Transformation focused technology events we've co-presented, sponsored or simply attended.

Summary: Filipe & Chris on AI, Claude, and the Future of Business Systems

Host Filipe interviews Chris, CEO of a 20-year-old system integration business, about how he and his company use AI day to day and where he thinks the technology is headed.

Daily workflow.

Chris describes debugging a personal project (built with Agent Zero) using Claude, running client research through Perplexity to brief a team member, and refining client documents in Claude via several distinct "personas" — for business development, commercial work, and delivery — each with its own instructions, uploaded skills, and brand guidelines so they "stay in their lane." He uses Perplexity's Comet browser mainly for research; Filipe is especially enthusiastic about Comet's ability to autonomously take control of a browser tab and complete tasks.

From skeptic to "AI-powered CEO".

Chris traces his path from ChatGPT to Perplexity to Claude, once his development team started using it heavily. The turning point was Claude's Projects feature, which let him define personas and iteratively refine them like an agile "retro." One persona became four. Chris, originally a developer, says this let him "release his inner geek all over again."

AI in software development.

The company's CTO built a multi-model pipeline: draft specs go through Claude, get reviewed independently by Codex, become code against internal standards, get reviewed again, then are built, deployed, and tested automatically. Chris insists human judgment about "what good looks like" is still essential — AI is "a better calculator," useful only if you know which equation to solve. Filipe adds that he cross-checks multiple models (Gemini, ChatGPT, Opus, Grok) against each other before finalizing work.

Data as the foundation.

After 20 years of large-scale integration work — including modernizing a major Australian airline's systems onto real-time data feeds — Chris's company is now bringing that expertise to mid-sized businesses. Rather than betting on one AI tool amid a "gold rush," the strategy centers on the data layer: extracting real-time data from SaaS tools (HubSpot, Xero) and legacy systems (Oracle HCM, Pick databases) alike. Their product puts a private LLM over this operational data, letting the company run analyses (Porter's Five Forces, data maturity models, Nash equilibria) and auto-generate feature specs that feed directly into development.

Chat vs. dashboards.

Asked if chat interfaces will replace dashboards, Chris says likely both. After finding Power BI too clunky for real-time demos, the team fed Power BI screenshots into Claude, which regenerated the charts in JavaScript within hours. Now the chat interface itself decides when to generate a chart, pulling data from Xero, the ERP, and the CRM to build forecasts on demand. Both agree businesses may increasingly build custom AI tooling instead of paying for full SaaS subscriptions, though Chris expects this to be cyclical as companies weigh token costs against subscription fees.

Personal use and closing.

Chris experiments with Agent Zero and OpenRouter for personal tracking, and uses the Oura ring for health data, hoping for tighter integration with tools like Apple Health. Filipe mentions a similar health "super app" idea he's exploring. The conversation closes with Chris sharing where to find him — DataSide.cloud and Sida4.io

00:00

**Filipe:** Mr. Chris, how are you?

**Chris:** I'm very good today. Thank you.

**Filipe:** Excellent. Thank you so much for joining. Chris, I'm sure we're going to have a very interesting conversation. We start with a simple question. What have you done with AI in the last 24 hours?

00:21

**Chris:** Okay, well, I've been busy. I think five minutes before this call, I was trying to use Claude to debug a problem I'm having with Agent Zero. A bit earlier on, one of my team members was telling me about a client meeting that he'd had with a particular organization. So I ran my standard opportunity review through Perplexity to pull out a statement. And I sent that to him, said, here you go, here's where you can start. And he said,

00:48

And I've also been using Claude because I run a few different personas in Claude in my role to try and refine some documents I'm trying to prepare for a client in terms of next stages of opportunity and how we could do some project variations on what we're doing at the moment. So I've been busy.

01:11

**Filipe:** Nice, nice. Very interesting. So you mentioned Perplexity and you mentioned Claude. Perplexity, great for research. It's incredible for real-time information. It's amazing. Have you been using the Comet browser?

**Chris:** Yeah, I've been using the browser. Yeah, I have some spaces. I haven't embedded it into my workspace and my email and things like that. I tend to use...

01:38

just for client research. So I've got a couple of standard prompts. I've got a space which says this is what I need to do. Here's my persona for that so that I can ask it to run in a couple of different modes for me.

**Filipe:** Nice. So the power of Perplexity, in my opinion, is with its browser, which is called Comet. Then you just start asking him questions and say, hey, take control over the computer and just do it. I don't want to do it.

02:07

And you can see it starts clicking and then going and filling information. It's incredible. It takes control of the tab and it's really, really powerful. I really love it. And then you mentioned Claude and you mentioned different personas. How are you thinking about this? Like projects, folders?

02:27

**Chris:** Yeah, I'm using projects at the moment. So I'm using the browser version rather than the code version myself. I've got a team at work uses the code versions heavily, doing some amazing stuff there as well. But for myself, you know, as a CEO of a business, I've got things I've got investors and shareholders I've got to deal with.

02:50

I've got it. I've got to go and do relationship management, business development. Then I've got to do commercials. And then I get involved in the engagement and delivery as well. So I run different personas for each of those. And then I've got one that I run for, you know, which is a project for when I'm trying to generate tooling for myself as well. So I've done some work there for proposal generation. So I've got,

03:13

And then I've got a handoff between each of them. So how they stick in their lane, you know, each one knows what it's meant to do. So it'll say, Hey, you need to, you need to go over here now. So that way I don't get it polluted. The one I probably use the most is my business development, which has got a whole bunch of things. I've uploaded a brand guides and things like that. So our marketing guys given me all the branding. So then I can generate high level documents and opportunity statements and

03:39

analysis because I feed in the content, the research from Perplexity into that persona to say, now do something, you know, whatever that might be with that. So that's kind of how I'm using it. Yep.

03:52

**Filipe:** You are an AI-powered super CEO.

**Chris:** I'm working on it, but it's amazing how far it's come and how I've been able to think about it in just the last six months, 12 months. I was scratching my head saying, what's this stuff going to do for me? I don't quite understand what I'm going to get from this.

04:11

And now, you know, I'm kind of at it the whole time. And I've got sometimes I'll start off a query and I've got it going. And then I'll pick up my mobile phone and I'm going to watch a rugby game. And I'm just going, I just need to get it to go a bit further. So there'll be a break in the play and I'll get it to go a bit further. So I think I might have got a bit crazy.

04:34

**Filipe:** Yeah, I can relate. So do you think that Claude made you this change of perspective into AI? It was the desktop app. It was really the power of Claude and Opus that really made you change your complete workflow and create all these

04:53

projects?

**Chris:** I think so. I think it was, you know, I started off with ChatGPT 12, 18 months ago and just going, okay, I'm still trying to get my head around to do it. And then Perplexity came along and I found it, I found it more interesting for deep research as opposed to writing. And then I guess, you know, our development team started doing some amazing stuff with Claude and,

05:19

And I just started to think about a couple ways they were thinking about it. Why couldn't I do something similar from an executive or a business perspective? And so just it clicked with the projects. And I realized I could create a project and say, this is who I am. This is sort of what I want to do here. Now write me the project instructions. Tell me what I need as skills.

05:43

And so then I'd upload the skills, I'd create the project instructions, and I worked out what I could do with that. So then I'd get some work along those lines and said, okay, a bit like you're having a sprint, you have a retro at the end of a sprint,

05:57

okay, go through and run a retro on how I'm using this thing and, you know, then upgrade it. So it's then evolved. So from one persona has gone to two to three to four, you know, and then, okay, how do I handle integration between these different personas? What do I need and what in terms of skill files? So it's kind of...

06:17

And I originally trained as a developer and an engineer. So it's let me release my inner geek all over again.

**Filipe:** Nice, nice. Yeah, and I mean, the whole skills world is incredible, right? There's a lot online, but even you can ask anybody.

06:34

like, hey, let's create a skill just for this. And what you said, it's really interesting, like the recap feature, like, hey, go through all our conversations from the last two weeks. What can we automate? What can we build like a skill? So I just tell you, like, hey, do this, and we create the files. So I think it's so powerful. I mean, I think you're in the right track.

07:02

that everyone in the company how are you thinking like in your company the use of AI how are you spreading it do you have like uh workshops for for your team how are you how are you taking this

07:15

**Chris:** At the business level, you know, I'm taking it softly, softly because, you know, trying to get it to work. But I'm slowly getting some of our exec team to lean in. But I think what's really been interesting is our CTO and, you know, what he's done. So we run, you know, for a long time. The business has been going for about 20 years. We've always been strong in the software development space and had, you know, business

07:41

leaned deeply into continuous integration and automated testing and things like that. So we've had a development process and we run a team, you know, we've got teams in Manila and what we've been using Claude is effectively to use different AI agents.

07:56

to sort of instrument how we've done code. And what's been interesting about that is at the code level is not being just saying, oh, it's just one model. So we have a development process where we'll feed in the draft feature specification and iterate that through Claude,

08:13

but then run a separate Codex review of that specification and iterate it. And then we'll generate the code according to the standards. We've got unit test plus code coverage, and then we'll run a separate Codex code review on that. And then we'll check it in, build it, and then deploy it, run all the automated system integration and system testing, and then run a retro agent as well.

08:40

And that's been really powerful. So what we're doing in the training is getting the team to lean into that and get confident in that. And people are a bit nervous as well. But what we've got is an experienced team. And I think that's been the key thing. So from an AI perspective, just trusting the AI by itself is not enough. You kind of need to know what you're looking for. What does good look like?

09:03

And so it's the human in the loop is critical. And my CTO would say, you know, Chris, you haven't cut code for a long time. You're a terrible developer, but it doesn't matter. Lean into some vibe coding. Do some of that to work out what it can do and experiment. When you've got something that you think is useful from a business perspective, hand it over to us.

09:25

and we'll make it, we'll deploy it the right way. So I think that's the key thing is not just trusting that, you know, what these tools produced by themselves is going to work. It's, you know, as I said, knowing what good looks like and then, you know, leveraging that. So I look at it and it's probably a bit simplistic. I look at all this stuff as just a better calculator. But if you ask it to solve the wrong equation, it will. You need to know what the right equation is.

09:54

**Filipe:** Absolutely. Absolutely. A few interesting things. You mentioned Codex. I mean, I think people shouldn't sleep on GPT-5.5. Super powerful. And I like that you're using both, right? It's like...

10:08

It's like you have two incredible minds. Why don't use both of them, right? I see them with different personalities, different training sets, different companies that build them. So I think that's super smart. I tend to put all of the models against each other. Like, hey, Gemini, ChatGPT, and Opus are saying this. What do you think about it? And then grab that answer and put it into Grok. And then all that answer, create a...

10:37

document to start the coding you know so I think that's that's interesting super interesting absolutely and the better calculator analogy I like it yeah I like it and how do you see the I mean as a CEO how do you see the future I I know it's a super difficult question but

10:58

With all these agents, how do you see your company evolving? Are you hiring? Do you feel that you need to hire less and people are more productive? Do you feel you're going to be increasing revenue with the same people and maybe that makes you hire more people because you can increase revenue? How are you thinking about this AI?

11:23

**Chris:** Definitely. I think, you know, my initial thing is doing more with same. And so, you know, how do you leverage what you've got? We definitely see productivity improvements, but us as a business, you know, what we've done traditionally is we've done system integration projects for people, for big enterprises. So we've just finished a three and a half year program with a, you know, a large airline in Australia,

11:50

converting all of their systems, their legacy systems to use real-time data feeds. And for those out there, you know, Kafka Confluent type stuff. So we've done this big end of town mission critical type systems.

12:02

And what we've seen out of that is, okay, how can we bring that thinking to mid-sized businesses as well? So what we did, and so from the AI perspective, you know, we see ourselves as there's going to be all sorts of AI tools, all sorts of things that people build, and it's a bit of a gold rush.

12:20

So trying to work out which one you're back is hard. But what we see and what our strength is in that data layer and all of these AI tools are going to need strong operational data that they can use. And so, you know, our focus at the moment is how can organizations get the data out of their SaaS, out of their legacy operational systems, which might be on premise. It might be cloud. It could be, you know, 20 years old, 30 years old in some cases.

12:48

So our focus is leveraging the learnings we've had over 20 years of system integration. And we were working on a product, which we've got deployed now, which essentially can take a real-time feed from an organization using connectors out of any of their systems. In the midsize, it might be HubSpot, it might be Xero, but it could also be an Oracle HCM product. It could be an old Pick database product.

13:15

And then it's a real time feed, which is event driven. And we were sort of starting off, you know, it's plumbing, but how do we show the sizzle? And we were thinking, I will do dashboards. We'll do whatever. Um,

13:25

But for kicks and giggles, what we did is we deployed a private LLM over the top of operational data and we use it ourselves. So we're sitting there having conversations with our own business about what's the data quality here? Where are the holes in the system? And being able to then improve this. Okay. And then asking questions like, okay, what's the best automation we could build between systems in our business?

13:54

And it'll go through and extract it. And it'll actually produce the first level feature and product spec, plus all the data, underlying data fields that are required to do that. And given what I just said to you about our AI automation, where I said we can feed in the base level specification and then take it through an automation pathway, we've got that first level spec just from the tool.

14:16

And, you know, we've done things because the language, the LLMs have got capacity to run, you know, Porter's Five Forces analysis, run me a data maturity model, you know, run me a Nash equilibria, you know, with game theory. If you can do this, it's amazing. So then we're getting data. And the point of it is, is we, you know, the product we've built, we can go to an organization, say, we will plumb in your connectors, you know, your systems, right?

14:45

into this you'll have your own private query engine that you can then start to say I can now use that as the basis to overlay your other AI tools but it's in a private environment because it's all contained self-contained within the cloud so it's pretty exciting and so we've gone you know we've got a product that we've gone to market there and then of course you can start to put agents over the top of that which will be the next level um so people can build their own AI enabled

15:12

products in their organisation that's custom to them. So it's pretty cool.

**Filipe:** Interesting. Very, very cool. Do you feel like this UAE

15:22

the chat interface and then you know the one that we know it's replacing dashboards do you think that that that's going to be the thing coming in the future or do you think it's both

**Chris:** I think there's a combination so one of the things we did is um you know we were we were trying to demonstrate to people using Power BI what would be possible with real time so we got this wired in and it was so clunky you'd

15:48

You'd show someone, you'd go into a HubSpot and you'd drag a deal through a stage and then flip back to Power BI to say, well, you can see it coming now, except you couldn't because the Power BI schema didn't support it. So that we started off by, we literally took screenshots of the Power BI charts we'd created.

16:05

and then fed them back through our Claude process so that we could generate the JavaScript version in line. And so then we enabled that. So within two or three hours, we'd replaced the whole Power BI with using Claude because Claude knew what the structure, the underlying database schema was. So a few iterations, it had written the code that we needed for the dashboards.

16:28

So we've got that going. But then what we did in the chat is we gave it the rules of how it could, if it was interpreting the data across systems, it could say, well, actually, the best representation for this is a chart.

16:43

And it will generate the chart in line into the chat so that you can then save that chart. So as a CEO, I can be saying, hey, tell me what the forecast is for my business, because it would know from Xero its underlying invoices. It would know from the ERP what the work in hand was, and then it would know from HubSpot, from the CRM, what the future deals forecast. So it's then building a chart of what it thinks the next three to six months looks like.

17:10

So, you know, that can only evolve and get better. And that's the sort of stuff we're saying, we want to bring this to all sorts of mid-sized businesses, because then you're using it to drive the improvement in your own business. And ironically, you're getting the data structures of the underlying SaaS tools and

17:29

which typically use 5% to 10%. So why couldn't you create a specification to replace them completely if that's how you chose to put your own tools in place?

**Filipe:** Interesting. Yeah, I mean, I think the software, the SaaS industry is going to be evolving rapidly.

17:51

It needs to evolve. If not, it's going to die, right? Like the new headless ideas, you know, that you do the interface. What you're describing, right? Like, hey, we provide the APIs and the data, and then you do whatever you want with your dashboards. And I mean, code is free right now. So you do whatever. I think that's a smart move.

**Chris:** I do think there'll be a bit of a cycle because then, you know, people go, actually, how much are all the tokens? How much is all this stuff costing me? And so what's the balance between this? Because, you know, you're replacing your SaaS monthly fees with your monthly tokens. And we don't quite know where that's going to land.

18:34

um in terms of your cost to develop your own stuff so you know it's like all these things it'll swing too far and then it'll swing back and you know there'll be a bit of a yeah somewhere in the middle

**Filipe:** the good thing for us for consumers is those prices are going to go down in my opinion right talking start token prices are going down we've been seeing that from the beginning and uh i mean more intelligent and

19:02

Yeah. Cheaper and then prices because the dilemma is going to be that, hey, should I build it myself, use tokens or I'll buy it from you, but it should be cheap. If I buy it from you, it should be cheap. So.

**Chris:** Yeah, that's right. Interesting. What a world.

19:17

and competition amongst models as well. And, and you'll get to a point where you go, I actually don't need to run on the latest model. You know, it'll be like, you know, you don't, you don't need the newest car out of the car dealership. I'm actually happy to buy a used model. That's three or four years old. Or in this case, I mean, three or four weeks old, baby.

**Filipe:** Yeah. I mean, GPT-5, it's a beast. Opus 4.5, it's still a beast, you know, and yeah.

19:48

so incredible Chris I wanted to ask you before we wrap up uh something about your personal life are you using AI in your personal life other than work is there an app that you're crazy your family your friends there's something your iPhone that's standing

20:05

**Chris:** I think I'm kind of using Claude as well for some of my personal things. And what I'm playing around with is in, you know, with Agent Zero, because I use various tools in my life. I try to track this and I try to track that. And so I'm playing around with a common interface that can, I can use. And I'm, you know, this is where I'm doing some of my vibe coding is actually using agents and,

20:30

with OpenRouter and all that sort of thing there. Sometimes a bit more successfully than others. I've found that, you know, Agent Zero, you do some vibe coding and it wants to eat itself alive. So, you know, that's, that's a bit of a slow burn, but you know, I'm always looking for ways I can, I can enhance and improve it. So, you know, I run some personal personas as well, which, you know,

20:56

I haven't got there yet, but I'm working on it.

**Filipe:** Nice. Yeah, I'm looking into, and also the founder, well, you've met Isaac. We're looking into all the data that these things give us, you know, put it into a nice dashboard and add it with our health, you know, like a super app that I don't think it's on the market. So, yeah.

21:19

**Chris:** Oh, that'd be interesting. I mean, you know, I use Oura myself. So I've got the Oura app, so I track my health data. And what would be great with some of these devices, you know, is to be able to tap into the Apple health data and the rest of it so that you could actually bring that stuff together in a common interface so you can see it. You know, at the moment, it's still a little bit closed loop where it's hard to get hold of.

**Filipe:** Yeah. Yeah.

21:45

If you go and have some blood analysis, you know, everything, everything, you know, there should be an app that knows everything. Yeah. Awesome, Chris. Thank you so much for this conversation. I really, really enjoyed it. I want to hand it over to you. Where can we find you?

22:01

**Chris:** Well, you can find me on a couple of websites. Obviously, you know, this is going to appear in a, you know, like its own branded site. The product I talked about, you can find that at www.DataSide.cloud. That's D-A-T-A-S-I-D-E.cloud. My integration business is called Sida4, and you can find that at Sida4.io.

22:25

failing that search for Chris Eldridge Australia on LinkedIn and connect with me if that's useful too um and you know always happy for a chat as you probably guessed from uh the style of this this this conversation

**Filipe:** thank you Chris have a great one

**Chris:** thanks very much Filipe bye.