Brainfood Live On Air - Ep407 - Disconnected AI vs Multi-Agent Ecosystem: How to build a TA Tech Stack for the Agentic Era

0:28nd you try and pull it back and We've done loads of these shows. I think we're now getting over 50 a year. We're probably at 60 or so. So we've got a few midweek ones happening. Next week, we've got three Brain Food Lives happening. So it's getting popular. Can we get one a day? Do I have enough energy for that? Probably not. But we're on our way there. Folks, welcome to the show. Let me just quickly do some audio and visual checks to make sure everyone can hear and see me okay. If you're watching this on Crowdcast, do let me know whether the audio visual is fine. You should see me and myself and our new co-host, Paul Bacon, on screen.

1:10Let me know if you can see him. And we should be broadcasting this on a number of LinkedIn as well, which I'm going to just check on my phone right now. Yes, you can see me on LinkedIn there. I hope the audio is fine as well. Let me know in the comments, by the way, if you can see and hear me on LinkedIn. Jovio, our sponsors today, are also broadcasting this show by their company page on LinkedIn. So if you're watching this on the Jovio company page, which I think you've got like a million followers there, it's like somehow acquired an enormous audience. Welcome to all of the fans of Jovio. Hope you're going to have a good time.

1:44Okay, let's get on with it. Let's introduce Paul. Paul Bacon, wonderful to have you on the show, mate. Who are you? What is it you do, Paul? So thanks, Hong, for inviting me. this episode of Brain Food is sponsored by Nescafe because I just got off a plane after being overseas. So I think I've been here, there and everywhere this week. So only landed the time up back in Heathrow. So I'm not sure if you're sponsored by Nescafe, hon, but maybe if the TA leader is listening, maybe you can start to be sponsored by me. I'm happy to get consumer-grade sponsors on here. Right now, it's a Starbucks mug.

2:29It's definitely a Starbucks mug. I think my mother must have stole this. But yes, I'm happy to get consumer sponsorship, mate. Yeah, nice. So yeah, I'm Paul Bacon. I'm the head of partnerships here at Jovio. So I've been doing this now for the last, well, last couple of years, basically. But I've been around the recruitment industry for 15, 16 years and spent a fair old chunk of my time at Indeed.com. so uh know the industry what i'd like to think uh pretty pretty well so yeah excited to be your be your co-host today hung um some might say the new ant and deck of recruitment let's roll it mate let's roll it i love adam gordon of course the standard and default co-host but you know what we're gonna mix and match it a little bit more like bts you gotta have a lineup you know people in and out of the of the lineup this is how it's gonna be um all right paul listen let's talk about jovio because jovio are a company that are also in the middle of expansion of scope you can say as some people might know jovio uh famously as hey aren't they the guys that do like programmatic advertising and so on which is still the case but there's more to it than that now in 2026 isn't it so tell us a little bit about where jovio are at in terms of what it wants to do and what kind of service it's provided to the market.

3:59Yeah, of course. Yeah. So I just thought in terms of for this show and podcast, and for what we're going to be speaking about a little bit later, I thought a pair of a couple of slides just to kind of like set a couple of how we're thinking about this agensic shift. And especially a topic that we're going to talk about today, is it single point vendor or is it multi-agents kind of like stitched together? so I thought if I shared a few slides with everyone hopefully that would get the kind of like juices flowing some of the some of the questions and conversations that we can have when we let let everyone in if that sounds good to you my friend yeah let's get the juices flowing for sure and by the way folks you know brain food live is live because I have no idea what's in these slides there's no it's like at your own risk folks I'm sure it's going to be yes entirely safe for work so here we go don't don't worry there's nothing uh there's nothing on support don't worry you see my screen okay yeah it looks good perfect stuff so yeah we kind of wanted to like couple of uh thought-provoking questions and how we're thinking about it um at jovio in terms of this agentic shift because some of the people listening in might think of jovio as uh the programmatic job platform, but we have over the last three or four years taken a massive seismic shift into full orchestration via our platform.

5:27But I didn't want to talk too much about what we did today and I wanted to kind of like where we're thinking like this agentic shift is ultimately going to go. And the starting point for us is that we've always thought that recruitment has always been people powered. And those sort of great teams that we've got out there has always been built on relationships, judgment, and trust. Now, the way that we think about it is that people powered is also meant that sometimes people are a little bit people limited. So this is, I'm sure that everyone listening in on the call today is ultimately what we want to kind of do with our recruiters time and talent acquisitions time is to be focused on the relationships and judgment speaking to particular candidates but because of the day-to-day so teams have the source they've got an outreach they've got a screen they've got a schedule by the time the 6 p.m rolled around in a recruiter's day is that whatever's left over is whatever's left for that kind of like more high value impact work.

6:39So what we think now with the advent of AI and agents is we want to make sure that agentic AI change of what people powered actually means. So those agents don't assist the work, they do it. So sourcing, engaging, screening, scheduling, interviewing, et cetera. And then all that time that's left over for recruitment teams is actually then going to be pushed onto more high value impact work. So this is kind of like your atypical funnel that you can see on top of the screen now. But in the advent of agents, each agent or agents that you could potentially deploy is then giving time back for those high value impact work.

7:25This is the way that we're thinking about it is all the time that we save with those agents that attraction, source, screening, schedule, and optimization, with all that time that those agents can take away, because it's actually just doing the doing, we talk about the 80-20 rule a lot, right? So we want to make sure that all of that time that we get back is pushed onto what we did, what everyone got into recruitment for, was human-to-human connections, judgment, and been able to do a fantastic job for those candidates. Now, what stays human, right? This is really important to us at Jodio, is judgment, relationships, and trust.

8:07So that's never going to go away, right? And we just think that in the advent of AI, making sure that those agents are doing the do and doing is said that we've got a lot more time that's actually pushed onto the more human work. And then what does that day actually look like? Right. So this is sort of the way that we're thinking about it. It's not sort of an exhaustive list, but it just kind of gives you an idea. So not having to do those mundane, manual, administrative heavy tasks anymore. This can all be reallocated to that high volume impact work. So rather than building sourcing lists, chasing applicants from first round phone screens, scheduling tennis, that back and forth, I'm sure that all resonates with the people on the call.

8:53What that is then left with is like having those great relationships with hiring managers. How do you actually get to read a candidate, not just the CV? How do we put that human to human connection back in? There's closing offers and negotiation, candidate experience and quality of hire and ultimately workforce strategy. So as we said, this is not necessarily an exhaustive list, but it's just the way that we're thinking about it here at Jovio. And then done right, we think that this is right. These agents that we're going to be deploying is going to make those connections more human, not less.

9:30Right. So every candidate is going to get answered. Every candidate gets a real shot. and then when recruiters actually do that human-to-human connection, they're not so beaten down with all that admin and all that stress, is that they can be fully present to have a true, meaningful conversation with the candidate itself. Now, one of the things that we're going to talk about today, Hung, is are we thinking about sins of the past, maybe coming into those multi-agent bloated tech stack? So when kind of like SaaS really boomed, obviously in the sort of the 2000s is people have those bloated tech stacks.

10:11So are we with a multi-agent kind of ecosystem? Are we potentially going into that world? And is those multi-agents, are they really truly going to speak to each other? Are they ultimately going to be able to have those handshakes so the left hand knows what the right hand's ultimately doing? Now at Jovio, we think that we've got this one connected agent connecting everything together. So to your point, Hong, yes, we still have programmatic job advertising, but we've now kind of moving, well, have moved to a full orchestration platform. So I'm not going to call out everything. And I just wanted this kind of like first few slides to hopefully be a little bit thought provoking in terms of kind of where we think the shift is going, that multi-agent ecosystem.

11:05But effectively, those four points in terms of what you can see on the screen. So we can one agent can conduct sourcing, outreach, screening an AI interview, scheduling madness and tennis. and also we can make sure that the humans actually supercharged with our co-pilot interview intelligence. So the idea is that we think having that one agent doing the whole orchestration is ultimately kind of how we're thinking about this because the left hand does know what the right hand is doing because it's natively integrated. So the idea is that, yeah, that only 20 % admin and then 80 % done on high value impact work.

11:54So for those listening, I'm going to put this into the chat when we get into the conversation. What we have is what we like to call our Builders Program. So for those of you that would rather build than wait, if you fill this out, you can be able to get early access to all our agentic hiring platform you get to speak to myself an amazing team that we've got so we can start to see how we can actually create that orchestration ultimately ultimately together so just wanted those few slides um hung to be able to uh yeah provoke some conversation for for us today and just give you kind of like a little bit of a sneak peek of what uh jovio has become is that that full orchestration no i totally appreciate you putting the effort in to to whack the slides together upon it actually really helps it helps to have a visualization to be honest with you so um so you're very useful very educational on the the product roadmap um great by the way i think one of the the downstream effects of ai is that it is going to lead to basically superior capability provided by vendors to the user base so we're going to be able to do more um it is going to supercharge us the question is um you know what does this agentic world look like in terms of how we build this tech stack um and that's one of the things that we want to discuss today um obviously jovial coming from a certain perspective to say hey listen um having a multi-agent single ecosystem might make more sense i think most vendors would say this um but then you know i'm interested to know from practitioner side um we're going to bring some practitioners on in a bit to help support okay what is the thought process behind this um you know particularly if you've got already like a multi-vendor type uh stack um how you know it's a big chew to get rid of it all in one shot is that something that's something you want to do or is it a piecemeal vision to go forward a question actually for the audience real quick um where are you currently with the number of tech tools you're using have you audited this um and what's your journey over the next 12 months like do you want to decrease that if so by how much percent um and what is the end point uh do you have a feeling as to what your end point should be with regards to the numbers of vendors that you were actually signed up to let me know in the comments on this what your thoughts are.

14:25And in fact, I've actually got this as a poll. So let me address the audience right now. Tech vendors consolidation, including LinkedIn, how many TA tech subscriptions are you aiming for in 2027? So this is a question for everyone who's watching this right now. Tech vendor consolidation, how many subscriptions are you aiming for in 2027? none is one of the answers zero you're going to vibe code the entire thing one you've got an ats and then you build everything around the ats two your ats maybe a talent attraction component ahead of the ats um three um and then maybe more than three um let me know in the votes where you're at with this folks i'd love to know excuse me i'd love to know what your vision is and what your ambition is uh and your thoughts on that that'd be fantastic okay good stuff no more than two from Christian at zero currently from today very very interesting okay let's bring on some guests here I want to bring on Mary Kay I was going to bring on Victoria Murphy as well but unfortunately Victoria is ill and not able to join so it's just going to be Mary Kay unfortunately right now not unfortunate at all because Mary Kay is absolutely awesome however if anybody else is up for joining the conversation do let us know in the comments I'm going to actually make the gallery open in this way.

15:44Yes, here we go. Open. Anybody can join the gallery. So if you want to come on to the show and you fancy it, do let me know and I'll pull you up there. I think you're all on defaulted on mute. But one of the wonderful things about Crowdcast is that it is a live stream. So we're happy to get people on if they're up for it. Or Mary Kay, I'm going to bring you on. I think you need to stick your camera and audio on if that's okay. And then we can at least have the practitioner perspective on this. Whilst we get Mary Kay set up, Paul, I wonder whether you can bring us some information as to what vendors are saying to you when you go with them for a multi-serve agent, single ecosystem approach.

16:29What's their cause for hesitation? What are the reasons why they don't immediately say yes? Yeah, I think sometimes it's just like the fear of putting everything in. in the same basket, right? Is that that's always like a little bit of trepidation for some customers. But I think it's the, I think kind of the way that sometimes you look at it is just that you have in your personal life, you have single point solutions, right? Because it's doing the job and it's doing that job extremely, extremely well. So I think once sometimes that they see the the handshakes that happen within the and the benefit of having that kind of like single vendor ecosystem then they they did really kind of like resonates um because especially with like i mentioned sins of that past is sometimes where they've had friction right and i can see mary smiling so um welcome mary k but yeah i'd love to get your as a obviously as a practitioner to get your perspective on that yeah yeah but before we do that hang on am i losing people or is it sort of uh i think it's my internet isn't it oh dear um i can hear you though it's all good

17:50oh dear i think it's my internet go ahead go ahead mary no i i was just going to say that um there's so many factors to consider, right? So there's, there's the value calculation of, you know, is moving to a single solution, A, is it going to work with my core data set? So in a large, complicated organization like mine, where we have big ERP kind of technology in place, like, that integration and ability to really layer on top is critical. And then the second piece for me is, you know, we've been working on AI enablement for a while now. And so we've already got some things built in-house that work really well for us.

18:46So shifting away from something that was totally custom to an AI solution that kind of makes us move has to be worth the trade-offs. So it's constant calculation. I think I love the idea of being able to have fewer tools, but I consider more, you know, does what we have create a seamless experience for our users, particularly our candidates, our hiring managers, right? Are they getting a seamless experience? which a single source of things points to seamless right and sorry Mary Kay we probably should have asked before you jumped on but just for everyone's context sorry can you just sort of mention um who you work for what you do and kind of like absolutely I can't introduce myself so no worries that's all right Mary Kay Baldino um long-time listener and fan of Hong I've been on some of these streams before.

19:54I lead global talent acquisition at Morningstar. There's multiple companies out there named Morningstar, but ours is the one that provides independent investment research and data to consumers and companies. And so we are a, since we are a large global complicated organization with seven different business units. We're very dependent on our infrastructure, which is Workday for human capital management. And so since I came on board a little over 18 months ago, we've been working to solve about a hundred different areas of waste in our processes, just a lack of automation, integration of ai um you know and i loved what you said about you know freeing up recruiter capacity and capability to focus on the value add work because that's exactly what we're aiming for looking at every single day what are they where is my team wasting time and energy that could be put into relationship building is it's really important to me yeah and it's the um it's the reskilling as well that you can potentially do with that because the IKEA just totally out of the recruitment space at the moment but IKEA that they asked their customers is like what would you want from from IKEA right and that was more interior design so with all the time to save through that through AI you could then ask that question well what could we do with that time and where could that be displaced in order to then that higher and higher value impact impact work.

21:42I totally, I totally agree. In terms of a question for you, Mary Kay. So you mentioned, you mentioned that it would have to be, it would have to be like really compelling, right. In terms of something that you've built in house. Could you talk about like what you're, what you're thinking in terms of like, what does that mean in terms of that? Not just wow, but what would it have to do? What it'll have to in your world? Yeah, I think that in the morning starts made some pretty big investments in giving everyone in our organization access to AI tools, right? Lots of people on my team have multiple access to multiple AI tools, But we're largely a Microsoft environment.

22:34So everybody in the entire organization has co-pilot licenses. And so it's been very easy to build agents in co-pilot studio and make them scalable at an enterprise level. So that would be kind of one thing is just what is the, we're leaning on existing infrastructure right now. So what is the reason to switch to something that is, you know, adding another tool into our existing infrastructure? Because it's built on, you know, in Copilot for us, it also allows us to pull in very easily a lot of supporting documentation, a lot of access to, you know, email strings with hiring managers, all kinds of other things that are super useful in continuing to refine the output.

23:32So accessibility, you know, who can use it, all of those kind of things are really important. And then just come. Yeah, go ahead. Come in on this, Marika. It's really interesting because Microsoft, you know, are they a competitor to every SaaS tool out there? Because basically you are using either Microsoft or Google, right? But let's use Microsoft as the example. suddenly they were all our co-pilot and they're saying okay, you know build whatever you want and then probably you could build a CRM or something sort of with Excel right, and it's like okay, so I remember back in the day, way back when I don't know whether anybody was a recruiter back in 2000s, that kind of early noughties period you were there, Mary Kay I was there and actually it's almost returned to type because that was pre-SaaS, it was pre-recruitment tech really.

24:32Your database was Microsoft Outlook. Like you would create candidate database off Outlook by creating folders and just stuffing it full of CV documents. That's right. And I wonder whether it goes all back to that. What stops that from happening though? Because I got a feeling it won't be that way. um uh like do we want sort of outlook to be the the place where everyone just recruitment work sort of exists or you know why is it that microsoft haven't created for instance a reasonable crm i dare say um or why is it they haven't created like an ats what's your theory on that uh mary kay or indeed paul bacon any thoughts yeah security honestly like the the ability to kind of guarantee a certain level of data security, data privacy, et cetera, is super important.

25:31Be able to produce the kind of reports and compliance and insight into what's going on. You know, I wouldn't be willing to build an ATS or CRM in, you know, Microsoft Copilot. Not today. I don't know if we'll get to a point where I trust the data security and data privacy enough to put all my data there. The things that we're doing in Copilot don't require any individual PII kind of data. we keep that in applicant tracking system and use embedded AI within the ATS environment to interact with all of the data privacy data security that we need but compliance would be my answer to that number one yeah so the users is currently the team currently using Copilot in your team, Mary Kay, would it be fair to say they're using AI agents in almost like an individual optimization type manner where, you know, they produce a small app, but perhaps that agent, they'd use it purely for them to get more efficient, but it's not necessarily something that rolls out to the entire team or the entire organization.

26:48That's where we started. That's what we started with, Hung. We started with, you know, everybody had access to tools. And so everyone was experimenting, right? And they were figuring out like, well, what can I do with this to help me be more efficient, more effective, to spend more time on the things I want to do instead of the things I have to do. And so, oh, I saw Lotus Notes in the... Yeah, I saw that. I had to write my... Sorry, that was a distraction moment. Anyway, that is where we started. We started with individual productivity and people were building some really cool stuff. So we kind of did an inventory of what everybody was building for their own use and picked the best two or three use cases.

27:32And I said, you know, to the team, like, let's make these scalable. Let's make these enterprise grade. And so we did that and went through extensive testing in order to get to a launch stage. But if somebody is doing something really great, I want everybody on the team to have access to that. So it started with individual, but we're now making scalable enterprise great. This is another question I want to bring before I bring Paul in, I want to question the audience again. Are you at the point where you are kind of sharing the individual optimizations, the individual agents into a kind of a sandbox that other people in team can then use um or are you creating a leaderboard of what are the great sort of uh ai automations that have been created and then ultimately deleting the ones that aren't the best and and isolating one one that's better i had a chat who's the director um into it good to bring on this call actually but she mentioned that's exactly what happened it was complete chaos everyone said look license everywhere everyone building like it was like you know half a dozen jd builders uh in in the team and it was like okay what we're going to do is rationalize it throw it all together we'll all workshop which one is the best and we'll just select the one that's the best delete the rest and say look this is now the standard so you've kind of pushed the build to the team.

29:04Yeah. Oh, God. Am I out? No. You're going to have to run with this. The internet is bad here. This is no bueno. But yeah, I echo what you said, Mary Kay, is that it's all the making sure that it's all security and governance, right? But I think the other side of that is everybody, we've all heard of Vibe coders, right everyone now is a yeah because you are literally being able to take take what was previously really difficult and was only the the people that have gone through that that extensive um extensive train over time and now all of a sudden with a with a laptop and whatever pick your pick your uh your model then you're suddenly then a then a vibe coder so the thing is is then do you become just so bloated with all of these all of these different tools where there's a tool to create a tool to create a tool and then it's just like a bit of a a bit of a chain a bit of a chain reaction but I think it's then understanding kind of like which like which tool is obviously going to be going to be best placed and and how that's going to be effective to the organization now I also think it very much depends on the on the type of business that you are right and I also So sometimes I think, and I'd like to get your interpretation of this, is like, could some of this be a distraction, right?

30:35Because if you're building tools, then to create like better processes, et cetera, is that a distraction of like maybe what your current day-to-day is? So it's, yeah, it's, I think, because of that, and I'm guilty of it, right? I love it. So my wife would come out with me because she's like, what are you doing? It's Saturday. and I'm just like, oh, I'm just building something with my best mate. It's kind of intoxicating because when you build these things, it's actually like, wow, look what I've done. But I think sometimes it's just that if everyone does that, then you create more noise, more conversation, and not anything that's relatively streamlined.

31:16So I don't know. It would be good to get your interpretation of that, I guess, Mary Kay. I'm thinking of a particular colleague. Who is so interested in AI. And he spent an entire evening building this AI animated video to introduce a concept. And I was like, what are you doing? How many credits did you burn? And this isn't like a sustainable thing that you built. So he was just really into it and was enjoying himself. It was like a game almost, right? And we have been on this learning curve. At Morningstar, we actually have an AI academy. And now we have like a 201 version of our academy. Almost everybody on my entire global team has been through our AI academy.

32:16So we went through a phase where everything was very much, everyone was vibe coding. Everyone was experimenting. Everybody was trying new things. And that was great because we learned a lot about what works and what doesn't. And what I saw was a lot of people, you know, jumping straight to AI as a solution before they'd even define the problem. So we've gotten a lot more discipline since then, and we've actually invested in a new role within our people and culture team that is entirely devoted to helping enable our entire organization to transform people and culture work using AI tools. Um, and so, you know, having a focus, somebody focused on this as their entire job is really transformational.

33:15It's really, um, really great because the work of making sure that you're thinking through all of the things around what's the right tool, what's the right platform. Do we buy enterprise grade software? Do we build our own agents? all of those things get thought through with somebody with real expertise so uh yeah that that's that's really interesting could you um do you have like obviously don't go into too much detail but like have you built like almost like a framework and like um we have like what like in terms of like like cliff note version this is like do you have sort of like key stages of like well we might want to do x i'm going to bend it through this lens and then what comes out the other point.

34:00I think it'd be really interesting to understand how you've done that and what framework you've put around it. Yeah. Well, so we do have a framework and we have a governance framework. We have a framework around looking at how we look at processes to determine what is just automated, can be just automated using, creating better workflows. What steps can we eliminate? So I have a Lean Six Sigma background. I love this kind of, I geek out over all the process pieces. And then what are the things that are really much better candidates for AI enablement? So where do we need to layer an agent on top of a knowledge base, things like that.

34:47So we do have a framework around that. And then we have, you know, documented testing protocols and things that we have set up as well. And we're very fortunate and we have strong support from our CEO and our board to really stretch and be experimental. But we also have a really good partnership with our legal and compliance teams around how we do that in a way that is responsible and doesn't risk people's privacy or our data security and integrity. interjection for me here mary-kay um i mean you've you've introduced some really interesting ideas here that probably sets the scene for second sort of other episodes of brain through life so for instance how do you encourage ai experimentation ai enablement amongst your team but how do you then rein it in a little bit um because as you mentioned with one of you one of the members different people are going to be into it but you might end up people i end up sort of having individuals that have literally gone off the deep end and as you mentioned Paul it's like they've actually come away from their job now um and they're doing something else like how do you manage that that's a very fascinating topic and probably it's a management skill that we're going to have to develop isn't it it's like you've got to hit the overview how do you do the development but how do you also maintain focus very very interesting challenge um and secondly you know how you've described the journey is is kind of like it's brilliant how that's kind of matured And I wonder whether it may require that initial messiness to get to where you're at now.

36:24You know what I mean? It's almost that you've got to give people the tools, go through the pivot, fall off the bike a few times, I guess. And then eventually you're going to figure it out. And you can't just study a book and say, right, I now know how to ride a bike. It's true. You know, it's not a theoretical exercise. you've got to get on it and then and then see who wants to go through the practical uh sort of use thereof interesting okay let's try and re reconnect us back to the main topic which is this concept of disconnected ai so we're looking at multiple agents um running free so to say versus single ecosystem pros and cons now i guess one of the the main things we've got to think about here is are the handshakes ever going to be good enough i mean if if so then we might end up in a situation where yeah you could vibe code or whatever you want and then you know because we're reliant on sort of the connections being so good that actually they can carry context through wherever they need to go um where do you versus do you need actually single vendor to maintain that context anybody in the audience have a view on this um i want to give mary kay a shot at this as well.

37:37I know you're not a technical expert, even Mary Kay, even though, you know, as a non-tech person, you probably don't ask anybody. Yeah. So I don't want to put you in a position where you have to make comments outside of expertise, but just from your own sort of, like, where do you think we end up? Do you think we're going to end up with where the connections are going to be good enough? I think we will get there. I do. I think that, you know, we're still early enough in the in the use of AI tools, that companies are going to get more and more sophisticated about this. And customers are already getting more and more demanding about how are these things all connecting together.

38:22And the handoffs are, as you said, the handoffs are the really important part. So I think that, you know, for me, I'm looking for, you know, keeping everything connected in one platform if it contains, if it relies on or contains, you know, PII or sensitive data and things that don't require that. I'm okay with having some less connected agents and we're sort of trying to take the best in breed ideas within the team and make those scalable at an enterprise level in that regard. But I do think it's really in talent acquisition, you're dealing with so much sensitive information, so many compliance things.

39:10I think it's important to build from a core and keep everything integrated. Yeah, absolutely. Because if, yeah, we've seen what's happened in the past, right? It's just that, yeah, people, I think in this day and age, like privacy and protecting that is absolutely paramount. So, yeah, that just said in the channel, Hong, that, yeah, there needs to be a big investment in workflow engineers and making sure that those connectors and functional automation specialists make sure, yeah, that makes a ton of sense. How do you connect those things ultimately together? I think, yeah, Andrew, yeah, protecting those processes.

39:56Absolutely. And what would you say, Mary Kay is like... Because I guess sometimes what are those processes do you think needs to be protected from questions of riffing off that? Yeah, I mean, it certainly at least starts with sourcing. So it's the earliest stages where you're bringing in candidate data. Right. So at a minimum, it starts with sourcing. I mean, there's lots of things that you can do with AI that are even pre that stage, you know, workforce planning, all kinds of other, you know, planning and organizing kinds of activities can be done with AI assistance. But at a very minimum, you have to have that kind of security and infrastructure and compliance kind of mindset once you start bringing candidate data in.

40:59Yeah. Yeah. And then it's the entire rest of the lifecycle, right? You don't do anything. Yeah. Without candidate data. Yeah. Do you think there's an option here to where we can maybe simplify the process a little bit and not make it as maybe as complicated as what it's been in the past? Oh, there's so many ways that processes can be simplified and sped up too, right? Exactly more. Yeah. Yeah. I mean, I'm a big fan of sort of pre-screening without human intervention. There's a lot of sort of knockout factors that help at least prioritize candidates and you can involve the candidates in that without needing it to be a live human to human conversation.

41:56assessment tools are really important i think i'm a big fan of using assessments as long as they're done properly so lots of things that can be way more efficient and if your company is doing any kind of you know work workforce planning and forecasting you can also work ahead to actually build good pipelines and you know and kind of get people stacked up, you know, just slightly in advance of the needs. So everybody's got key talent roles, roles that you hire for pretty continuously, you can be ahead of the game on that, which also simplifies things tremendously. Is there a situation, guys, where the vibe coding is a sign of immaturity of tech stack, so to say?

42:51You know, back in the day, I'm just thinking thinking about like vibe coding like converting a google sheets into your ats i mean i always did that when i was starting from zero you wouldn't buy a tool everything be in a spreadsheet is pre-gdpr by the way um and then you'd you'd sort of um you know you'd you just do it until you you hit a certain number of people that system breaks down you have to buy a tool so is vibe coding similar in the sense it's just a better way of just diying the solution but you get to a point of size and scale like for instance where you guys are at morningstar mary k i think i don't know how many thousands of people you are but you know into the into the four figures five figures employees it becomes untenable or even impractical to start building stuff from scratch particularly non-expert people you've got off-the-shelf solution there um you can push some of the compliance risk to the vendor i guess even though you know you're looking at the eua it looks like it looks like the operator is also responsible but let's say they can share in the effort uh to stay in compliance um although then again i take that back because maybe the best way to stay compliant sort of down the pipe is to build it with yourself um because at least then you don't have the vendor making updates outside of your outside of your own compliance rules um because they're going to do that on their own understanding or interpretation of whatever jurisdiction they care about most.

44:19But anyway, that's a conversation for... That's a complicated question. Yeah. Although I would say the best way to stay compliant is to not let anybody else own your compliance. What does that mean, Mary Kay? Don't treat your vendors in a way that would imply that you need them to stay compliant. You should be managing those tools aggressively on your own. yeah i get what you mean you've got to police it yourself and police yourself and your vendors so yeah yeah paul a question for you um you've you've mentioned sort of the vision of multi-agent ecosystem that obviously makes sense at a certain level um however what about the prospect of interacting with other agents produced outside of that ecosystem is that part of the roadmap for for jovial are you thinking about that in any way or do you think actually no people have got to make a hard choice to just go one or the other no i don't think um i don't think necessarily right but i think what to kind of like this conversation like what we i think what's really important to have is like a really strong backbone of of agentic right because we've we've the main themes that i've heard from speaking to like this conversation so far today we're talking about compliance we're talking about governance you mentioned harm like gdpr etc then to everyone's point whether i think it's really important to have um a vendor that gives you that that really strong backbone and and skeleton right and especially with those handshakes between like you said mary k like PII data, I think that's then really important.

46:08But then I think offshoots of that, then I definitely don't think that is necessarily a bad thing and it then just becomes part of a wider ecosystem. So I would guess, I've kind of just sort of like a bit ambiguous in terms of my answer, but I definitely think that there is a value in both. But I think to get that strong making sure that um you can put your head on the pillow at night knowing that that it's everything everything's fine and you're not going to be you're not going to come with a with a with a huge fine i think there's that there's this huge huge value in that right and you'll be able to then kind of like get other people to do the legwork so you again what you can focus in on what your business is is all about and driving that forward right it's the that's why kind of like vendors exist because you can lean on them to then be able to focus in like sort of like pushing your your business forward so i definitely don't think it's i definitely don't think it's that with more agents but i think it is massively important to have a really strong core yeah okay go ahead okay i was just gonna say i've worked in in smaller organizations where you know morningstar has over 11 000 employees um i've worked in smaller organizations where i I wouldn't have had the luxury of the kind of technical folks in my team who are capable of building and scaling agents internally and, you know, would absolutely have needed to work with a venture to get to that.

47:49Yeah. Yeah, absolutely. Absolutely. What's your own thoughts on that question? Your own question back to you. Yeah. you know what my brain tends to be emptied for a reason paul um and um yeah the the less the less i the the less i i look internally as to what i think the better um because i need to observe what everyone else thinks the question is that you know answer is i don't know i mean i do think it's a matter of scale like i do i i notice a few people in the chat was saying look we're zero tech stack we're vibe coding everything i do think that probably is the way to go if you're a very small solopreneur small team um let's say you're an agency if you're setting up a recruitment agency day one probably you're going to vibe code the entire lot um and then you know you're going to be fairly confident that you're not going to be in breach of this out the other even though you are but you're probably not going to be a prominent sort of victim thereof.

48:50So, you know, this is not something I'm recommending, of course, if anyone's doing, but I can understand the practicalities of hacking a few things out and actually getting some good progress there. However, I think the journey that Morningstar have gone down is probably about right, depending a little bit on the infrastructure that you've been provided with. So for instance, a lot of companies haven't given every employee co-pilot licenses. They've been much more restrictive in different ways. So the more regulated you are, probably you're going to lean a little bit away from build your own, a little bit more to buy in and out.

49:28And that probably means more ecosystem type commitments. Although you can still imagine people buying multiple point solutions. I'm interested to know how point solutions might survive. There are some companies that I've got confidence in simply because they are addressing sort of a super niche environment that I think a generalist provider might struggle to adequately resolve in every case. So for instance, interview scheduling, you might imagine, okay, ATSs or any kind of core product will probably produce something as sort of an interview scheduler, which will suit lots of businesses out there.

50:11But does it suit, you know, a multi-location, a multi-language, a multi sort of panel stakeholder, a very significant recruitment process that might involve, you know, 50 ,000 interviews a month or something? Probably not. So at some point, a niche player point solution might still be operating, even though it becomes a much smaller market for them, but probably higher value customers that would pay for that service. Same, I think, with interview assessments, you know, you've got the screening software, maybe applicant screening becomes pretty much commoditized by core platform, I guess. um i mean i can i can see that but that's different from sourcing i think that's yeah you know when you're managing applicant flow coming in i think producing a screener for uh the cv or resume review is very similar to knockout questions as you mentioned mary k you could you could keep the assessment fairly light um so the filters porous uh but you're knocking out people that are irrelevant or asking them to establish whether they are relevant i think that's totally fine as well um and uh sort of component um but then you know you're hiring specifically for technical skills that you need to you need to assess or you're hiring for a psychology profile a psychological profile that you really want um probably point solution might come back into play there.

51:48So yeah, it's very interesting. George LaRocque's mentioned some... Go ahead, Depple. No, no, sorry, go ahead. I was going to mention George LaRocque here, because whenever George pipes up, it's always worth paying attention. George is actually coming into the show a couple of weeks down the pipe. We are going to be talking about this topic, VibeCode Solutions. Is it even like Bible? So we're all excited about it, but how Bible is it? And George has mentioned something here. He's talking about the opportunity i'll read i'll read out his comics i think it's worth reading for the opportunity isn't agents themselves it's the platforms emerging to govern the agents for instance who gets the agent access which frontier models get used how does it all plug into existing stack note that point how it plugs into existing stack or rather replace um and and the ones letting hr compose their own apps without vibe coding i guess that's an element of control like where are you allowing HR or TA who are non-experts to produce these agents?

52:49And can it legitimately be constrained? Because, you know, you don't want them to start stack ranking CVs, for instance, or you don't want them to, you know, get out of control. Like we've seen sort of in prominent, you know, as the Frontier Labs themselves have kind of rather strangely, unashamedly announced that they've created agents that have gone and hacked other companies and, you know, produce all this stuff. Look at us. We're so great. We found this hack. Yeah, it's like, yeah, maybe you shouldn't have done that. But feasibly, like you can feasibly imagine a local person in the TA team actually creating an agent that runs amok.

53:31And how do we even police that? Like, what is the policy there? So I wonder whether actually Paul and Mary Kay could ask that question, one from the practice. side I think it's very interesting Mary Kay your thoughts on how do we kind of make the constraint there how do we do it yeah well so within our environment we've put you know a number of different controls in place first of all you know you can build in in co-pilot for just your own kind of productivity and and that's fine but we do have an inventory of what you're building right you know once build an agent we we know what it is and and what you're doing so if we see something that we think has you know warning flags around it we can go off and investigate and talk to that that person and see what they're doing and why shut it down if we need to um and then that's actually how we discovered some of the things that people were doing that were really great as well.

54:34And, and when said, let, let us help you make that scalable because everybody should be doing this. And in order to build things that are larger and more powerful, people have to have access and it goes through a fairly rigorous approval process and they have to be able to tell us why they need access to build something that's more powerful or that has more connectors in it so um so just being able like it's not a it's not an open free-for-all on sort of builder grade access to these tools um it does have to go through a rigorous examination yeah i i completely i completely echo that like the internally with us we've got center of center of excellence in terms of our GPT accounts where everyone uses the same workspace, very much like you would do with Google or Microsoft.

55:34So you've got that level of control. And I think just from like I was chatting, Mary Kay, is just that, yeah, what guardrails, what framework do you have around that to make sure that stop and ask yourself, is this worthwhile even doing to begin with, but also that you're going to be compliant? Is it going to be safe? is it going to be useful so i think it's yeah it's a real shift and george you echoed it in the comments just that us having this conversation right is that there's a little bit of like doomsday that oh it's going to take all the take all the jobs having this conversation new frameworks and excellencies more people on compliance right it's it's gonna is this is gonna have to be yeah make sure that we put some put some guardrails to your point hung that doesn't run amok and then go hack into another system for sure.

56:30Yeah, it's really interesting. I wonder whether there may be a future. Again, I tend to go onto the side of dystopia simply because that's my, I find it entertaining. So, you know, I wonder whether people would end up creating like AI agents that actively seek out other AI agents and kind of arrest them and dust them down to see whether they're actually compliant. Do you know what I mean? Like one of the ways to manage it, it becomes a little bit like police state, but you can't allow like Wild West because everything will go absolutely completely wrong given how powerful AI agents might be. So you have like a Judge Dredd style sort of police state within your AI agent world and they just go stop people that are outside of the law and prevent sort of egregious transgressions, I guess.

57:23But then that leads on to, you know, do we want to create that vibe amongst these AI agents that, you know, might be emergent and sentience, you know? Do we want to teach them this type of behavior? I don't know. Yeah, interesting philosophy. But folks, we are coming to the end of this conversation and it's a fantastic chat to have. Loads of amazing comments as well, which is brilliant to see. I wish I had a bit more ability to bring people on screen randomly, but there was some tech issues we've had, some chopping in and out. And I think a lot of that might have been done to get basically the gallery in.

58:02But I think in future, I might try and do that. Like trying just randomly on spec, get people on. In which case, folks, make sure you have your makeup and your webcam on. You may well be called upon. You may well be called upon. But so many great comments. We want to make this dialogue. So make sure, for instance, that you're connected with everyone who's watched this show. That's something that we're always very keen for everyone to do. The reason why we do this is because Bring Food Live has to be a show that starts a conversation or ends it. So if you want to continue the chat as to whether we want to move to a disconnected AI future with best of breed point solutions, with great connectors, or do we shift to multi-agent ecosystem with a single sort of platform at core?

58:53That's a conversation you need to continue to have with your peers, as well as the vendors who are coming in. So way to do that, very straightforward, grab your LinkedIn URL, share it in the chat stream now and make sure you connect with everyone else who's done the same. If you're watching this on any of the LinkedIns, and I think at least a dozen people are still hanging on in my LinkedIn. So go and do the same. Just share your LinkedIn URL into the chat stream or the comment thread, should I say, and then just connect with everyone else who's done the same. And yeah, the rule of thumb there is just to make sure you connect with everyone and then make sure you've got a cohort of people that you can go and have this dialogue with as you see fit.

59:35Okay. I've got to let people go. Mary Kay Baldino, wonderful to see you. thank you so much for joining the show brilliant contribution as usual um paul bacon brilliant to have you on the show as well fantastic to have your input uh great slide deck as well and yeah really interesting to just have an exploration as to where um the world of recruitment tech is going we'll probably do more of these i think i'd like to get a bit more technical some of these maybe get a few product makers in as well um and just debate the issue you know where we where we go of this. Absolutely. Cool. Nice one. All right, Paul, listen, I'll let you go.

1:00:10Thanks. Get some rest, mate. Thanks for joining. Thanks, everyone. Bye now. Great stuff. Okay, folks, that's it. I hope you've enjoyed it. Thanks for bearing with some of the tech issues there. I'll take a look into why that was a bit tricky. But on the other side, thank you so much for watching. We're back next week. We have a triple header. Indeed, there's three brain food lives happening next week, I highly recommend you join them. They're kind of unusual times, Tuesday, Thursday, and Friday. On Tuesday, we're going to be having a one-to-one interview with a super exciting CEO and someone outside of domain who's created, I think, one of the most interesting products in the market to date.

1:00:52It is Matt Wilson of Jack and Jill. Jack and Jill, by the way, are a kind of a candidate gets an agent, employee gets an agent, the two agents talk to each other to find the right fit. Really interesting product. And it's going gangbusters here, sort of in the US and the UK. On Thursday, we're going to be talking about how to skillfully assess, how to assess for skillful use of AI. Everyone uses AI today. I use AI. Am I a skillful user? I can tell you right now I'm not. But how do you actually assess for that in an interview? I'm in a chat with David House, who's the senior engineering manager over at G2i.

1:01:32And we're going to be talking about how AI has changed the way they're finding and evaluate software engineers. Must attend for any tech recruiter out there. And then Friday, we're back with Brain Food Live. What are we talking about on Friday? Let me just give you a quick update because I'll tell you very shortly. It is another fantastic show, folks. We're going to be talking about, is Vibe Coding Tools the real reality? Is it viable? George Rock is on there. Again, We've got a bunch of technical people and recruitment people joining. Make sure you sign up for all of those individually. Follow the channel for more updates.

1:02:05I'm going to release October in due course. And then we're going to be all in there. Guy, you haven't missed the link for those sessions. What you can do, though, is follow my channel here on Crowdcast or follow me on LinkedIn. And all of those things will be there in any case. I'll share the link right there. And that's it. OK, I'll leave you to it, folks. Have a good weekend. I'll chat to you soon. Thanks for watching.