How to Increase AI Adoption in Your Workplace
Jack Houghton, Chief Product Officer and Co-founder at Mindset, and Matt Burns, Co-founder of atlas copilot, explore the importance of bottom-up AI adoption, the role of AI agents in enhancing workplace efficiency, and the impact of AI on knowledge management. They provide strategies for solving common content management issues and discuss how AI can create personalized learning experiences.
In today's episode of the HR Leaders Podcast, we welcome two exceptional guests: Jack Houghton, Chief Product Officer and co-founder at MindsetAI, and Matt Burns, Co-founder of atlas copilot.
Jack and Matt delve into the importance of bottom-up AI adoption, the role of AI agents in improving workplace efficiency, and the impact of AI on knowledge management.
They provide insights into how organizations can harness AI to solve common content management problems and enhance dynamic learning pathways.
🎓 In this episode, Jack and Matt discuss:
The journey of creating adaptive technology for users
The role of AI agents in enhancing workplace efficiency
Challenges and strategies in enterprise-level AI adoption
How AI can create personalized learning experiences at scale
The impact of AI on knowledge management and contextual understanding
DISCOVER WHAT EMOTIONAL SALARY MEANS – AND HOW YOU CAN MOTIVATE EMPLOYEES BEYOND PAY.
Great recognition is more than just a thank you program. By leveraging frequent and meaningful recognition, Achievers drives business results that matter to organizations like retention, productivity, and engagement. Our platform makes it easy for employees to recognize each other anywhere, whether in-office, remote, or on-the-go.
The Achievers Workforce Institute reveals that two-thirds of employees have one foot out the door in 2024. The top reason for job hunting? Better compensation. But money isn’t the whole story. Employees are seeking not only monetary salary, but emotional salary too.
Jack Houghton 0:00
Everybody looks at adoption for a very top down view, and have a very impatient view a lot of the time. So I think there's a lot of people who are the innovative people that love new shiny product, because it's a b2b technology with b2c dynamics. People like to adopt it and use it themselves. But actually, if you move away from that enterprise level adoption, you actually see people using the technology every single day, but not telling people about it. It's fascinating. You probably saw it, the LinkedIn Microsoft report, 60 odd percent of managers are still trying to figure out how to measure ROI and agenne AI product. And it's like 78% of knowledge workers now using one every day and not telling people I guess there's a lack of vision or bravery a lot of the time because companies are trying to justify an ROI. And as a result, people just don't wait for them. I think the report branded it, bring your own AI obviously, bring your own BYOB, bring your own AI to work.
Chris Rainey 0:54
Jack, welcome to the show. How are you?
Jack Houghton 0:55
I'm good. Thank you very much. It's great to be here.
Chris Rainey 0:57
Matt, how are you my friend? You know, I'm doing really you do anywhere from Canada,
Matt Burns 1:02
we we launched a company. It's morphing. And I hadn't been to Ben before. So I wanted to make sure I saw that too. Yeah,
Chris Rainey 1:11
amazing. Before we jump in, everyone on the show knows Matt at this point, but we'll also tell them a little bit more about yourself, and your background and journey to where we are now.
Jack Houghton 1:21
So yeah, my name is Jack Horton, Chief Product Officer and co founder at mindset. I guess a little bit about the journey. Someone's really started with one fundamental first principle question, which is, how do you build technology which adapts to people and people not adapt to technology?
Chris Rainey 1:39
And that one for a while? Oh, God,
Jack Houghton 1:41
you get asked these questions. I always get asked these questions of who are you? What do you do? And you have to regurgitate them all the time?
Chris Rainey 1:48
What is the problem that mindset solving? No, it's not?
Jack Houghton 1:51
It's a good question. Yeah, I guess there's there's probably three big problems we see. Number one, most people have to rely on keyword searches. It's infuriating. You know, it used to be really good in the 90s. But these days, it's incredibly frustrating. So you rarely ever find what you need. So are these big companies with lots of content learning libraries, knowledge bases, that people are users can find what they need. Number two, is typically when you find something if you're lucky enough to find something, sifting through all that content is infuriating. You know, it's halfway through that video, if I paid a pound for every time someone said, that's a really good idea, it's halfway through that video, probably be able to buy an expensive lunch. But it is a real problem for companies with with a heck of a lot of users. And number three is typically people don't know where to search. Is it my LMS? Is it my G Drive? Is it God forbid SharePoint? Where do we go? Where is it? Where is it? So there are three big problems that we see in the world that we we get to work on fixing. And we get to work with some of the largest learning providers and software companies, large content and media libraries, media companies, and increasingly, businesses. So
Chris Rainey 3:04
amazing. I think we're just as excited to have you on the show. And also have Matt here is, as HR leaders, they're just inundated with new technology, new shiny objects every day. AI is the new buzzword. There's always new terminologies every day, there's a new one, right? It's a bit of a minefield right now for HR executives. So I wanted to use this episode as a good opportunity to set really just help them on the journey. And we break down some of these complex things that we hear. Yeah. And jump into it. Let's just kind of start off with talking about the current landscape. Now what is the current state of Gennai? Deployment at work? And what you're seeing there, start there? And then we can kind of build upon that? It's a
Jack Houghton 3:50
really good question. I guess it's it's as a technology is quite unique, because most of the time new technologies, small little startups, a camera focuses on this niche product that nobody cares about, or big companies don't care about on a market that also nobody cares about. And then over time, it grows bigger and bigger and bigger, and takes over the world. Whereas with Gen AI, the only people that could afford to build large language models were the Microsoft's the AWS is the massive companies. So the dynamic we're in is these huge expectations from a public market perspective of where's the revenue coming from? You spent $50 billion here, where's the money? And therefore, a heck of a lot of expectation for quick returns, and a huge amount of attention. Whereas I think, actually, if you look at where we're at today is everybody looks at adoption from a very top down view, and have a very impatient view a lot of the time. So I think there's a lot of people who are the innovative people that love new shiny products, because it's a b2b technology with b2c dynamics. So people like to adopt it and use it themselves. But actually, if you move away from that enterprise level adoption, you actually see people using the technology every single day but not telling people about it. It's fascinating that you probably saw it the LinkedIn Microsoft report 60 odd percent of managers are still trying to figure out how to measure ROI and our Gen AI product. Whereas it's like they know they're using every day. And it's like 78% of knowledge workers, and are using one every day and not telling people, how do you measure it? Exactly. And I guess there's a lack of vision or bravery a lot of the time because companies are trying to justify an ROI. And as a result, people just don't wait for them. I think the report branded it, bring your own AI, obviously, bring your own BYOB, bring your own AI to work.
Chris Rainey 5:36
Oh, my God, Matt. Well, you've seen it's
Matt Burns 5:37
similar phenomenon we saw in the early days of the pandemic, where we talked to people about hybrid work, when you ask most employees, are you more productive at home or in the office, they said, I'm more productive at home. And when you ask their managers, are they more productive at home or in the office, they said democracy in the office. So there's been a disconnect systemically between managers and employees for quite some time. And I think that's another example of where you see that right now, employees find value in using AI in the flow of work. And managers are concerned about how they're going to use AI in the flow of work. They're concerned about things like privacy and data security. And these are very, very valid concerns. And as they explore and navigate this world of artificial intelligence, it's just simply not acceptable as a manager or a leader in an organization to say, we don't do AI, you need to create safeguards and need to create processes and need to create checks and balances so that it can be deployed, and with integrity and with ethics and appropriately, but it needs to be deployed, because organizations that don't deploy AI will have a significant disadvantage to even startups that do have AI and Chris, when when we launched Atlas co pilot, how often do we use that solution in our own company, every day, every day, all day, because like Jack, we're also trying to solve three problems, we're trying to solve the problem around getting the information to the people who can use it most effectively, which in our case, are people leaders, they're working in organizations, they're supporting employees, they're supporting one another at a very difficult time to be a leader. And they're up to here with information and tasks and responsibilities. And it's not realistic to expect them to recall every possible lesson and knowledge and detail in their organization. So the ability for artificial intelligence to overlay on company policies and procedures and communications, to overlay best practices from leading thought leaders and people who are in the role and other organizations is a significant advantage when it surfaced in the flow of work. And if you can trust the sources, and you can trust the safety of the sources, it's going to give leaders an incredible advantage. So the ability to synthesize all that information into a digestible piece of actionable data to curate the right kind of content. And I think, Chris, what you and Shane have done really well for almost 10 years is tell really compelling stories. How do we surface really good information, then not just valuable, but as also engaging? Because if it's a chore to go through asynchronous elearning courses, no one wants to do them. They don't want to do it 12 questions, survey on health and safety, and then click the box and move on. They want to be engaged, we actually want them to retain the knowledge. And the way you do that is by surfacing knowledge and information in cool formats, videos, through interviews, in summits, I think we've done a really nice job pairing those three things together. And but the proof will be in the pudding. Do we have people that ultimately take that information and then deploy it? And those of you in new and unique ways? Yes,
Chris Rainey 8:21
the thing, right? It's weird, because we've gone through this evolution of when I first started in my career, it was like the access to knowledge was our challenge. Right? And then only a certain layer of leadership had privy access to X information, right? Then YouTube came, then Google came, and we all have access to unlimited knowledge. It's all out there. Right. So now our problem becomes knowledge overload. So how do I find the answer to my question? And that whether that's a piece of content, whether it's a research paper, whatever it may be, basically, but the missing point that a lot of companies, I think, talking about ROI is and how does that lead to an action that has impact on the organization? And we're already seeing I hate to mention Apple, Apple's keynote yesterday, Apple intelligence about their entire focus was talking about here's how Apple intelligence helps create an action because they've integrated that into all of the apps, so they can all talk to each other BI directionally and take an action. And I think when we get to talk about ROI, I think that's not a conversation that many of the HR leaders I speak to don't get that bit. elaborate a bit on that.
Jack Houghton 9:35
Yeah, I mean, it's a really interesting one, I guess it it feeds into I don't want to go into too much for now. But it comes into a lot of, I guess what Apple have done is AI agents, which take actions across systems but I guess from a return on investment thing, I think people are still behind the times I there's a CFO joke, which is if every app saved us this much time and money we'd be we'd be literally constantly profit making it over single department, because that's that's been the way most people sell. And I guess the really sad thing is, is that people aren't spending enough time digging into, I guess the actions as a result of a Atlas tool or whatever we're doing. Because most people have to hide from it. Because if they tell their it boss or their team that they're using this data AI tool, actually, what happens is that people say you're not allowed to use that. So actually, if you think about use cases, and obviously product management is our job is to find the right use case. Basically, it solves a problem. Everyone's going what are the right AI use cases? And actually, most of the time your team have already invented about 500. And if you measure the ROI from the actions of that 500 different use cases, because you know, that trusts it. Exactly, exactly. So there's there's a lot to dig into in the question of actions. But on Apple intelligence, especially, I think people again, because it's a consumer technology will force every single person who HR leader or manager to go okay, now I know I get it, you know, I don't have to wait for my team manager to deploy this. It's on my phone.
Chris Rainey 11:01
That's going to be crazy, right? I think this breakdown, and let's jump in as well. What our AI agents, what our agents, we keep hearing this terminology, constantly. And a lot of these I speak to you, especially in HR spaces, like what agents, bidirectional agents, when you just break that down.
Jack Houghton 11:18
So there's two parts, I'd say to that answer. There's what are they? And then how do they actually work? And I think I won't go into both of them straightaway. But I think both are really important to get your head around it, because they sound like a James Bond thing. Yeah, AI and Agent shouldn't be put together really, because it's not something that people want to adopt. They don't trust it. But essentially, most people today use a non agentic workflow. So agentic is what people often say. So if you hear that word, you know, agent. And it basically means I prompt an LLM, ask it a question. It doesn't involve me. And it comes back. Imagine if you ask an AI to write you an essay. So the AI just literally writing new one without taking a breath, any thought or even asking questions about what the essay should really involve. And I guess the difference with an agent is the ability for it to mimic humans. So plan, do something iterate, test it, improve it, complete it. And that, fundamentally is what an agent is. And the way it all works and pulls together we can do dig into more detail is a series of prompts. So it's like, imagine this, like an ability to prompt in many, many different different ways, without a human ever realizing to achieve that objective of mimicking humans. Yeah.
Chris Rainey 12:30
What's the main difference between an agent and assistant? So a lot of companies already have assistants that are built into their, into their companies already. What are the main difference was put in
Matt Burns 12:42
practical terms. So an agent in the context of Atlas, is you ask a question, you get a question back. Now, why is that important? Because we can't expect people to be prompt engineers. They're not math because like you and I, Chris, that write 400 word prompts, and put it into our platform. They ask questions like, build me a talent strategy for my engineering department. Or I need to think about how I build a conversation script for a tough performance chat with one of my colleagues, they often don't introduce the context necessary to be able to get the best possible answer. So with agentic capabilities, it comes back to you doesn't answer the question, but actually comes back with a question and says, Hey, in the context of talent, tell me a bit more about your industry, your geography, some of the challenges that you're facing around talent. And by illuminating more context, it ultimately delivers a better result. But the important thing around that is it does so in a chat interface. Because we can't expect people to upload essays around context, or spend hours behind a computer, looking for speed, quick answer, but by having a conversation, it feels light, it feels easy. And then ultimately, it delivers them that better results. So the agentic capability in that context is complex in his description. But it's even more transformative in terms of response. Because you get an answer that's more context rich, that's more in the flow of the conversation. And then the user is more happy with the response.
Jack Houghton 14:03
I think you the opponent have a better response is the key one, I mean, the there's a study produced by Andrew Ng, if you ever search for him, incredible guy. With GPT 3.5. On a single coding task, it was under 50% accurate per task. With GPT. Four, it was just over half an exact 60% with an agentic workflow on GPT 3.5. It was nearly 100% accurate. So I mean, we because we've got a project will be deployed to millions of people with a learning content library worldwide. What do they care about? accuracy, consistency, reliability? We noticed very early on that that was going to be the one of the things the benchmark measures when everyone comes down a little bit about this is really cool. And we get into how does it perform to your appointment. That's why agents are so important. And there's about four or five different components that make up an agent of which the questions back is really The important one, but that's where the value is gonna get delivered. And that's where we're gonna see AI really deliver the value people hope, and are potentially scared off, to be honest.
Chris Rainey 15:09
Can you write down those three things? Yeah, so
Jack Houghton 15:14
exactly there was planning is one of them. So actually teaching AI to think slow, slow down, plan your process of steps of how you're going to execute that. So if someone says, What's the dress code policy, this place it can go and ask itself, actually, I need to find out a set of information to answer that correctly. So that's planning to ask it to create that process, indirectly linked with planning is the ability for it to reflect. So read prompting and asking how does that process step look like? How's the output? Would you improve that? Is that right or wrong? And it's incredible how much increase in performance you get, just by those two things. So you've got short or long term memory.
Chris Rainey 15:50
I was about to say, when I was when I was about to ask you about that. Where's the memory come into that sort of long
Jack Houghton 15:54
term memory? Exactly. So the ability for to store data as variables. So every conversation is remembered, but long term information on you as a person? Yeah, so integrate into these different platforms. You also have what's called tools. We'll all see tools very soon, we've got big, big leaps here. But a tool could be a calculator tool, a web search tool, it could be a holiday booking tool to integrate into different platforms. So it knows to execute a tool. So use a tool like a human, I'm gonna use this tool to complete this job at the right time. And this is all programmed into you, as a human as user don't need to think about. Yeah. And I'm gonna get
Chris Rainey 16:31
back to our point earlier about it, then being able to create ads to actually execute on tasks, exactly as at the moment, and a lot of that isn't. I mean, that's the sort of Next Wave asked him to do. That we're seeing right? Yeah,
Jack Houghton 16:42
that's, that's where that's where the big leap. I mean, we're, we're certainly we're be there almost immediately. It's the big, big, big push for us at the moment. I mean, I always, I always want to slow people down. Because most of our life as humans, especially knowledge work, is asking other people for information, searching for it, analyzing it reflecting and doing something with it. And the other half of our work is doing a task. So we've not even captured the value from the search world yet. But we're already obviously driving ahead because we're humans are cool humans. Yeah. Never, never bought.
Chris Rainey 17:13
Yeah. What offers
Matt Burns 17:14
the opportunity for us to get the benefits of specialization and generalization at the same time. So one of the limitations right now of using open AI, and again, I feel like I'm being cruel to open AI. They're absolutely transformative technology, pushing the path. But in the in the context of asking questions for people, leaders, or in specific industries, open AI is being trained on a broad knowledge set. So the ask a question, you get a general response, which is amazing. But if you want to ask a question about a specific context, you have to introduce that context, or through agents, we can introduce the contents through agents, syllabus in terms of an HR leadership team, we can create agents for total rewards, talent acquisition, people analytics, and we can ask those individual agents questions about their specialization. If you are a people analytics agent, please help me develop a questionnaire on how to develop a strategy for analytics in my organization, and here are some of the variables. And then that agent could confer with all the other agents in the same way in HR, lt would meet weekly to discuss organizational issues, they in real time can confer. But that people analytics agent kinda gets the 51% vote because it's the expert, and can come back with all that context, but also with specialization. And then the end to Jack's point, gets you incrementally closer to a better answer. So for the user, for the for the HR practitioner, for the people manager, they're getting an answer that's just that closer to perfection, which is ultimately the what we're trying to get to as as a performance and a platform is to give them those context rich answers in the flow of work, so that they don't have to go back and go back and forth five 610 times that they can just get the answer and take an action. Yeah,
Jack Houghton 18:46
you're on a really interesting point is that multi, it's got a multi agent system, they are terrible at naming things in the
Chris Rainey 18:51
direction or do they call it? What's the actual terminology? So let me call it by direct.
Jack Houghton 18:56
Robert bidirectional is good. It feeds information and can do things both ways. But um, so it's called a multi agent system. And as I said, agents terrible name but is they call it a swarm
Chris Rainey 19:09
is saying, hey, selling swarms. Make everybody distressed you it sounds like the siege in the matrix when a swarm robots come together today, there's not caught up. But that's
Jack Houghton 19:23
where the that's where I've I've looked at where agents are going is some multi agent systems exist? Now we've we've got it. But it's only in small doses. So it's about increasing some performance, but it's hard. So you basically got this manager agent that tells other agents which agents are needed as part of a process. And that's where the next big innovation it's I guess, 5050 in small doses works really well. We're not all cracked it. I think that's going to be a big thing for the end of 2024. But it's, as you said, it's where you have 15 different agents specifically set up with many different, let's say, workflows, like you can follow step by step sequence sequences to do things And then for a big task that would normally take a week suddenly could be reduced down to a few days.
Chris Rainey 20:05
Yeah, or an hour. I think when we think about this in the context of work, I think one things we haven't touched upon, which I think really game changing is the is now how this is allowing us to provide accessibility in a way that we never done before. From a language perspective, from just being no from giving us access to anyone in the organization. Before we couldn't do that at scale, like it was impossible to do that I think what actually really supports 97% The world's languages. What are your thoughts on that? And Matt, as well, as a former CH, Ro, yourself? Who had to manually create every single course in every single?
Matt Burns 20:48
Every single translation team? Yes.
Chris Rainey 20:52
So I mean, right, like, that's something I think we shouldn't.
Matt Burns 20:55
And working in Canada, we have two official languages, you have to communicate everything in English and in French. So yeah, it created a lot more complexity in the organization that now suddenly becomes easier with artificial intelligence. Yeah, I think one thing I'm thinking about a lot these days is source, and source attribution, but also source diversity. A lot of the lessons I learned in my career came from a lot of the same places that are familiar to all of us, Harvard, business publishing, McKinsey. And then watching organizations that I aspired to learn from, like Netflix and Google and just learning from what's happening in the market. And now, I'm asking myself, what's happening in Africa, that's really innovative, what's happening in South Korea, that's changing the way we think about things. Because of for I have to travel to those places, or have personal relationships. And now I can ask artificial intelligence to surface insights from those parts of the world, and combine them within my own regional context, to tweak it and make it actionable. These are things that required time and effort and energy, and they still require those things. But the energy now is spent on deployment and change management, and embedding the practices longer term, which are things we never spent enough time doing. We always spend time in the front part, getting something deployed, and then we walked away because we know the next project. But now we can spend the time on the project actually embedding these practices and creating better cultures. Yeah,
Jack Houghton 22:13
but it's I love I love how you kind of summarized it as basically, how do we find out and bring together ideas from everywhere. Because I mean, I'm a bit of a history nerd. And if you look at the most successful periods in history, for culture, they're usually periods of shared language, and ideas. So pack six Amika, you know, a period of relative peace, where ideas can be shared through a single language from every single person to create something that would be impossible to great leaps in science, or philosophy are these incredible leaps. And as you just said, those is suddenly not worrying about languages, but everybody's input being equally shared, every single person. And I always say, imagine an employee of seven minutes having the knowledge of an employee of seven years. It's a really powerful thing, no matter where you're from, what language you speak.
Chris Rainey 23:03
How do you think the introduction of voice which is going to be coming in the next few months is going to impact that?
Jack Houghton 23:13
We've said it for a while. I think, first off, the interface between us and technology is the most exciting thing. One of the most exciting things for me, personally, I think, I love the changing nature of the interface. And voice is a next leap in that I mean, we've designed everything to have as minimal UI actually, in many ways, because we knew that very soon, you'll just talk to your computer. So you don't want to have UI that gets in the way of that prevents that being possible. And I think SAS is really going to have a real watershed moment. Because if you look at the SAS earning polls, so workday, 15% Share loss, a big company suddenly being valued by less because of deal compression, but I'm gonna predict AI purchasing power is influencing that. And really, what you're gonna get is the people's expectation of experience with technology is simple in a conversation. And Gen AI first companies will have an advantage there. And I think that's why they ended up getting bought out by the big ones. But I think it's a really interesting period of time, how we're just going to talk to computers at work, and it's going to understand me, and it's going to go and do something helpful. And I don't have to worry about UI or interactions. And again, you can talk to any language, any language, it's, it's, and you actually look at that and say, speak to me in a certain way. Yeah,
Chris Rainey 24:29
I love that. Coach me, does a tone you can even ask, Hey, talk to me and coach me in this way. Yeah, that's crazy. So
Jack Houghton 24:36
hilarious video of someone getting taught how to speak Chinese but she said can you speak to me like you're a my boyfriend. She's like, Oh, Baby, I love how you said that are very funny ways of just learning all of a sudden, on
Chris Rainey 24:50
a serious note for organizations that also can convey the tone Yeah, culture, their brand, their mission, their purpose and really start to feel part of it. organization are
Jack Houghton 25:00
exactly right. I mean, you put guardrails on it. So that doesn't happen at work.
Chris Rainey 25:05
Of course, yeah. It's interesting, because you're going to also see that coming into the home, right? So I have like, obviously, we saw it yesterday with Apple intelligence, that you now can just talk to your phone, your we will be able to talk to your phone and have that conversation. But you can now imagine, once, obviously, Amazon is going to get around to it that every single Alexa in the world, people are just going to be communicating, it's just going to become a normal way of working, that will translate we're always a bit behind in the business world, in terms of bringing that consumer grade experience. But when you look at traditional SATA SAS platforms, we don't want to pick you picked on one of them. But there's there's lows, right? You go in there is so Salesforce, there's Salesforce, there's 1000 tabs, there's a million drop down boxes, and it's like, now, all that you UI almost becomes irrelevant. But mean, you can just say, add the deal to deal stage one, add his toes, makes send Matt an email about it, etc. And it just takes that action. So now how what happens is all of that
Jack Houghton 26:13
started, you can learn from how things happened previously. So when content the cost of creating content dropped dramatically, you know, you can it still costs a lot to produce high quality content. Example. But what happened is the big BuzzFeed of the world, they didn't get replaced by the new BuzzFeed they got replaced by influences or 1000s of small companies. Yeah,
Chris Rainey 26:35
media, traditional media and news channels directly laced by social media influences
Jack Houghton 26:40
big sources, as technology becomes cheaper to provide with AI. And as expectations changed the big sales forces if they do, if they do, you know, it was with trying to predict like black swan events, you know, it was but if if they do collapse, they won't be replaced by another big Salesforce, I think it'd be many different providers. So but yeah, that's just that's one hot take. But that's the
Matt Burns 27:03
thing about the reason why that is, and I think anything that's new, has a challenge around trust. And as people see the potential impacts and how AI will permeate through all parts of our lives, both personally, professionally, they're rightfully a little bit concerned about having access to Alexa. And hearing every conversation, we've got me. There's been lots of discussion around tic toc and the impact around how that information is shared, who has access to that information. So I think one of the things we have to look at from an adoption perspective as we move in parallel, and sometimes as technologists, we can sometimes miss this step, because we're bought in, we understand we're totally there. But we can sometimes forget that people are still beginning this journey. So I look back on some of my change management expertise, working in large organizations deploying any technology and say, Okay, we're all starting from different spots. Let's take people on the journey, let's anticipate some of the concerns, they might have privacy, security ethics, and let's have a position on those things. And let's take some actions in those areas, so that we can give them confidence in knowing that those concerns will be addressed. And then they can spend their time thinking about the benefits that can get from the platform, and not dwelling so much on what could happen because they're not wrong. But there can be actions taken to prevent data security, you can get yourself GDPR certified, you can get yourself ISO 27,001 certified,
Chris Rainey 28:17
if you want now isn't now with the new one.
Jack Houghton 28:20
It's just, it's a great business model. That's all I'm gonna say it's a fantastic numbers,
Chris Rainey 28:24
I can't remember what they knew.
Matt Burns 28:25
And we need forms of regulation to harness this technology. Because put in the hands of the wrong people, it can also be used for exponentially not great things in the societal best interest. So I'm one I'm all here for doing things for the right ways. And part of that means taking the time to take people on the journey.
Jack Houghton 28:44
Trust just is a really hard one in this space, I think because because it's as I said at the beginning, it's a technology that went from zero to everywhere, into our public psyche so quickly, so fast. And the media who loves clicks, it's very easy to constantly tell the stories of the bad things a
Chris Rainey 29:02
fear monger and they love it. And people go back to the Wall Street Journal, or
Jack Houghton 29:07
the Wall Street, the Wall Street Journal, I mean, the the sad. Gemini is losing steam. They said it's too expensive. There's no more innovation. And there's a there's reasons for that. There's, there's low adoption and low revenue, and it's kinda like it. Obviously, we're talking Wall Street Journal, which is a public public investment. So they are going to look at it from a quarter by quarter basis.
Matt Burns 29:29
What are they going to say when they launch their own AI?
Jack Houghton 29:34
We work with media companies who do the exact same and they of course they're gonna do it, but But it speaks to a lot. And I think obviously, we're in the space so it's easy to be in a bubble. And I'm sure people listening to this. If they listened to the whole thing. They're probably also interested in trees, maybe in a bubble, but it's easy to forget the wider narrative sometimes and I think the Wall Street Journal captured it quite interestingly. And it's, I always find it interesting to read. But she it's kind of I think, oh Anya holding back all the time. 18 months ago was when they finished with GPT. Four, and they've not released anything major other than voice. So I think they want to keep everybody just near them, then jump ahead.
Chris Rainey 30:11
It's almost like though you do have to have one company to force everyone else out. So what happened? Open AI launched within days, every single other platform launched Google. Oh, okay. Like, out of the woodworks? Where's? Where's this been hyped in exact, so they were kind of forced in the market? And you're gonna see the same thing? Yes, soon as they drop voice. Yep. Everyone else is gonna drop
Jack Houghton 30:36
Apple straightaway partnered with them. Yeah,
Chris Rainey 30:41
we just we partner with open AI, like, you know, and, you know, to security about Elena, Apple spent 30 minutes of that presentation talking about security, the fact that it's not trained, they're not training your content, the model on your content, the fact that it doesn't sit on Apple servers, you know, they went on for a long while about that, which they should. So it's really important. And part of our conversations myself and Matt, that is one of the main concerns from our clients. And we have to obviously explain that. And again, it's such new territory, that most people aren't going to be experts on security and
Matt Burns 31:15
to give people comfort, it's new territory, but it's not the context isn't new. So if you're deploying any cloud based software solution, you're going to have the same kind of concerns. So if you're deploying an HR S, or a payroll software, in fact, I would probably be more sensitive to a payroll transformational project, because it involves people's wages. And if you've ever been in charge of HR or supporting HR when payroll didn't happen on time, I promise you, it's not a fun place to be. Because it affects literally everyone in the organization. And there's meaningful cascading effects when people don't get paid on time. They have automatic withdrawal set up, they have mortgages, they have expenses, etc. So I think that piece is, is critically important. I also think that as we consider the path forward, we have to look at how we deploy this in a way that bridges strikes the balance between doing the right thing, but also challenging our assumptions around this, because if we simply limit ourselves to what we're comfortable with, we're going to miss an opportunity to do some really transformative things.
Jack Houghton 32:18
You're You're spot on. I mean, because I think human inner loops, an interesting conversation that you see a lot in human in the loop. What do you mean by that? So how can I say, especially when we're talking about actions getting taken? When should the human be involved in that process to stop review, feel too much human in the loop makes the whole point the process a bit pointless? Automation is taken away, because there's need to be a human involved all the time. But too little, obviously, is security risk.
Chris Rainey 32:45
What you saw with Google, right, was the recommendation of glue equally.
Jack Houghton 32:49
Yeah,
Chris Rainey 32:51
I deployed it. And I'll retract that, like telling millions of people to eat glue. So there, there needs to be some god warehouse. Yeah, there
Jack Houghton 32:59
has to be so much powder. And as Matt said, a lot of the security stuff is standard. In many ways, it's exactly the same as it's always been. I think what we're now doing is a reinvention process as well of how we interact with models. So we've got a buyer sweet, like a way of looking at what data does this agent have? And what can it because obviously, if it has access to information that is biasing it in a bad way, that's gonna be bad. And we've got people who put, say, 15,000, or 11,000 pieces of content in there in a 24 hour period, huge amount of potential bias. So those are the things that I think will will soon care about more, you know, how accurate is it? How consistent is it? What's his bias? Is it straying off into random directions? Yeah. And how do I test that at scale to identify those outliers, which is something that we've been putting a lot of time in as well. Because really, you can't really simulate 300 conversations or 300,000, or a million conversations as a human. You have to get agents to test the age.
Chris Rainey 33:54
Think back to Matt's earlier point. That's what's exciting about what you're doing at mindset. And what we're doing Atlas co pilot because it's, it's based on our knowledge that we put into it. So you're not just getting a random answer back from a worldwide web. And it could could have grabbed it from a random Reddit. Comment somewhere where people say all sorts of weird and wonderful things, right, is trained on our corpus of content. We've uploaded every single every single piece of content is being curated by us in the team been checked by us and his team, uploaded by us in the team. But of course, we're not, you know, manually looking at the insights and doing that, because it's not possible. But that's where the AI does the hard work for us. Well, I
Jack Houghton 34:33
mean, how many hours have you probably spent making sure the experience is perfect? hundreds of hours and making sure that experience is perfect? Yeah, exactly. You know, it's, it's what matters?
Chris Rainey 34:41
Yeah. You said to me before that the way that most people think about AI adoption at work is wrong. Could you repeat that? Like what did you What did you mean by that?
Jack Houghton 34:53
I guess it's the the b2b to b2c sort of idea, which is I think people are obsessed with top down adoption. We like to measure how many companies have adopted it as across the company. And they're just so slow, you know that they're much slower than a consumer who just picks up buys it and does it. So I think the way we're measuring adoption right now is in the Wall Street Journal is a good example, the measuring adoption and revenue from an enterprise perspective, most of the money will be coming from individuals. And over the next 12 months, already, you're seeing every company is now starting to adopt. So I think adoption, you can view it from a bottom up or a top down approach. And I actually really liked the bottom up approach of measuring adoption, because that's where the use case value gets identified. What is the right prompt library to create for a team? You know, based on what questions people are actually meaningfully trying to answer?
Chris Rainey 35:41
Yeah, it's kind of one of the reasons that when we're building our agents right now, we're actually basically crowdsourcing, how we do that, for example, we have about to launch our people analytics agent. And we've spent months working with 20. Vice Presidents are people analytics, and the world's leading global brands, to help us build that Prop library. People that live and breathe it every single day. So that when we put that into the hands, so that's not that's from bottom up is to people to actually living it, right? Yeah. So when we deploy that and put it in the hands of our users, they know that it's coming from people that know that live and breathe the job, and understand and
Jack Houghton 36:18
the questions are there for them to learn. Because ultimately, like when you give someone a blank slate, like Google, first question, what is the color of the sky, so they don't know what to ask. And this is such a powerful new technology sometimes. And that, and some people are really good, some people on it, they know what they want. And then some people just need that guidance. And that's why proced like, that type of thing is so powerful, it gets people adopted really quick. Yeah.
Chris Rainey 36:42
What do you see the next wave, we spoke about voice mat as well for you. Would you feel like the next focus on wave of innovation is going to be around.
Matt Burns 36:55
I think I'm interested in dynamic learning pathways, in particular, the opportunity to surface a personalized set of content, articles, learning products, to help broaden and deepen your knowledge. But also in the flow of work, I think it's not as realistic to expect people to check out of the office for two weeks or six months or two years to go to a traditional university experience. I think people need to acquire skills in a number of ways. And I think one of those ways can be in the flow of work, where you're applying those skills in real time, and therefore, embedding that knowledge and to me the opportunity to learn. And the opportunity to do so in an agile way is something that I think a lot of people are looking for, at a time when a lot of skills are becoming replaced with new kinds of skills. And I don't know about you, but I'm asking myself, like, what got me here, which as a corporate executive was running really good meetings, is may not have as much relevance in the future state that involve AI. So how does it have to evolve? And I've been doing a lot of research on large language models and their applications of different industries, and asking myself lots of questions around macroeconomics and how they might affect broader work trends. And those things to me are really interesting. And they're made easier by artificial intelligence, because I can pull on a curiosity thread, and then go deeper or broader based on what I would do, which I normally go down a YouTube rabbit hole, or, you know, Google rabbit hole. But now I can get more curated, more specific more in the flow. And it feels less like I'm wasting time on social media and more like I'm investing time in my future.
Jack Houghton 38:27
Yeah, that's, um, it's quite an exciting time. And I think I was just reflecting on the question, I think there's a few different things for me, I guess there's two parts is the stuff that we're working on that I'm incredibly excited and stuff that I see coming very soon. Stuff that I'm excited about is the ability to just simply describe to an agent, the experience it should provide to users and people. So literally in say, store, their name is data forever as memory to help influence and support that journey. The Pathways piece, create a survey using our own assessment using all the content you got access to don't make it too hard. And just running it and taking people through experiences. It's a key part of that agentic workflow. And I think when you think about this, in the context of deployment is the other part I'm really excited about. So right now we're in the early stages of deployment still in the world that people are just seeing, you've got on device agents.
Chris Rainey 39:24
Yeah. Apple, Apple on device agents. Yeah. That was a big, big part of that announcement that they're trying to make. Say that, hey, most of this is gonna happen on device. Yep. And then when we need to, then we will sort of go out to GBT for a prompt.
Jack Houghton 39:38
Exactly. And that's an agent. That's what they're doing. That's an agent there. You've got on computer agent. So in video run, Microsoft just launched it onto your computer. You've also got so many different we've got multistep agents, just where we all are, which is that how do you take people through a multi step process without us having to craft these new experiences every time There's all these new forms of agents that are emerging cloud based, every website is going to have their own agent, which will store your passwords. And so you don't need to log in anywhere, it'll just know. Yeah. And
Chris Rainey 40:10
we've seen that even with the websites, agent that Apple spoke about that now, when you visit our website, the agent gives you a summary of the website, and gives you the key takeaways in a little box that pops down. So rather than you have to scroll through the website to be like, what's the address, so what's, what's this, this, it just says, here's all the top line information immediately accessible for you. And now the ability to do that with our employees in the flow of work to surface those insights of them. They don't have to go and search in an intranet to see what their benefits are, didn't have to go and message HR wait three days to see me holidays got left simple things like that, that frees up HR leaders to work on strategic work. And actually, you know, the work that they want to be. And I'm with Matt, I love the idea that for the first time we used to use we can use AI to create customized learning experiences at scale is just a game changer. Yeah. So rather than having a static online course, where you go, module 1234, and four, the modules are not even relevant to you, and you're going for it to move into a dynamic learning pathway, which Apple also built, demonstrated in their thing where you can say, here's all our content, create me a course around how to lead remote teams, and it goes and grabs that content and builds it specifically to you. And because it knows you're in this particular country as local context to to how you do it in that region. And now you have a personalized experience, to be able to do that, that is going to be so exciting. I
Jack Houghton 41:40
know, I've just been working on the inside. Once you've gone through that process, people you can't really go back. You can't you can't like you can't know imagine and an alternative scenario where this isn't the way of doing it. Yeah, and the way we should experience it as people in a conversation with a voice, you know, call
Chris Rainey 42:00
for a wrap up for all the HR executives and listen probably gonna go way beyond HR executives, really. They're gonna be listening and parting advice you get for them. And then we're gonna say, where can they connect with you and check you out? Create
Jack Houghton 42:11
a culture whereby every single person knows they can come to you with ideas for different use cases, different ways of doing things, and do it with open arms. That's that's always my parting advice to every single leader out there right now. I'm on LinkedIn. Go to mindset.ai on Google as
Chris Rainey 42:31
well, that find
Matt Burns 42:32
advice. It's hard not to find me these days.
Chris Rainey 42:34
That's your advice. Yes, hi devices is harder to find me these days.
Matt Burns 42:38
And when people ask me what AI is most powerful, where the most powerful AI answer as answered before, which is I think it's really good at synthesizing large sums of information and curating the right information, and then helping you take actions that create really compelling stories.
Chris Rainey 42:55
Yeah, love it for everyone listening all those links below to connect with Jack mindset map, as well as also, if you're not already using Atlas copilot, you can start using it right now for free. There's a link below. Go check it out.
Matt Burns 43:10
I use it two hours a day.
Chris Rainey 43:13
Thanks so much for the comments. It was great chatting with you. It
Jack Houghton 43:14
was it was a pleasure. Thank you
More from the HR Leaders Podcast
Tony Buffum, VP of HR Client Strategy at Upwork.