"Hi, this is Matt. Just before we start the show, I want to tell you about a free white paper that I've just published on AI and talent acquisition. We all know that AI is going to dramatically change recruiting. But what will that really look like? For example, imagine a future where AI can predict your company's future talent needs, build dynamic external and internal talent pools, craft, personalized candidate experiences and intelligently automate recruitment marketing. The new white paper, Ten Ways AI Will Transform Talent Acquisition, doesn't claim to have all the answers, but it does explore the most likely scenarios on how AI will impact recruiting. So, get a head start on planning and influencing the future of your talent acquisition strategy. You can download your copy of the white paper at mattalder.me/transform. There's been more of scientific discovery, more of technical advancement and material progress in your lifetime of advice and all the ages of history." Hi there! Welcome to episode 625 of Recreating Future with me, Matt Alder. The hype around the arrival of generative AI was off the scale crazy. Now things have calmed down a bit and the use of LLMs is starting to normalize, it can be easy to think that the hype was overblown. However, we're about to see some developments in the technology which will change our mindsets completely and start to bridge the gap between the initial hype and our current reality. My guest this week is Jack Horton, co-founder of Mindset AI. Mindset AI is using cutting edge AI technology in the HR and learning spaces. Jack talks us through the potential of autonomous AI agents to change the way we work forever and help some of the initial hype around AI become our living reality. Hi Jack and welcome to the podcast. Hello Matt, how are you doing? I'm very well thank you and it is an absolute pleasure to have you on the show. Please could you introduce yourself and tell us what you do? Pleasure to be here. My name is Jack Horton, I'm the Chief Product Officer and co-founder at Mindset. I guess a little bit about myself and Mindset. I've been in people learning HR and early curious recruitment for many different years in different businesses and a little bit about Mindset is kind of the culmination of a lot of my experience personally and the experience of the rest of the team. It's kind of a perfect storm and what we do is we essentially are a, we help companies launch their own AI co-pilot for typically their knowledge and learning and content libraries and internal knowledge as well. So enabling employees or users to suddenly ask an AI agent or a co-pilot or a chat bot, whatever you want to call it, different important questions to help them find exactly what they're looking for inside of massive content libraries, learning libraries, knowledge bases inside JXG Drive or SharePoint and for the AI to be able to ask questions back, take them on a guided search experience and help serve up those easy to digest slices of information inside their knowledge base for them to understand and learn about. Obviously over the last 18 months, all anyone wants to talk about is generally AI and some of the things that it's making possible but also thoughts about where it might go in the future. Based on the experience of what you're building and the knowledge that you have and the team that you've got, I better time stamp this. We're talking mid-May 2024. Where are we with AI right now and what's likely to happen in terms of developments that we should look out for over the next 12 months? There's a different, a couple of different ways you could answer that. I mean, for those on the inside, I always say right now, it feels very much like we're in this kind of slope of enlightenment, plateau of productivity, area of GNAI. So if you've ever kept track of the Gartner hype cycle, which is always a really interesting process to kind of watch and fold, when the likes of Chuck EBT came along, I guess everyone assumed it could do everything and solve all problems. Everyone's going to lose their jobs or, you know, those type of, that type of dialogue suddenly emerged really quickly. And really, really soon after that, you realize the kind of the limitations of everything. Now, we've been building what, not AI, machine learning for a number of years. And I guess what's been really interesting to really answer your question is that there's a multiple combinations of new forms of technology that's not just GNAI related but makes GNAI really, really useful. And I guess that's where we're coming to at the moment from, I guess, the inside camp. People are building in it every single day, all day, every week. In that it's now becoming really clear what sets of technologies need to be brought together to make this incredibly valuable for people and not just like a chatbot. And I guess we're really now at the, I guess, the edge of that really. And I think that's where people are going to see major differences over the next 12 months to where we've been in the previous 12 months. And I guess it's really the best analogy is saying it feels less like a chatbot as an experience and more like a human that's actually helpful, understands me and can really deal with my problems. Yeah, I think that makes a lot of sense because we did see an extraordinary amount of hype when this sort of technology became generally available. People were predicting that everyone would lose their job in a matter of weeks and all these things would happen. And I think that the danger sometimes with these technologies is they do go through that process where there's hype and then that doesn't, it doesn't live up to that. And gradually they come back. And I think what's interesting with JNAI in particular is it seems to be going through that very, very quickly. So in some ways, people might feel that things may have been a bit quiet for the last few weeks, but yeah, it does feel that there are some really interesting things about to happen. Yeah, I guess, and I think, I think I always probably miss underestimate where the rest of the world is up. You know, obviously, again, it's really tough sometimes being on the inside of the camp. And naturally, all of our customers are therefore the, I guess, you put them in the early adopters, technically, you know, there's still although adoption spiking, I mean, I saw a crazy start, it was about 92% of Fortune 500 companies are using some former JNAI. But we're still in that early adoption period, really. And there's many people in the world that just haven't even barely used anything well, and you're not even just Chuck GBT. And usually the good litmus test is saying, you know, have you heard of Claude? Have you, you know, whatever you just just naming for other large language model out there? Have you heard of them? Have you engaged with them? And I think that's why we're still quite early on in the entire journey. But I think it's a particularly exciting period because really the combinations of technologies that are now emerging will make so freely possible. I'm being really vague there, but I guess that will probably come out through the rest of the discussion. I'm going to pick on that up on that straight away, actually, because I know that one of the things that is a key part of this is this whole concept of AI agents. Tell us about them. What are they? How do they work? Are they here right now? Is this something for the future? Give us the kind of low down on it. Yeah, it's a really interesting one. I guess the reason it's really interesting is because all generative AI endeavors will mostly be agent-based. It's an agent framework, they follow. And agents are, they have a set of capabilities that move beyond just chatbot, because ultimately the objective we all want to achieve is not have AI answer questions on chat GPT. I'm going to use the most common denominator that most people would have known. That's not what we all want to be. We actually want to have something that understands us, does more than just answer a question that's often misaligned with our intention. We want just to, this is my problem, go help me solve it. I don't even want to have the cognitive burden necessarily of understanding how to really always solve that problem. I want to be helped in understanding how I could solve that problem, rather than me do most of the cognitive work and then have very low level, simple tasks executed, write me this, summarize that, which are all incredibly powerful. And there's an immense amount of value that's yet to be captured there. But actually that's why AI agents are going to be really important, because actually they're the gap between the bridge between this kind of the world of a possibility of what people like to talk about and where we're at today. So if I break down some of what they are is essentially an AI agent has a few of unique capabilities. So an AI agent is able to, and this, by the way, still shocks me, it's amazing to see it, is able to apply logical reasoning and create planning and processes and strategies. So one task or a question from a user, if I ask you something like, let's go really simple in a HR context, I guess, what's the dress code policy? Ideally, it's quite a simple question, but actually it's got a lot of nuance there, where you're based, what gender, what office, what company, you know, there's a lot of information there. And an AI agent is able to apply a logical reasoning process, much like a human, to go, I need to ask or find the following sets of information, and it's able to create a process strategy. So that's a really, really powerful capability. And watching it do that is really, really incredible. And it uses the large language model itself to ask itself back and forth questions to understand if those steps are the right steps to take. So that's a really, really powerful power. And that's just one component. I guess the other, there's multiple of these, but the other component is they have what's called tools. So tools are a really important word now. A tool is an API, it's an ability. A tool could be a, let's say, it could be a, something as simple as a summarize something, or extract data, or extract tool, could also be a search the web tool. But essentially, you're able to provide its specific tools. So when it actually has that process, it's able to use different tools along that process to execute a task that exists not just inside the chat bot context, but might exist on the internet, or in workday, or in another system, i.e. information that needs to retrieve, or information needs to push into those systems. And underpinning that as well is the ability to have workflows. So call it process flows, workflows, much like a Zapier, much like those like traditional robotic process automation, but in a way that's much more smart, because you don't have to dictate every minutiae of step, it's able to infer the steps it should take. So it's a very quick summary of what an agent is, and the most powerful, the most important thing here is the combination of both memory and engaging with other agents. So that's the kind of final two parts. Just to clarify in the first instance, so is this technology that's available now? Because I've seen a lot written about agents, and very often people are talking about this is what's possible in a year's time, this is what's going to be in chat, GPT-5. So where are we with this technology right now? It's here. I mean, that's what we are. We're an agent platform. And we use the word co-pilot, because that's the word people know. Most of our customers are kind of, let's say, one of our customers just launched, which is like the biggest library of HR leader content in the world, that just launched a co-pilot for that product. So people just understand the word co-pilot from a marketing point of view. But we're agent-based today, so the technology is here now. And that's just fascinating stuff, because I think I'm really reflecting on what you said there about when you interface with a large language model, chat GPT, whatever it is, is this kind of process of steps that you're sort of making a path for it to give you information. And I think that one of the misunderstandings or misconceptions that's around in the industry at the moment is that's it when it comes to AI. So I think people look at that sometimes and think there's no way that this is ever going to replace my job. It might automate things that I do, but it can't really think in the same way that I can. But actually, in a few months time, that's going to be very different, isn't it? Yes or no? I mean, it's a really interesting argument. So I mean, I'll give you a very tangible example of where the limitations are and put in a HR context. So we've got a partner in HR. So one of the products we have is essentially AI, let's call it HR automation and support. So a co-pilot for every employee, basically within a HR context, we've got a partner called a quick AI that's incredible, that runs all of that. So they're doing amazing work. So the limitations and positives and exciting part, I guess, is dictating as you describe a workflow of, let's say book me a holiday or fetch this information from SharePoint, be able to build that workflow for an employee. So they don't need to think about anything or what system they're using. Just one question, book me a holiday, push the information, books it for them, done. So that's really exactly what you write, but how does that replace people? And I guess the fear is that that replaces people at scale. So I mean, when you've got thousands of those workflows, and then the second part of the agent framework is when agents can talk to other agents. So this is an innovation that nobody's really fully cracked yet when we're on that journey. Agents talk to other agents. So imagine once we build up many, many different workflows for completing many important tasks and workflows that are a bit more complex than, you know, write me this quick article. An agent can pull in other agents to execute a process. So you might have an agent for HR, for employee wellbeing, you might have an agent for marketing and content writing. At scale, when this gets to that point of scale is when you have agents talking to other agents that can complete these. And I guess that's probably where people fear that it's called a swarm of agents. That's the technical term. Wow. Yeah, it's nice. It's quite a horrible name. It doesn't inspire excitement, but I guess people must fear that suddenly that can replace people. But I find it a non-com, a not a very compelling argument overall, mainly because people say, well, companies will start firing lots of people and that will happen. You know, that will happen to many jobs. But if you say to companies typically, okay, with 30, let's say 30 developers, I know there's a big, if you've heard of Devon, the AI agent for programming came out recently. There was a massive splash. Let's say this agent for programming. If you said to a company, combining Devon with your programming team could 10x your output. What do you want to do? Do you want to go with the developer team and just have Devon or reduce it and just have Devon? And you'll have four extra output or 10 extra output. I think companies typically want more output. They just want to press that button more and more and more and more. So I think really, that's where I find that more compelling is how do we get people to do 10 times more? You mentioned, as we sort of gone through the conversation, you mentioned a few use cases that we're seeing in HR and recruiting and things like that. It tells a little bit more about what some other use cases might be, what might be possible in the near future. It's actually in many ways a product management question of use cases. It's really interesting that we're all going through this main question of people get very excited, which is, oh my god, we could do so much and then go actually, what is the specific thing it can do? So it's always a really interesting question. I guess really it's a job to be done thing. So interesting things that it can do, let's say, take us end to end, not just to write an article, but do the research on the web, identify every single websites, high performing keywords, second, identify all the trending topics on different platforms aggregated from an important website, pull into those, this is in one workflow, pull into those and identify that as a theme and a blog or an article or a paper brief, then turn that into an article, but use and reference this template and then publish that to Slack every single morning for my team. Good example, multi-step workflow that essentially traverses your internal knowledge base, external knowledge sources and pushes it to the place that your team need it to then adapt and publish and go live. So there are some of the really powerful ones. Yeah, I think that's where I guess that's, most of the time really, if I break it into two use cases, ask and do. Ask is searching for information inside of my knowledge base and also external, but inside of my knowledge base. So an example, integrate, so mindset, integrated into SharePoint, into G Drive, into these different sources and ask a single question and get and pull the answer, but also the source document and that segment of the thing that I was looking for. So that's ask. And you think about the amount of questions you ask to every single department in your company. So it could be employee and HR related questions and recruitment related questions, product related questions, commercial related questions, everything. And then you got do, which is that workflow I kind of described. So you break down all those do tasks, suddenly. And I think they're the two primary big areas of work. And I think people get very excited by the do, including myself, I'll admit, but actually the most powerful use case in the world right now, in my opinion, is still asked by a mile. I think that's where the most value can be gained quickest because you very quickly get into the question of use cases of a lot of work has to get put into that. But the world of ask is really powerful. That's a really interesting point. And obviously, there's so much going on, as you say, you're kind of inside of all of this and trying to sort of make sense of how to use it moving forward. For people who are listening, who really will appreciate that actually this is going to be such a key part of their job and their career moving forward. How should they be thinking about it? And how could they stay up to date with the technology and kind of really plan for the future? It's a really interesting question. I mean, so most importantly, it's get really close to it now is quite important. And a good example. So we were talking to an ATS provider. They've not yet done much, but they're really interested in their product team. We integrate our agents into technology platforms and people learning in HR a lot of the time for helping users get information. I guess the most important thing is to stop moving from that, like talking about internally to actually just trying to bring experts in, whether it's a vendor, whether it's talent and people, but get into it really, really early on, that as soon as possible. So that would be an organizational perspective, I guess. From an individual perspective, to be honest, a lot of the time, here's my take is AI is just becoming SaaS. You can be as close as you want to it really, but it's just understanding how to use the tools that are available a lot of the time. I think when it first came out, everybody was kind of, it was magic, but now it's just getting put into every SaaS product everywhere. So it's just really understanding very quickly how you can actually think from an individual level, like, what's my workflow and what tools will I use in this new AI world to be productive, to be successful, to be driven by my career. And that's honestly fundamental, especially from a recruiting HR perspective, I think. Yeah, I think that's interesting because you can see already that it's on its way to becoming invisible, just in terms of how it's kind of being adopted by the tools that we use and, you know, is driving a lot of things that happen every day. And I suppose the question for people is, as that happens, what are the implications for efficiency and the way that we work? So it's much more of a strategic view on, you know, how to position functions than just understanding individual bits of the technology, isn't it? 100%. I mean talent acquisition in particular is a bit of a tough one because it's a quite highly regulated space, actually, from a Jedi perspective. It's where the most regulations are laws have kind of been put in place almost. But I mean, if I was to give advice, I mean, really, really quickly, I would say most importantly is there's usually in each team within your recruitment function, paycheck, whatever it might be, there's usually one or two people that are kind of real, let's call them the early adopter innovators that get incredibly excited by this type of technology. What you want to do is essentially turn them into a champion, create a communication space, to start posting tools, things, ideas into there, to really start to engage the team, because that's how you get the team involved involved and bought in really quickly. And from there, what you can start to do is essentially go, okay, what are the things that are a problem in our team? How do we drive that in terms of efficiencies? And when people start adopting these type of thinking and they get excited, they have champions that are backed by the leaders who are championing this. Certainly, you can now start to get a team that's quite educated. So that question of use case, people already have 10 ideas already. And then I always recommend people to say, why don't you host a let's call it a hackathon as a team? And you go, what are our jobs to be done? What do we hate doing? What do we love doing? What are we really good at? And how do we build a kind of, let's say, AI framework for our team around that? And that's where you can start to be strategic, because you can obviously go, okay, these ideas brilliant, we can't do them for six moments, but we'll drive huge efficiencies. And certainly what you've done is bought a team in, you've got them involved, you've got them excited, you've got them educated, and they understand the implications. And once you get that team level involvement and education, you can then start to think as an organization, how do I move the needle with AI as well, rather than jumping in at the deep end, I guess. Final question for you, and probably the question I've been dying to ask you all the way through this conversation, what is the future hold? Where is this going? Where might we be in sort of three, four, five years time? There's certain things you can almost guarantee. Mars law will continue. So everything will become incredibly much more efficient at scale, and it will become more powerful and intelligent, essentially, it will just keep getting better and better for some time anyway. I mean, that law is yet to be proven wrong. So when we think about large language models, we can assume that most things that people really want it to be able to do, will be able to be done in within a year's time, basically, and all the things, whether it's avatars or voice or this or that, most of it will get incredibly good really quickly. And I think in 12 months time, I think what will probably be is that AI, probably 24 months, AI will be up to the everywhere. I mean, again, we from ATSs to HR providers to LMSs, we've already embedded in. So that'll be happening all over the world. And it will be in every single person's workflow. I think really all that will happen is now all the combinations of technologies that have been brought together by ourselves and other vendors in the world will start to really become valuable for people at scale. And I think that's for me, the missing gap right now is there's been some value attained from AI. It's now capturing the full value. And I think that's the next 24 months, to be honest, is actual widespread adoption everywhere. And that's the biggest impact. And I think 100% that comes to AI agents. I'm pretty down certain whereby AI feels less like a chatbot and more like a actual helpful assistant that's always there with you. Because if you look at wearables, where we're going with them, there's lots of things that are trying to limit the friction between me and talking to a computer. So anything that can limit that friction, basically, is good. And I think every organization, every team will just have an agent interface into them, basically, for every person to engage with. Jack, thank you very much for talking to me. It was an absolute pleasure. My thanks to Jack. You can follow this podcast on Apple podcasts on Spotify or via your podcasting app of choice. Please also subscribe to our YouTube channel by going to matalder.tv. You can search all the past episodes at recruitingfuture.com. On that site, you can also subscribe to our newsletter, Recruiting Future Feast, and get the inside track about everything that's coming up on the show. Thanks very much for listening. I'll be back next time, and I hope you'll join me. ♪♪♪ This is my show. ♪♪♪ [ Silence ]
Get my free whitepaper "10 Ways AI Will Transform Talent Acquisition" - Download Now
Trend Spotting - Find out how my digital course will help you shape the future of talent acquisition in your organization - Click Here
The hype around the arrival of Gen AI was off-the-scale crazy. Now that things have calmed down and the use of LLMs is starting to normalize, it can be easy to think that the hype was overblown. However, we are about to see some developments in technology that will change our mindsets entirely and start to bridge the gap between the initial hype and our current reality.
My guest this week is Jack Houghton, Co-Founder of Mindset AI. Mindset AI uses cutting-edge AI technology in the HR and Learning space. Jack talks us through the potential of Autonomous AI Agents to change how we work forever and help some of the initial hype around AI become reality.
In the interview, we discuss:
The huge developments coming in the next 12 months
How will removing the current friction in accessing Gen AI accelerate adoption?
How Autonomous AI Agents will bridge where we are and where we are going.
Tools, APIs, and logical reasoning
How do AI Agents communicate in Swarms to build workflows and intelligently manage tasks?
Ask use cases and Do use cases.
What are the limitations and the implications for jobs?
The importance of taking a strategic approach
Where might we be in 3-5 years?
Follow this podcast in Apple Podcasts.
