#119 Google Cloud Next 25 roundup

[00:00:00] Dara: Hello and welcome back to the Measure Pod. Listeners will probably remember me. I’m Dara, I’m back in the, in the hosting chair after a bit of a break from the Measure Pods, but I’m back now with a new host, my co-host, Matthew Hooson. So firstly, Matthew, welcome to the hosting Chair on the Measure Pod.

[00:00:34] Dara: Yeah, thank you for having me. Pleasure to be here. Yeah, no, it’s exciting. So we, as I mentioned in what would’ve been our last episode, were mixing up the format a little bit, shaking the bag. So we were gonna be joining as the new co-hosts, and we might change the format a little bit, which I guess we can maybe cover off today a little bit, although this might be a little different.

[00:00:58] Dara: This will probably be a slightly different format to what’s gonna follow, because obviously we’re introducing ourselves or I’m reintroducing myself. and then we might start it, it’ll kind of start to settle into a new format, probably from the, from the next one. The other thing is we’ve got quite a big topic to cover off today, so we probably need as much time for that as we can possibly give ourselves.

[00:01:20] Matt: Yeah, we thought we’d have a nice easy start with the next 25 and then Google reinvented BigQuery pretty much. So that’s not giving us a huge amount of wiggle room. 

[00:01:29] Dara: And that’s just BigQuery. Yeah, we’re probably not even gonna get to touch on any other updates, but, we’ll, we’ll see how we go. But anyway, little, little quick Rero.

[00:01:39] Dara: So as I’ve said already, I’m Dara, listeners will probably know me unless you’re newer listeners and you’re thinking, who’s this guy? I used to be on the metro pod back in the day, and now I’ve, I’ve returned my triumphant. That’s a bit ambitious actually. We’ll see. We’ll see what happens. We’ll see.

[00:01:56] Dara: We’ll see how it goes. But it’s my return, whether it’s triumphant or not, we don’t know yet. Matthew, do you wanna give yourself a bit of an intro? 

[00:02:03] Matt: Yes, I am Matthew Hooson. Chuck my second name in there as well for, for Fish. I work at Measure Lab. I’m head of Engineering and Technology. I think I may have been a couple of times guests on the measure board in the past.

[00:02:20] Matt: I can’t remember now, at least two. I’ve done at least two guest spots, so I feel I’m perfect. Qualified for this 

[00:02:27] Dara: auspicious occasion. I, I, I think the reason you’re slightly confused is I think one of those, you maybe had, had a few beers, I think your most recent Oh 

[00:02:36] Matt: three, 

[00:02:37] Dara: yeah. Oh, you weren’t counting 

[00:02:38] Matt: That one.

[00:02:39] Matt: Okay. No, I forgot about that one. But for the reason you just mentioned, yeah. Yeah. It was super week and I, it was about one in the morning. I think that was 

[00:02:45] Dara: recorded. 

[00:02:46] Matt: Don’t I? 

[00:02:46] Dara: Might have had a couple of beers by 

[00:02:47] Matt: That point. Yeah. So, so you’ve 

[00:02:48] Dara: been on two episodes making some sense, and then probably one episode making not a lot of sense.

[00:02:53] Matt: Well, that’s for the audience to decide if I make sense of any even, 

[00:02:55] Dara: But yeah. 

[00:02:56] Matt: Yeah. 

[00:02:56] Dara: Okay. Well, listen, we, maybe we’ll give a little teaser of what’s gonna come in a, in a, in a kind of typical episode now in the, in the new formula, it’s, it’s not, you know, there’s no point over egging this. It’s not gonna be vastly different from what we’ve done before because, you know, we have a formula that works pretty well.

[00:03:13] Dara: It’s obviously gonna be talking about analytics and data. Sometimes we’ll have guests, sometimes we won’t. and we’re gonna include some kind of industry news or. News to do with Measure Lab or upcoming conferences that we’re gonna be at. And then we’ll, we’ll tend to have an interview or a topic where, we’ll, we’ll dedicate the bulk of the episode to talking through that topic.

[00:03:34] Dara: Maybe having a few healthy debates along the way. I’ll probably just disagree with everything you say without actually being able to back it up in any way. That’s, that’s the modern world post truth. Okay. So the, the, the, the news is the topic today. So we might as well just dive straight in. and this is all of the announcements.

[00:03:53] Dara: We won’t probably cover all of them ’cause we don’t run unless we’ve got, I mean, I’m thinking this’ll be about four and a half hours long this episode. What, what do you think? Yeah, it’s a lot. 

[00:04:01] Matt: Yeah. This is for me, I, I know I mentioned it very briefly at the, at the, at the start of the call, but we’re talking about next Google Next 25 here, which was on, was it the seventh?

[00:04:11] Matt: Sorry, the ninth, 10th, and 11th of April. Yeah, last Vegas. Yeah. 

[00:04:17] Dara: Wednesday, Thursday, Friday last week. Slightly confusing because it was live at that time, but we, because we were watching it remotely, we were waiting till the next day, each day to get the on demand content. So it was, yeah, well at least it confuses my small brain when, well, it did 

[00:04:34] Matt: look like they were having sessions at two, 2:00 AM in the morning, so that was a big, maybe they were results to be honest about.

[00:04:39] Matt: Maybe they, they do, 

[00:04:39] Dara: they are hard workers over in the US nipping off to the casino in the breaks just to play a bit of blackjack and then back to us.

[00:04:46] Matt: I think the killers played what at the, at next 25. The Killers were the big special evening guests at Next 25 and the, in the Las Vegas sphere. So there you go.

[00:04:56] Dara: Wow. Us They should have got them on, they should have got them on the stage and asked them their thoughts about, you know, the latest agentic updates. 

[00:05:05] Matt: Yeah. It couldn’t have been worse, I dunno how many you watched. I got through quite a few. But the thing with these tech conferences is they have some of the most awkward.

[00:05:17] Matt: Wooden presenters for these talks? Well, like us, we try to have this like us. Yeah. Try some sort of pretend banter back and forth and, and awkwardness and some jokes. They’ve just written on some napkins, but it was pretty awkward. So maybe the killers would have a 

[00:05:33] Dara: better job. Yeah, maybe. Although what I did like about it was because e every now and again, you know, it’s hard enough to keep your, at least I find it’s hard enough to keep my focus.

[00:05:41] Dara: ’cause some of it went over my head and then I’d start to tune out a little bit. or, or some of it was just a little bit, you know, at times, a little repetitive when you’re hearing about the same updates over and over again. So when one of them, so some of the presenters, had this real like, you know, big, big energy.

[00:05:58] Dara: So they’d come out and be like, Hey, gimme a woo, or, or something like that. And it would kind of like wake me, wake me back up again. So it was, I did kind of appreciate the kind of slightly different styles of some of the presenters. ’cause it kept me, kept me focused or, or just being like, what? 

[00:06:14] Matt: Yeah, yeah, yeah. There was one person, I should have written down some names, but who did an escalator pretended to be on an escalator to see a computer and then came back up in a lift at the other computer that woke me up just from confusion, really. Just like, what the hell are you doing? But yeah, it worked.

[00:06:29] Dara: Yeah, that was a, that was an interesting one. Yeah. Yeah. Alright, so we’ve got a lot to cover. So do we want to, do we want to do a bit of a broad, kind of pick out some of the kind of overall highlights? Or do we actually just want to. Drilled straight into the kind of areas that are a little bit more relevant to like what we do at Measure Lab in terms of kind of data analytics, BigQuery updates.

[00:06:56] Matt: Maybe we just, we just fly through a couple of headlines. ’cause there were two keynotes, if I’m not mistaken. There was the keynote, keynote and then there was the developer keynote. So there’s a, and in the main keynote there was stuff that was big and impressive and potentially, potentially affecting our sector, but not necessarily, exactly bang on the money, like the developer keynote and a lot of the other talks after that.

[00:07:22] Matt: These were all big query based and, and agent, agent Agentic, agentic. I’d love to have accounted for how many times that was said, but I should have done that before we started. You could build an agent to do that. I could build an agent. If they would give me access to agent space, I’ve asked them about six times and I don’t seem to have managed to get into it yet.

[00:07:41] Dara: Yeah. I thought it was just me, but I’m glad I’m not the only one in the, in the, in, you know, outside the club desperately waiting to get in. 

[00:07:48] Matt: Yeah. And it’s like looking at the rope, letting all the good looking, seeing all the good looking people go in while we hang around the bins with the wrong shoes.

[00:07:54] Matt: Yeah. so I’ve got, I, I can check out some very quick headlines and then it, you tell me anything I’ve missed this, this is more like on the, the keynote thing. Yeah. So 75 billion CapEx spent for next year, which struck my ear. That’s an absolutely huge amount of money on cloud and AI that they’re spending.

[00:08:17] Matt: Yeah. and I think that’s pretty similar for most of these tech firms. I. Sure I saw it, I saw an article recently. I’m, I’ve been doing really good at just sourcing random things that I can’t remember or vaguely remember reading recently. But yeah, a lot, there’s a lot of big spending going on.

[00:08:33] Matt: They are really going heavy in on cloud and particularly AI and agents. Not that shocking. They announced this new Ironwood TPU, which tends to be a tenor processing unit, which is what they’ve been using to underpin a load there, it’s like a Google made tenor processing unit that underpins a load of their tech.

[00:08:50] Matt: And as we all know, that is 42.5 x of flops. I mean, I knew Yeah. Whatever that means. but yeah, they’re, they’re rolling that all out, that new architecture 20. They, I just found this interesting, that I’ve never heard this as a metric or a measure before. I think they were comparing Gemini 2.5 or 2.5 light to GPT-4 0.0, and they said it was 25 times more intelligence per dollar.

[00:09:18] Matt: I just thought that was an interesting brand new made up metric. Intelligence per dollar. Yeah. I mean, how 

[00:09:24] Dara: do they even, how do they break that down? 

[00:09:28] Matt: Yeah, that’s not, maybe let’s just move on. It’s just a number. yeah. And then, an agent agent. Agents agent kit. So they talk quite a lot about agent care and, and this, this ability for agents to talk to agents and their support of cps, which is like model context per model, protocol model, context protocol.

[00:09:50] Matt: Thank you. Thank you. agent space, which we’ve kind of just alluded to, and then just an absolute pile of BigQuery stuff is what I kind of had as their overall headlines. Dunno if I missed anything there. 

[00:10:05] Dara: Well, there was a few, the few other thing, I mean, there’s probably lots of other things, but there’s a few other ones that I thought they didn’t, they didn’t necessarily, you know, you know, you know some of ’em you think, I can see why that would be a big thing, but it doesn’t, like, it’s not immediately relevant to us.

[00:10:17] Dara: So like the wide area network wan, is that, is that right? So Google is opening up their own, their own network that they have that you, you know, that they use for our services like Gmail and photos. So this ultra low latency, you know, high speed network that they use, they’re opening that up to enterprises.

[00:10:39] Dara: sounds interesting. I mean, I had no idea what kind of enterprises that would be applicable to. I assume it’s gonna be costly. It’s probably very large enterprises only. but I thought that was vaguely interesting. Sounds impressive. Yes. Sounds impressive. Sounds impressive. Yeah. And in a similar or in a similar vein, the kind of on-prem solution, their partnership with nVidia.

[00:11:03] Dara: So if you’re in a kind of very, what’s the word? Where do you need high security? Like in banking or banking? Yeah. So if you need to have an airlock, am I saying the right things here? You airlock. It’s something to do with the airlock. 

[00:11:19] Matt: yeah, I think I, I, yes, I think so, but you gotta keep, you gotta kind of keep it air, you gotta keep it airtight from the outside in and air, you can’t have any sort of interference.

[00:11:28] Dara: And yeah, so that, I, I, I guess the reason that peaked my interest, or at least I paid attention to it, is like, is that, along with is if these are maybe slightly disconnected, but I felt like there was a bit of a theme where, you know, in the past at least the impression I had was like, Google almost went about thinking, well, you know, we’re gonna do things our way.

[00:11:48] Dara: We don’t necessarily have to be open source or we don’t necessarily have to provide you with, you know, on-prem solutions, anything. It’s like, it’s in the cloud. That’s it. And I got the sense from this that there’s actually a lot more focus now on kind of, you know, collaboration, also allowing people to do their own thing a little bit more.

[00:12:07] Dara: That seemed to be a bit of a theme running through the whole thing. I thought rather than just, this is our way and you have to adopt that, or, you know, do something else. 

[00:12:15] Matt: I feel like what, that’s kind of, that’s kind of the vibe I’m getting from Microsoft recently. Like with co I think co-pilot recently the, the, the coding assistant, Microsoft locked that down a lot and made only available in like visual Studio code and stuff like this.

[00:12:31] Matt: Whereas a theme throughout this, like we may touch on some of the AI stuff later, but the model garden and being able to use like Claude and Meta and all these different models in, in the platform being able to have like on-prem versions of, I dunno how you have an on over from on-prem version of cloud, but that’s probably just semantics.

[00:12:51] Matt: It’s just a very small cloud. Yeah. It’s just a tiny little, tiny little cloud in the room. Yeah, it definitely felt like Andy talked a lot about sort of. Cross cloud collaboration and networks and security and just being able to, to be able to have a specific, say marketing warehouse or a whatever warehouse at a different set of data sitting away from everything else that may be sitting in AWS or Snowflake and try and foster that, that collaboration and that movement of data around.

[00:13:20] Matt: Because as far as they’re concerned, I suppose if someone’s really embedded in another platform, if they can’t get ’em to migrate over, then they may as well try and may as well try and, empower the marketing teams to build their own little solution over the fence perhaps. 

[00:13:34] Dara: Yeah, I think that’s it exactly. It’s opening it up to a lot more, I guess, individual users, but, you know, businesses and, and business users, that was real, it was definitely running through the whole thing. They’re trying to make everything easier. They’re trying to make it more accessible, you know, the agent to agent protocol as well.

[00:13:50] Dara: kind of driving this like multi-agent, multi environment, you know, doesn’t matter what tech you’re using or what frameworks you’re using, you’re gonna be able to build agents that are gonna communicate with. Each other. You, you know, so it’s in terms of where you store your data, where you access your data from, what agents you’re building.

[00:14:06] Dara: It’s trying to kind of push towards this, I guess, version of things where it doesn’t really matter and you can, you know, you can store your data in one place and then use Google technology to do something else. I’m just saying things now. Just, just, just, just throwing out words. Interoperability, that was another word.

[00:14:25] Dara: Big thing. Talking about synergy. 

[00:14:27] Matt: Yeah. Yeah. But they, they just have the, might they have the might of this technology, of this cloud technology, their experience with artificial intelligence, their, their, this fantastic platform in big BigQuery, I suppose for them, making it open source and giving people more opportunities to bring more things onto that platform is, is a no brainer.

[00:14:49] Matt: They don’t, they don’t necessarily. I have to worry too much about building some old singing, all dancing separate products. They can just flex that Google muscle. They definitely seem to have caught up in the AI race quite a lot. There’s a lot of top, a lot of leaderboards recently, which we wouldn’t have thought maybe 18 months ago.

[00:15:08] Dara: No, you wouldn’t have. It’s remarkable, isn’t it? I think people were starting to fear they’d been left behind and, you know, underestimate Google at your peril. I think that’s probably the key point there. It’s always been the lesson. 

[00:15:21] Matt: Yeah, yeah, exactly. Yeah. It’s what a tattoo says. 

[00:15:23] Dara: Yeah. Before we, we are inevitably gonna lead towards some kind of big query and how AI is, is, is, is, and Gemini is in BigQuery.

[00:15:34] Dara: but just a couple of other things that I picked up on but don’t know a great deal about. The new Firebase studio sounded to me like it’s their kind of vibe coding. Environment. Just to throw out yet another buzzword is that, you know, I, I watch parts of the video of the guy introducing it, but I, I, I didn’t watch the whole thing, but it’s like a, you know, you get your, you get your interface and you can use natural language to, to basically build apps within, within Firebase.

[00:16:03] Matt: Yeah. That’s, that’s the, that’s the vibe I got as well. They do, they have got another, I dunno if, I dunno if you’d describe this as vibe coding, maybe you would the, this Gemini code assist Kanban thing, which is in a similar space, which maybe we can touch on a bit later. But the Firebase thing, that’s what I understood it to be as well. Yeah. Just sort of natural language building on that existing Firebase framework for application development. 

[00:16:27] Dara: And, and, and Gemini is just in everything. I mean, every single thing they talked about Gemini is the, is the, is the common denominator, isn’t it? I can’t remember seeing many demos they did. 

[00:16:38] Matt: Where at some point they didn’t interact with Gemini in some way, shape, or form. I mean, I was watching most of the sort of BigQuery data ones, but it did feel like they were talking about it from security, talking about it from IMS perspective, talking about it from costs, talking about it from database management, talking about it from data. It is literally absolutely everywhere. 

[00:16:57] Dara: I was gonna just talk about Vertex AI and the different kind of multimodal, you know, the fact they say now they’re the only one that does text video now, what is it?

[00:17:07] Dara: Video images, text to audio and audio clips. I dunno. The, the, the, the, the difference between CHIRP and the other one is a bit blurry. Chirps are one of them. Lyria. Is that the lyria? Is the audio? Audio? I don’t, I don’t, but my only experience, I tried all of these. ’cause I thought, well, why not? This is, this is kind of fun, fun.

[00:17:30] Dara: Well, those are videos. I’ll create some silly videos. Yeah, yeah, exactly. Yeah. but someone, I don’t think this has been fully, ’cause the version that was demoed isn’t the version that I’ve got access to. Like it was restricted in some ways. I couldn’t upload an image to c then create a video from that image.

[00:17:45] Dara: It wouldn’t allow me to do that. I obviously tried to upload my own voice as well to create an AI version of myself. 

[00:17:52] Matt: I mean, why wouldn’t I be just in time for this podcast? 

[00:17:54] Dara: Just in just, maybe this, maybe this is it. but that didn’t work for me either. but anyway, in theory at least they’ve got all these upgraded, yeah, multimodal models.

[00:18:06] Matt: That is one of the frustrating things with, with next is like, especially now, like there’s so much cool stuff. You are like, oh. You kind of jump over to your cloud just to go and have a quick play with it and realize that it’s not there. Yeah. And they’ve got a lot of, vague coming, soon, at the bottom.

[00:18:23] Matt: A lot of their decks describing these tools. So it’s like, ah, I really wanna have a play with that, but I can’t get hold of it. Then who knows when it will eventually appear. 

[00:18:32] Dara: Yeah. Or how you get early access to, or is, I find it really confusing as well. Some of the things they’re talking about are obviously, you know, some of them are even all already existing features, but they’re enhancing them and then you don’t know if you’ve got the original version of the announced one.

[00:18:44] Dara: And then some of them are brand new, and whether you have access to them or not. And then sometimes they do like a gradual release, and other times they just say, this is available now. so it is, it’s, it’s a little confusing. Another one that I went to look at, which I guess we will talk about is the, you know, where you can build an app from, big from a BigQuery notebook.

[00:19:05] Dara: Yeah. I didn’t have, I couldn’t, I thought they said that was available and I couldn’t, I couldn’t see it. 

[00:19:10] Matt: No, I think I did the same thing. I don’t think I could see either. And I know we do have, we do have like the preview stuff turned on. and we are, we are partners, so we do sometimes get early access to some of these things, but I, yeah, I didn’t see it.

[00:19:23] Matt: We’ll see, we see we’ve done these things here, like sit on the agent space waiting list. Yeah. Forever. Yeah. Our people play and I sit on the outside. 

[00:19:33] Dara: Okay, so let’s move, let’s move the focus onto, onto BigQuery then. And there’s, there’s, I mean, there’s a lot to cover here, isn’t there? 

[00:19:42] Matt: Eh? 

[00:19:43] Dara: No. Yeah, there’s a couple of things.

[00:19:45] Matt: Yeah, it did. Yeah. What I think we were both saying before we, before we started today, like we kind of had this intention of looking much more, much broader across things and maybe understanding things that may be a little bit more peripheral on data and, and some of the stuff that’s going on elsewhere in the platform.

[00:20:02] Matt: But, it was just, it was more than enough to, to even get through and watch just the BigQuery announcements alone. Without thinking about anything else, without even really necessarily thinking about all the stuff that’s in Vertex AI and the agent space and the, all that stuff. So a lot. 

[00:20:20] Dara: So one, if, if I can maybe, tee things off or, or, or set this up with a, with a kind of que I guess it’s a question, or I’ll turn it into a question.

[00:20:30] Dara: At least I saw this phrase. I, I don’t think it, it, it came up in some of the videos. but it was in a, there was an article, published by somebody at Google kind of closely after, one of the big query talks, I guess, and they’re using this phrase, autonomous data to AI platform. So BigQuery is not just BigQuery anymore, it’s now this autonomous data to AI platform.

[00:20:56] Dara: Complete rebrand. Yeah. 

[00:20:58] Matt: Yeah. They, they, I think they kept some at the one I watched earlier. That’s, that was the tag eye at the very beginning. Mm. Like. BigQuery, the autonomous data, AI platform. I, it’s not very catchy. No. and it, I was gonna ask you the same thing, like what does that mean?

[00:21:17] Matt: because it is, it is a bit, techy and business speaky, but it, it definitely, I think it speaks to just how many different agents, how many different, sort of autonomous behind the scenes systems and, easy accessible point and click solutions they’ve put in there, stuffed into BigQuery now to try and make it as simple to get to activation.

[00:21:46] Matt: But I suppose that our version of that activation being AI, LLMs, quizzing the data, getting insights in that way, but with the conversational analytics agent they mentioned. 

[00:21:57] Dara: I mean, I’m, I tend to be, I dunno what to call myself naive. I’m wondering whether to be kind to myself or mean, but let’s just say I’m naive and maybe a bit optimistic, but I watched some of the demos and I thought, wow, this, like, basically you just don’t need to have any actual coding skills anymore.

[00:22:12] Dara: You just, you know, you go into BigQuery, open up a canvas or something. Yeah. And, and you can use all these, specialized agents to do the heavy lifting for you. And obviously the demos are rehearsed and they’ve got things lined up and they’ve got their, their prompts copy and finally copy. But yeah, yeah, there were some very, very kind of packed clipboards I noticed, like, let me just right click and add something from my clipboard and it’s this perfect prompt that they’d probably spent hours fine tuning.

[00:22:43] Dara: but it really looked like with these, you know, with these agents that you can now like end to end so you can. You know, it’ll make suggestions on how you might want to clean up the data that you’ve got. and then right the way through to actually, you know, well, helping you create pipelines and then actually analyzing the data.

[00:23:04] Dara: At one point in one of the demos, the guy didn’t even ask it to do this, and it just went off and used the data science agents. Yeah, I saw that. It was in Canvas, wasn’t it? Yeah, yeah, 

[00:23:14] Matt: yeah. 

[00:23:14] Dara: Yeah. 

[00:23:15] Matt: Yeah it was creating, just started creating new visualizations and new SQL queries and Python scripts and stuff.

[00:23:22] Matt: I, I think that’s what they would like us to think, and I think that’s the intention. It’s they, they’ve got what, they’ve got all the GA four data and the connecting up all these other sources. Thinking about it from a marketing analytics perspective anyway, going into BigQuery, you want this to be the centralized hub of data that you then manipulate and build and create all of these great tools off.

[00:23:45] Matt: So I. To do that, you have to make it as accessible as possible. And all of these tools do that. So to just still label the, some of the tools you were talking about there, we had like the data engineering assistant, which was to sort of help you build out transformation pipelines and, and sort of get the data into shape and, and join things and, and do all those kinds of stuff.

[00:24:12] Matt: You had the data science agent, which was around, data exploration, cleaning analysis, getting it into the right format for one of a better word. And then the conversational analytics, which I guess is on the other side of this whole thing. My gut reaction is it’s all cool and it’s all, it’s all very, powerful and it’s gonna put it, it’s gonna put data into a lot more people’s hands and suddenly a lot more people are gonna have access to stuff.

[00:24:45] Matt: If you don’t have a strong foundation of the data in there in the first place, you haven’t got robust pipelines coming from your, your siloed sources into your, into, into the warehouse. If you haven’t got, you know, robust, repeatable models and all these sorts of things, you’re gonna end up just using all these cool tools built on top of crap data and that’s gonna lead to a lot of problems.

[00:25:10] Matt: So, yeah, they’re cool, and yeah, they could be really powerful, but it, to me, I was, they didn’t mention a huge amount about that bit, the fundamentals of, of getting the data in there and making sure it’s robust and, and accurate. 

[00:25:23] Dara: Well, I was, yeah, I was really curious about that as well, because some of the stuff, it’s interesting, isn’t it, like, at least from my kind of perspective, you will have taken a lot more in than I did, but I was watching some of it thinking, wow, this is really thin.

[00:25:34] Dara: I want them to kind of go deeper on this. And then other times I’d be thinking, wow, this is way too. Technical for me. but there were a couple of things mentioned that there wasn’t a huge amount of detail on, like, they talked about this governance layer. I dunno if that’s right, I can’t remember if that was exactly, it was termed, but there wasn’t really much about it.

[00:25:52] Dara: and they also did talk about how the, and they did show this in some of the examples, but like the data engineering agent can make suggestions about cleaning up the format of your data. But I, I, I guess that’s only gonna take it so far. what I didn’t know, and a, a question I was gonna ask you to get your take on is like, did you see anything that suggested that it would actually be a really meaningful help to you as a, you know, say you’re a business and you’ve got, you know, you’ve got your data and it is a little messy and you do have issues, will any of what you’ve seen actually help to get better data in the first place?

[00:26:30] Dara: Because it seemed like they were suggesting that, but there wasn’t a great deal. It wasn’t a great deal of meat on the bone in terms of how far you could take it, as opposed to just, oh, this form or this column doesn’t have a clear name. You could maybe rename it. Or, these dates are differently formatted, let’s clean them up for you.

[00:26:48] Dara: You know, almost like the suggestions you get in Excel, but like, is it actually gonna help you, you know, if your data’s really bad in the first place, is this gonna help or is it just gonna make the problem worse potentially? 

[00:27:01] Matt: Yeah. I still, I still, I think there’s, it depends on the problems that are in the data, but if, if, I’ve been saying recently that data engineer’s job has historically been just putting out fires, like crap just gets sent into a centralized place and then the job is just to go like, right, how do I fix all of the problems and issues that are, that are in this data and get it into a shape that is of some use to be able to be used for whatever it’s gonna be used for.

[00:27:26] Matt: So I think that there’s, there’s still gonna be those instances where, where rubbish data comes in and maybe, and this is. Maybe I’m going off on a tangent here. Maybe all of the vibe coding and the, and people making their own applications and, and their own data sources is gonna exacerbate that problem.

[00:27:43] Matt: And even more unstructured, messy data is gonna be made available as it’s only gonna be an increase in that stuff. So I, I, I, I think I agree. I don’t think I saw anything of substance to fix those kinds of problems. and I think you would still need it. It’s like with anything, you can augment yourself with this stuff as a professional and you should, in my opinion, I, I wouldn’t bury my head in the sand or dismiss it and go, nah, I, it’s a lot of, it’s a lot of crap.

[00:28:12] Matt: I can do it myself, in a couple of hours. I think they are useful if you augment yourself and, and use it as part of your processes to begin to clean up and, and get data into shape. And not necessarily just completely lean on them, just to autonomously do it, which I think some people may do because they won’t have those underlying skills.

[00:28:33] Matt: Mm-hmm. I’m not sure if I answered your question then, or if I just went off on a random rant. 

[00:28:39] Dara: No, you did. And, and, and I, I guess these things are on a, this is a little bit of a spectrum as well because I just started thinking about like if, if, if someone’s limited in terms of their own skills or, or if it’s a company and they’re limited in terms of their resources.

[00:28:52] Dara: Well, if some of these tools kind of allow them to inch themselves forward, like if they maybe is, maybe is a kind of, maybe isn’t the right, quite the right thing to say, but I kind of think if someone’s got bad data anyway, then if it can make it slightly less bad than at least they’re taking, they’re taking a step forward.

[00:29:10] Dara: So if some of these tools, like you’re right, obviously you don’t just wanna, like, nobody should want to just press a button and then trust what comes out at the end without having any. Layer of kind of validation on that, or without knowing what happened between the time you pressed the button and you know, you then used the output.

[00:29:30] Dara: but if you, if you, if you do use these tools, if you, you know, if you don’t have the resources to take all of this stuff really seriously and do everything according to best practice, then it might help you get a better understanding of that data and where the problems are. And then gradually, almost like, I dunno, again, maybe this is slightly idealistic of me, but I kind of think if you are someone who’s a bit of a jack of all trades, even in a, in a company where you are the data person, you could at least use these tools to kind of plug some of your own gaps where maybe you don’t have the skills to do, you know, full data engineering, but you know enough about what you’re doing with the data and you kind of use this, I guess that is augmentation really, isn’t it?

[00:30:12] Dara: That, that, that is what we’re talking about. 

[00:30:15] Matt: But you’re right. There’s no, there’s no, there’s no point to just sort of go, well, if we can’t get it, if we can’t get it a hundred percent right, then we’ll just never use it and don’t bother there. Certainly for smaller companies, I, I, I think I’d always advise a smaller company to just start to build.

[00:30:32] Matt: When you start getting the data in there, just do it with good foundations and at least then, you know, if you find problems and fixes and, and, and you wanna reshape things, you’ve got good foundations that you can adapt and build from. And then you can explore these tools and go down blind alleys and go down dead ends and come back to the start and figure and evolve and grow and mature with it.

[00:30:53] Matt: If you just, you know, if you just chuck data in and just hope for the best and start making big old decisions based off it, maybe it’s not the, the greatest, greatest idea, like part of it is I. I don’t wanna, I, I I, I’m also, wary of, labeling entire professions that I don’t know masses about. But, the conversation analytics stuff looked really impressive if it worked.

[00:31:22] Matt: If it works the way it works. ’cause you can imagine at that point you’ve got the data in, you’ve got the data transformed and cleaned and, and inner shape, and you kind of are put in a chat interface into everybody’s hands to kind of make the ask, answering of very specific questions available to the masses rather than having to build out specific single dashboards, to try and cover broad questions and answers.

[00:31:51] Matt: Cool. But then again, you know, I’m not deeply into visualization, and data science, but of course I thought I looked cool. 

[00:31:59] Dara: Yeah, I did too. And, and I guess the more you build up the, so I don’t know, again, I might be getting the, the wrong end of the stick here, but it is this where a knowledge engine plays a part as well in terms of building up that context.

[00:32:12] Dara: Does that work with conversational analytics or is that something different? 

[00:32:15] Matt: I think, sorry, I, I think what I took the knowledge engine to mean, so this is BigQuery knowledge engine, was that it understood your, your whole data ecosystem within BigQuery and then it was able to sort of augment.

[00:32:32] Matt: All of the various agents within BigQuery kind of get it, to make better decisions when it’s writing code for you. So I suppose if you’re a data engineering assistant, it has an understanding of the various layout of your data, lineages, schemas, metadata, et cetera, and can make a better decision on how it writes that code to get better answers.

[00:32:52] Matt: That’s what I took that to be. And I think part of that would obviously lead into the conversational analytics thing as well, but I think it also is in the marketing data scientist agent. Yeah. 

[00:33:03] Dara: And the data engineering agent. Yeah. Yeah. No, I think you’re right. And also the thing, I just remember that, that it does as a, it also because it’s using Gemini surprise, surprise.

[00:33:14] Dara: It’s, it’s, and it’s got the search. So it, it can, it can, it not only knows your data, but it can also, it can do the advanced reasoning that Gemini can do and it can access live information. So one of the examples in the demo is something like, can you categorize this data into. I can’t remember what it was, but it was like some established industry norm for customer segments or something like that.

[00:33:38] Dara: And it, you, you didn’t in the prompt have to explain what they were, because it obviously used the, it used the model to actually figure out what that was. So it can kind of mix together your data and then use its reasoning and potentially pull in information from, I guess live information from, from the internet as well.

[00:33:58] Dara: Yeah. Which is ground powerful. 

[00:33:59] Matt: Is it grounding They call it grounding go. Yeah. Grounding with Google search and being able to take it outside of check things. What? Yeah, because they’re all trained up to like 23 or something, aren’t they For these models, but to get something more recent and not, some old code base or whatever.

[00:34:17] Matt: There was also, I just wrapped up this agent thing, I suppose. They look like the inside of BigQuery notebooks. So these are the Python, call out no Python notebooks that sit inside a BigQuery. There look to be a load of new updates in there in terms of Gemini where it just had like a little chat bot at the bottom and you could just sort of type away.

[00:34:41] Matt: They’d just be populating all of the, all of the cells. Similarly in canvas. Canvas I think is just like a visual layer on top of a, on top of a notebook, but you could, again, just be typing away. I think you talked about it earlier and, and it just like spinning off and creating amazing organizations and things.

[00:35:01] Matt: Yeah. and there was, there was the data preparation thing, which I actually had to play with earlier that does exist. I don’t know if it already existed before, before the conference or not, but that is like, there used to be something called Data Prep, which was owned by a company called Trifecta or something like that, that was in.

[00:35:21] Matt: In Google Cloud, but like a paid for additional service. I got very heavy data prep vibes from it. So I wonder if they just gobbled that up into BigQuery, put a lot of AI on it, put a lot of AI on it, and yeah, you could, it put those suggestions forward of like, yeah, there’s too many nulls here, or This is long, we’ll filter out, get rid of that or join this together.

[00:35:42] Matt: And you can kind of schedule that and have transformations that look pretty interesting and cool as well. 

[00:35:49] Dara: Yeah. 

[00:35:50] Matt: One of the, one of the, 

[00:35:53] Dara: one of the, one of the points about this, we said very catchy or autonomous data to an AI platform. One of the points that they were kind of calling out was that they’re making, quote unquote, they’re making unstructured data, a first party citizen, which sounded great, yamo for quality, you know, like it’s, you know, it’s about, it’s about time.

[00:36:18] Dara: Unstructured data got the, you know, the, the, the, the respect it deserves. No, because of the structured data in a corner. Exactly. Exactly. but I didn’t fully, I mean, I dunno if you, if you got a chance to look much into that, but I, IO one bit of a demo I did see was where, she was basically accessing, it looked really cool.

[00:36:39] Dara: So she was accessing like normal kind of table data, but then also PDF PDFs from I think a sales system or something, and she was able to merge the two together and then get it to extract information outta the PDF. And she was doing this all in, I think maybe it was Canvas or it was just in. Regular query?

[00:37:00] Dara: I’m not sure. Can’t remember now.

[00:37:01] Matt: I think this is, and I might be wrong. I’ve got so many just headlines written down. Yeah, same. Yeah. Of, of like whatever they’ve named a thing. But this, there’s, there’s, there was something they announced called BigQuery multimodal tables, which allows you to have structured and unstructured data within the same table.

[00:37:19] Matt: which sounds like what you’re describing there, where you could, that you could show, I, I think I remember seeing it like images, text, it was all just in this, in this one table, which I guess is very useful for multimodal ais to make use of, ’cause they don’t have to just deal with structured data.

[00:37:37] Matt: They can also understand what’s inside of an image. They can understand what’s inside an audio clip, what’s inside a PDF. So yeah, that did look it, it’s it all. Fairly substantial. 

[00:37:48] Dara: Yeah. And again, it all, it all looks so easy when they show you in the demo, but like, what is that gonna depend on what information you’ve got in the PDFs or, and then if you start mixing like a different, you know, let’s say it’s the same, let’s say it’s an invoice or something, but you used to have a different format of invoice.

[00:38:04] Dara: Is that gonna mess it up? Like, how, how, how safe I are these things gonna be to use, or, you know, ’cause they’ll, they’ll have just taken, you know, as you would with a demo, you’re gonna check in advance and make sure, okay, that information is in that PDF, these PDFs do work. It’s not gonna come up with some random error.

[00:38:20] Dara: When I try and, you know, merge these two different data sets together, God knows 

[00:38:26] Matt: I’d, I’d, I, I, it’s, it, there’s so much stuff that we do, that’s released in BigQuery on a regular basis. Plus the stuff you have to just do on a daily basis as a job. It’s hard sometimes to get. To get in and get stuck in, and explore it all.

[00:38:40] Matt: And I think that this, this time, this time in particular, I don’t necessarily remember that being this much around BigQuery before I might be wrong. but it does feel like it went all in this time on, on BigQuery. so I’ve got, I, we’ll find out if it’s safe or not in two years time when we finally get around to looking at it.

[00:39:01] Matt: And it’s completely, it’s completely, it’s completely, yeah. 

[00:39:04] Dara: Yeah. But it makes sense given what you were saying earlier, like the, you know, the, the them really, really gone to town on BigQuery. it, it’s like that, it does feel a little bit like that’s at the center of it all in a way, doesn’t it? It’s like that’s the linchpin and it’s, it doesn’t actually matter the whole thing again about it being multi-cloud and all the rest of it.

[00:39:26] Dara: It doesn’t, it doesn’t, it doesn’t really matter, does it? It’s like that this is the thing that’s gonna get you, ’cause they said something like, and they were being really kind of. Vague about it, but you know, it’s obvious, isn’t it? But they said that, you know, BigQuery has got five times the number of customers than, and they just said the other two leading.

[00:39:44] Dara: Yeah, I found that confusing. ’cause I was thinking like AWS presumably is bigger, but Yeah. But so they said data, what they say Data warehouse only or something. So I thought maybe they all include very specific wording. So it’s probably these two random, two random kinds of competitors that aren’t the ones people would think of.

[00:40:06] Dara: But anyway, the point I guess is that they’re, they do seem like this is a key area for them, isn’t it? To, to get people. So, putting Gemini into the middle of it and allowing you to, you know, mix different formats and all the rest of it, like, it, it seems like a bit of a no brainer, doesn’t it?

[00:40:23] Matt: It’s, I, I, that’s why I think that’s why they’ve got such a big advantage, especially from an, from an enterprise perspective, like for so many. I mean, when did three point t 3.5 come out? Like two and a half years ago? Something like that? Three years ago. and it’s all very much trained on whatever data and you just kind of tap away black box of data.

[00:40:43] Matt: It would give you relatives, it would give you answers. And what Google seemed to be really pushing at is to just make, just to get enterprises, encourage enterprises to stuff their data into these models and activate their data with these models. I think that’s half of what the on-prem firewall, air locked, whatever we called it, stuff is about for them to be able to pump data into their own hosted models, the Model Guard and being able to host your own version of Cloud sonnet in the same place, your apps being deployed, and having them just betting big that BigQuery is the center of all of this, because that you can’t date a, ai, M’s.

[00:41:24] Matt: ML modeling is all completely useless without data. Data. so. They have the best, well, I, I might be biased there as somebody uses BigQuery on a regular basis, but they’re one of the best, if not the best data platform around in terms of data warehouses. 

[00:41:43] Dara: Presumably, I, I only just thought about this now, but presumably all the, you know, to my untrained eye, I could look at all the stuff they’re doing with like the, agent space and the, you know, agent development kit and all that stuff and think, oh, that’s very separate.

[00:41:58] Dara: But I guess it isn’t because all of that, if you, if you’re building agents with, in agent space and it’s all using your business data, all of that’s sitting in, in BigQuery. Yeah. And then you’ve got Gemini running through the whole, the whole 

[00:42:12] Matt: thing. Yeah. All of the, all of the stuff in like the agent builder, the agent space, the, the, the enterprise tooling and automations that you build can all be based off of your hub.

[00:42:26] Matt: I keep having to look at my page to remember what the hell he called it again. Autonomous data, AI who do, yeah, it is all based on that data and joining structured unstructured and getting data to clean it. It all makes sense. It’s almost like they know what they’re doing over there. You’d think so, wouldn’t you?

[00:42:47] Matt: Yeah. There’s a couple other things as well that just popped into my head that we haven’t mentioned, which was the vector Vector search. Yeah. Yeah. Pro vector search. I believe it can identify where there’s similarities in the data or duplication and things like that, that’s what I took from it, is to be able to scan across the entirety of your BigQuery data and, and identify things.

[00:43:15] Matt: I suppose vectors for people who aren’t familiar are used heavily in. AI and, as in rag retrieval augmented generation, so pulling out embeddings and vectors to ground responses from LLMs in, in some sort of data source, some sort of reality. So maybe you can imagine having some sort of agent looking at all of your data and using vector search across the whole thing.

[00:43:41] Matt: I don’t know, but pretty cool. And then the other one is, my last one maybe, was the continuous queries. Oh, I don’t think I, I don’t think I saw anything about that one. 

[00:43:56] Matt: So they, this, this came out a little while ago, but I, they, they mentioned it again, I don’t know if they’ve updated it or they’ve, you know, sate Gemini to the side of it.

[00:44:05] Matt: But it’s the idea of having BigQuery be a real time data source as well. So it can ingest data in real time, but it can also, Undertake transformations and modeling on that data in real time to then be providing insights in real time to people for reporting or for ML or for whatever. a lot of you used to be able to, well you still can do this stuff with like, data flow where you’ll be sending data through and transforming it and there’ll be like a real time pipeline of getting us through.

[00:44:37] Matt: But you can do that in BigQuery now with this continuous BigQuery continuous queries trademark. 

[00:44:43] Dara: Oh, such catchy names, aren’t they? 

[00:44:45] Matt: Yeah. They all start with BigQuery as well, which is weird. 

[00:44:48] Dara: Yeah, maybe they can just drop that at some point when, when everything is BigQuery they can just drop that. Yeah. 

[00:44:53] Matt: When, when, every shop on the high street and every assistant in your home is BigQuery. I think that was it. And we talked about BigQuery governance as well. Didn’t that, didn’t we? That which seems to be metadata generation was quite interesting. Annoying. ’cause I just built an app that did metadata data generation and column name descriptions and stuff like that. But it seems. Nick that idea.

[00:45:14] Dara: So it’s always the way Google doesn’t, don’t ever, don’t ever build any apps for improving No. Any Google existing. Don’t ever improve. Yeah. Don’t try, don’t try. Just, just don’t try. Just don’t try. Yeah. It’s the moral of the story. It’s the message of the podcast of today. 

[00:45:30] Matt: Yeah I think that, almost certainly we’ve missed, but have you got, have you got anything that might, I missed on the BigQuery side. 

[00:45:37] Dara: And I did have something I just came to know, when you ask somebody if there’s anything else, that’s when it, that’s when they forget it.

[00:45:43] Dara: So that’s when they tap me. So it might come back in a second, but I did wanna ask you about the, the, the meta, what did you call it? The meta gender, auto meta data generation? Is that what you call it? 

[00:45:56] Matt: Yeah, yeah. 

[00:45:57] Dara: BigQuery, ETM. Is that another area though? Where is that gonna depend on the quality of the, like how good a job is that gonna do?

[00:46:06] Dara: Are we back to the same issue of if you don’t have. Good quality data, then it’s basically gonna make the wrong inferences and it’s gonna generate the wrong metadata potentially. 

[00:46:19] Matt: Yeah, I mean the one, the one that we created that measure up, like looked at say five, five random rows of a data set, the column titles and things like that, and it was, it would take a guess at what was in that column, write a column description, and importantly have a human in the loop to say like, this is what I think it is.

[00:46:40] Matt: Are you happy with this? Do you want to sort of validate it before it would then go and write it into BigQuery? I didn’t see a huge amount of, I suppose there is a human in the loop in terms of you having a conversation, but like you said, it just went off and started doing things. Yeah. At that point, it didn’t feel like there was a huge amount of humans in the loop there.

[00:46:58] Matt: So I guess it all depends on if it’s just going and doing it or if it has got that kind of, you have thought about those checks, cross checks and stuff, because I mean, some of this. We joke about it going off and doing its own thing, but on an enterprise data set, like how, how do you make sure it doesn’t, I dunno, query some raw data and start doing some gnarly stuff on it and you rack up a 15 grand massive cost.

[00:47:25] Matt: Yeah. Bill and I, I assume there’s firewalls, but they didn’t really talk about them, as far as I remember. 

[00:47:32] Dara: No, I guess that doesn’t, that doesn’t sit, that doesn’t in it, yeah. Yeah. That, that, that just follows in the, in the detail afterwards, doesn’t it? 

[00:47:41] Matt: That’s the two, that was the 2:00 AM sessions. Just loads of nerds talking about their firewalls.

[00:47:47] Dara: Yeah. Yeah. But, but, but all but, but all of that kind of taking that even a step further, like the, you know, with the agents, like when, when, when people end up building these kind of multi-agent systems that are, and they’re truly, ’cause I think it’s, is a, is it fair to, this is something I picked up in one of the talks as well.

[00:48:05] Dara: It’s like they were talking about how agents at the moment are. They’re almost like slightly glorified assistance. They’re not really being that proactive, but the idea for them in the future is that they will be kind of, they’ll, you know, they’ll move to semi-autonomous and then they’ll be fully autonomous and they’ll go away and they’ll figure things out and they’ll update based on new context and new information.

[00:48:27] Dara: So how complicated and scary is that gonna be when your business has these multi-agent systems going off and activating and you, you don’t necessarily know what’s happening. You know, you’re, you’re basically handing the keys over to, to ai. 

[00:48:43] Matt: Yeah. Maybe, maybe it’s, maybe, no, maybe it’s just convincing people that everybody else is doing that to force up just to give you a fear of missing out and everyone just runs and starts.

[00:48:51] Matt: Well just do us Yeah. Automated data at things and Google makes months, tons and tons of money. You hope, I mean, we’re not, like you’re saying half the stuff is coming soon or not out yet. You hope that, you hope and think probably there are. There are checks and balances and, and things you can do to make sure it doesn’t go crazy.

[00:49:14] Matt: Like with conversational analytics, for example, I had to play with it earlier ’cause you can, you can add agents that exist. Conversational analytics, you can add agents in that and you can specify the tables that it has access to and give it instructions and tell it what’s in each different field and things.

[00:49:30] Matt: So, but, but like you said, people could just be randomly playing with it and, and there might be too much assumption of people thinking things through and doing best practice and people are gonna come, come across when they don’t do that and just start pointing it at, at big old data sets. But maybe that’s true now before ai probably plenty examples of, of some, some, some wet behind the ears analyst querying the wrong table and costing the company a fortune.

[00:50:00] Dara: I can see that now. So I. Just maybe a lot easier to do it in the world of agents and Yeah. But then again, if there is kind of cost controls and you know, if there’s kind of layers in there to, to kind of protect against that, then you’d hope that that would, you know, ’cause a lot of these that would, you know, there was talk about like some of these agents, some of the folks in these agents is already around, not just kind of cost optimization, but some of ’em are around security as well.

[00:50:26] Dara: So it’s like, you could have agents that actually check. So you could even have an agent, this is getting very meta, but you could have an agent that checks what the other, you know, a supervisor agent that’s checking what the other agents are doing and then alerting somebody if you know, like the agent police.

[00:50:43] Matt: Yeah. That absolutely, I, I’m sure that is like a, a genuine theory that they’re trying to put in place for like these big agents that they’re scared are gonna become a sentient, like an agent to check that it’s, knock it on the head with a wooden spoon if it starts to become sent behavior yourself. Stop it.

[00:51:00] Matt: Yeah. I think that to me, all the whole thing is massive and fascinating and places scary, but it, it, it reiterates, reiterated for me and highlighted for me how much a growing need there is to get good, robust data into these places. And, and that’s a real opportunity for people like us and, and other people to help people get that data into the right shape so they can take advantage of all these, these new tools that Google are trucking out, left, right, and center.

[00:51:30] Matt: So it was, it, it, it’s, it’s reassuring in a way that all of this stuff, all these announcements will be better with better data. I’m excited and optimistic and, and think, yeah. Next, 

[00:51:43] Dara: Yeah. Yeah. I think I am too. and I did remember the question I was gonna ask, which wasn’t, it wasn’t about something that you’d miss, but.

[00:51:51] Dara: It’s thinking about this, like vector search and knowledge engine and this, the fact that now you can have a mixture of structured and unstructured and all different modalities and whatever. It made me, maybe this is a naive question, but I, my thinking would have been up to this point that businesses would usually use BigQuery for, you know, tabular data and that would’ve been the main focus.

[00:52:18] Dara: You wouldn’t necessarily, although I’m sure there are lots of businesses that probably do do this, but you wouldn’t have all of your assets in, you wouldn’t have like images and video files and text files. You wouldn’t have all of that information stored in Google Cloud storage necessarily. But now there’s gonna be more and more of a reason to have all of your business assets potentially, unless I’m misunderstanding this, but if you have everything in Google Cloud storage.

[00:52:47] Dara: And you can access all of that information through BigQuery. And then on the gen AI side, you can output anything you want in all of the different modality. I’m saying modality. Am I just making up words? It sounds really good. Keep saying it. I’ll keep saying it, whether it’s right or art. So you, you know, you could, you could query, you know, it opens up the possibilities of, you know, you could do some analysis there and then saying, you know, identify the customer types that have the highest average order value and profile them into a customer segment.

[00:53:21] Dara: And then, and then allow me to create a customer ad campaign using the right assets that, you know, basically you could create a custom campaign and target people and activate all of that. I’m not saying you can do all of this now necessarily. Maybe you can, but certainly that’s gonna become a lot more commonplace.

[00:53:37] Dara: So there’s, there’s reasons to have more and more of your business data. Which again is, well, it works out great for Google, doesn’t it? For Google? 

[00:53:45] Matt: Yeah. Yeah. I mean, to, to a certain extent. Like it used to be that you’d have, you’d maybe, maybe cloud storage would be a, a, a middle point before it went off into BigQuery.

[00:53:57] Matt: Maybe you’d have, that’d be where you, you dump all of your raw data and then you have a transformation layer. But then recently, particularly with semi-structured or structured data, you’d be dumping all that into BigQuery straight away and then using a data form to transform it and just use the power of BigQuery.

[00:54:11] Matt: And it almost feels a little bit unstructured. It’s like that Chuck, you’re unstructured there and there, just chuck whatever, Chuck the bin in, take the kitchen, sink, chuck it all in. Yeah, it was some big wheel out autonomously for you. 

[00:54:23] Dara: It’ll change it into ai. Yeah, I mean, what I took from the next 25 is just put everything in there and then you can do whatever you want without knowing any code and not worrying about governance or security or anything else.

[00:54:35] Dara: There’s no problem. Nothing can ever go wrong. Happy days. Hey Presto, ai. Pops out the other end of the conveyor belt. Dear Gemini, make my business better. Thank you. I’m off. Happy days. Yeah. Okay. Any, any, any kind of final thoughts or, I mean okay, well here’s a question for you which might kind of help you with final thoughts.

[00:54:59] Dara: Like what, so as you just said, and I found the same, you watch it, lots of, it’s really exciting. Some of it’s kind of directly relevant and applicable. Other stuff is just, wow, that’s kind of cool. That might be useful for something at some point. What are your, like what are your practical kind of thoughts or either your own or what would you give as advice to somebody who works in the data and analytics space based on these updates?

[00:55:24] Dara: Like, what would you be, what’s got you thinking in terms of actual day to day, you know, what differences is this gonna make and how might this change not just your. Day-to-day job with somebody working in a company in a, in a data role.

[00:55:39] Matt: I, I think I kind of said it before, but I, I, I think for sure, don’t ignore them. I think that it can be a real tendency, whether it’s through fear that, that this is coming for your job or something and you just want to bury your head in the sand and pretend it’s not there, or it’s through just, just, you’ve been doing it for so long and it’s like a second nature and you just ignore it.

[00:56:02] Matt: I think that’s dangerous and, and that could potentially lead to problems down the line if you just completely ignore these things. So just exploring it and just feeling how it can maybe assist you in your day to day. Like look at maybe your data scientist, an unbelievable data scientist, maybe using the data scientist agent can save you an hour in your data cleanup process.

[00:56:24] Matt: ’cause it can just skip through a few steps and just do some of the really easy, repeatable stuff. Nice and simple for you. Similarly, if, if you’re, you’re engineering and you’re, you’re doing bits of, I, to be honest to you, data engineering one is the one I’m least clear on. Ironically, I wasn’t completely sure what it was doing, but it seemed to be able to do some pipeline stuff, explore it, see what it can do, see how, how it can make you more efficient.

[00:56:49] Matt: In the first instance, you’d assume it would shift the, the repetitive and the low hanging and, and give you the, the, the head space to concentrate on the stuff that really needs the human brain behind it. That’s from a personal perspective and I think from a practical perspective, it’s just making, just making sure that the data that’s going in there.

[00:57:15] Matt: Right. as much as possible, if you can own and understand the point at which the data is being collected and generated, and ensure that that’s correct, whether it’s to speak in our, in our cadence, having a proper, robust data layer, implementation and collection process. So, you know, by the time the data’s hitting BigQuery, it’s in at least some half decent shape.

[00:57:36] Matt: That’s gonna make your life a lot easier, and it’s gonna make these tools a lot simpler and, and able to be utilized better, utilizing more efficiency. So yeah, just, just understanding. It’s all from the, from the far left of the process to the far right, from the, from the generation of data to the activation of the data.

[00:57:58] Matt: again, that as robust as possible will make these tools much more useful and much less likely to cause your headaches. 

[00:58:05] Dara: Yeah. Did I wrap up? Yeah, good takeaways. Yeah, I think the, on the, on the, on your first one around like people fearing it’s gonna, you know, take their jobs or whatever, it’s like the, the, the phrase that I like is, you know, when they talk about toil and like reducing toil, this idea of like, this kind of repetitive, kind of boring work.

[00:58:23] Dara: And it’s like that’s, that’s really where it can add the advantage, isn’t it? It’s like, it’s not new . Well, let’s not get into whether it is or isn’t gonna take people’s jobs, but I think it’s like, if it can, if you can use it to automate and basically just delegate those repetitive, boring tasks that you do.

[00:58:42] Dara: A lot of it, you know, a lot of the examples and the demos, they were showing you what it was doing, that it was just using it to do some of the heavy lifting for you. So you don’t have to, you might even have the skills to do that, but that doesn’t mean you need to, you need to do it each time from scratch, does it?

[00:58:57] Dara: So if you can just. Offload that, get some of that heavy lifting done. and then as you said, you focus more on, you know, using the what, what you, what you need, the human brain, your own creativity, your own context or whatever. Just apply that on top then. Alright, well maybe we’ll do a, I think we’re probably, that feels like a good kind of natural stopping point.

[00:59:17] Dara: Maybe we could go on for probably another six hours and still not cover everything. But maybe we do, when the dust settles on some of this stuff and some of it’s maybe available to us. ’cause some of it is very much like, oh, you will be able to do this, but you can’t, not yet. So maybe when we’ve had a chance to get our hands on some of this stuff, maybe we do a bit more of a deeper dive on some of these individual features or on a group of ’em.

[00:59:42] Dara: You know, if there’s a few that kind of fit nicely together, we could do a bit more of a deep dive. 

[00:59:46] Matt: Absolutely. Yeah. I think it was. Tons of exploration to be done, but it feels like a pretty rich vein. Yeah, rich vein of content, which is what you want on the podcast book. Absolutely. Yeah. That’s what you need.

[00:59:56] Dara: This will keep us going for a little Months Stein out on this. Yeah. That’s it for this week’s episode of The Measure Pods. We hope you enjoyed it and picked up something useful along the way. If you haven’t already, make sure to subscribe on whatever platform you’re listening on so you don’t miss future episodes.

[01:00:11] Matt: And if you’re enjoying the show, we’d really appreciate it if you left us a quick review. It really helps more people discover the pod and keeps us motivated to bring back more. So thanks for listening and we’ll catch you next time.

• Spotify
• YouTube
• RSS Feed