733)
Thanks for having me. Cheers, Art.

Aaron Phethean (40:48.161)
Awesome.

Okay, that is really good. I think, you know, quite close to a...