Latest / CXOTalk / Will AI Agents Replace Your Entire Workforce in 2025? | CXOTalk #876
Transcript
- 0:00Are AI agents an enterprise savior, workforce apocalypse, or
- 0:06just another tech bubble waiting to burst?
- 0:09Today on CXO Talk episode 876, we cut through the noise with
- 0:15Bill First, CEO of HFS Research and one of the most respected
- 0:21industry analysts in the world. When we talk about AI agents and
- 0:26agentic AI, what what are we actually referring to?
- 0:29What do we mean by that? It's really the ability to
- 0:32replicate human behaviour in software.
- 0:37It's as simple as that. Whether it's mimicking our
- 0:39voices or supporting us in doing our day-to-day work, It's, it's
- 0:44really like the augmentation of humanity and software.
- 0:47And we we talk about the blending of, you know, humans
- 0:50and technology. This is really where it's at and
- 0:54it's it's it's something that, you know, we dreamt about for a
- 0:58long time, but there's only really starting to come into
- 1:01reality, but at an alarming pace.
- 1:03I don't know, you know, we see, you know, a lot of fun things
- 1:08flying around on X and LinkedIn and all these types of things.
- 1:11It's just incredible, you know, how much development there's
- 1:15been in voice and video in in just the last six to nine
- 1:19months. So we're we're going through a
- 1:21complete revolution and Agentic is right at the front of that
- 1:25from a technology perspective. Why is agentic AI so important
- 1:30and at this particular time? I'd like to rewind back to the
- 1:35early 20 tens when you might have heard of a technology
- 1:39called RPA, Robotic Process Automation, which got very big
- 1:43and very height within the technology world.
- 1:46We actually coined the phrase alongside a company called
- 1:49Blueprism when we launched it in 2012 and we did the first
- 1:52analyst papers on it. And at the time we were talking
- 1:55about RPA replicating human, human behaviour in software,
- 2:02which would allow us to scale more effectively, threatened
- 2:07elements like offshore outsourcing, because companies
- 2:10could technically consider having less offshore resources
- 2:14when you can automate a lot of this stuff.
- 2:16But the problem with RPA was the technology didn't scale well.
- 2:20It was very brittle. But the the concept was there.
- 2:23But that was really all about following instructions, easy,
- 2:27easily. It was about eliminating manual
- 2:30effort waste, you know, which was wasted on repetitive tasks.
- 2:33Then everyone remembers the influx of gem AI nearly nearly
- 2:37it's going to be its third year with ChatGPT really came public
- 2:43nearly three years ago and that really changed the game in terms
- 2:47of it became the productivity amplifier that accelerates
- 2:51creative and analytical work that really bottlenecks humans.
- 2:55It's the ability to create content and this is like one of
- 2:58the first times we've had non-technical people have that
- 3:02ability to start to create content, create data, augment
- 3:07their work, create code. Even.
- 3:09You know, there's a lot, lot of discussions going on around how
- 3:12much code can be eliminated now because of Gen.
- 3:14AI. And that was all about creating
- 3:16based on prompts. Now we're into the Gen.
- 3:20AI phase, which is about understanding goals and figuring
- 3:24out how to achieve them. So Gentic AI is a collaborative
- 3:28actor that removes the need for constant human oversight of
- 3:33complex processes. It's self directing in many
- 3:37respects. It coordinates multiple tasks.
- 3:39It transforms entire workforces, it creates new organizational
- 3:44paradigms. But it's not about the fact that
- 3:47it sounds great. What's exciting about a genetic
- 3:50is it really does work and it's and it's working at an alarming
- 3:55pace that is making in, in reality, many people are
- 3:59comfortable. Some people are loving it and
- 4:02they're embracing it and they're realizing, wow, I can do my job
- 4:04so much better. And I'm I'm an analyst.
- 4:06I can tell you how Jane Turk and Jenna are helping me do my job.
- 4:10But this is the most, I think, impactful wave in this AI
- 4:14continuum that takes us to the next phase, which we're terming
- 4:19artificial general intelligence, which is much more self-directed
- 4:23intelligence that overcomes human cognitive limitations
- 4:26across all domains. And eventually, you know,
- 4:29artificial super intelligence, which is about computers
- 4:33outperforming humans. We're not there yet, obviously.
- 4:37But you know, I watched Terminator, Terminator One with
- 4:40my son the other day. I hadn't seen that in about 30
- 4:43years. And that brought me back.
- 4:45They actually predicted in 2029 was when humans, computers
- 4:49become self aware. So that if anyone's got nothing
- 4:52better to do this weekend, watch the rerun of Terminator One.
- 4:55It's uncanny. Oh, they got the timings right
- 4:59on this thing. So I want to latch onto a
- 5:01particular point that you made. You said it really works, right?
- 5:07Can you elaborate on that? Because that's kind of the magic
- 5:11point, right? This is not a lab based theory,
- 5:16it's something that actually works.
- 5:18So tell us about that. Most companies right now have
- 5:21created some stand alone single agents.
- 5:24So that's but that's an agent that can handle maybe one
- 5:28specific task or function. So that could be like an e-mail
- 5:31writer or even a meeting scheduler, things like that.
- 5:34Even like copilot, you can use copilot right now to summarize
- 5:37your emails and remind you to do things.
- 5:41Or you can use Fireflies, which is a really popular tool, or
- 5:44LinkedIn, not so much LinkedIn, Zoom AI needed to summarize
- 5:47meetings. Those are single agents, believe
- 5:49it or not, and they're already working.
- 5:52People are already, you know, pretty excited about or did you
- 5:55turn your fireflies on? So we got a good summary of this
- 5:57meeting. Where this starts to get really
- 6:00exciting is when we start to build functional multi agents,
- 6:03which is multiple agents work together within a single
- 6:07business function. So that could be a sales team of
- 6:10agents handling prospecting or qualification and follow-ups,
- 6:14that sort of thing. And eventually we get to
- 6:17something we're calling horizontal multi agents, which
- 6:20is where you get different agents collaborating across
- 6:23various business functions and and even other supply chain
- 6:27partners. So that could be sales agents
- 6:30working with marketing and customer service agents.
- 6:33So you're actually building out capabilities and business
- 6:37functions beyond one single function.
- 6:40It works because you just got to try it like I, I, I'd love to.
- 6:46There's a demo, it's called Super Film, I think it was where
- 6:49you could actually put an avatar of me in our research website
- 6:53and ask me questions and I would literally in my voice dig into
- 6:58our research and communicate them back to you using using
- 7:02voice. You just got to see it to see
- 7:04how effective this is. Now, is it perfect?
- 7:07No. Is it as accurate as talking to
- 7:10a human being? Not yet.
- 7:13But in many respects, we're creating agents that are
- 7:17becoming very, very supportive in our jobs.
- 7:20I mean, I'll, I'll tell you, for example, I wrote a piece on
- 7:23tariffs the other day and I put together here's like big
- 7:27tournament and things about terrorists and people might have
- 7:29even read it. And for a bit of fun, I pumped
- 7:31it through Chatchy, BT pro and and outcomes.
- 7:37I said, can you pump? Can you can you replicate this
- 7:39using Phil first voice just for a bit of fun.
- 7:42And it came out sounding like the sort of thing that I would
- 7:44have written. And then I asked it to turn it
- 7:46down a bit, that sort of thing. And then I produced another
- 7:49piece. Well, I said, can you produce me
- 7:51a chart that shows life expectancy in the US and versus
- 7:54other countries and health issues?
- 7:56And it starts putting information all over the place.
- 7:58And then you start to train, train the model.
- 8:01So start starting to become your own personal agent.
- 8:03And it's getting to know what I need.
- 8:05And then it's like, can you produce this in the HFS?
- 8:08Find some colours and you're programming in the colours to
- 8:11use and everything. So you're really building out
- 8:14something that can literally become your go to at work, you
- 8:18know, so there's so many different uses.
- 8:20And, you know, I don't even know if we're going to call these
- 8:22agents in another 6 or 12 months, but this is just how
- 8:25we're partnering with technology now where we don't have to go to
- 8:29people to get things done all the time now.
- 8:31We can get so much done ourselves and then we as human
- 8:35beings become the creators of that content.
- 8:39Like we say, I've got a big business meeting to go to
- 8:41tomorrow. You're going to get so much of
- 8:43content that you need you. You set your agenda to the
- 8:46technology, it gathers you to what you need and it allows you
- 8:50to then curate that to make you effective as a human being.
- 8:54How is this different from having an interactive chat with
- 9:00Chat GT? So what's unique about agents?
- 9:03But before you answer, I just want to remind everybody, our
- 9:07regular listeners know this, that you can ask your questions.
- 9:11So right now, if you're watching on Twitter, pop your question
- 9:15using the hashtag into the into Twitter using cxotalk #cxotalk.
- 9:21If you're watching on LinkedIn, pop your question into the chat.
- 9:25This is your opportunity to ask one of the top analysts in the
- 9:30world pretty much whatever you want.
- 9:32So take advantage of it. And we have some questions that
- 9:35are coming in now. But first again, so you're
- 9:39describing an interactive process.
- 9:42You go to a meeting, you give it notes, you then say modify this
- 9:46or modify that sounds like ChatGPT.
- 9:49How are agents different from LLMS and and our usage as we
- 9:53know and love it today? ChatGPT is useful because you
- 9:58can use it for a specific prompt.
- 9:59So I need some information on this or that get some
- 10:02information quickly produce this, do that, do this, do that.
- 10:06An actual agent is the virtual Co worker who is completing end
- 10:11to end processes for you. So it's self directs and
- 10:14coordinates multiple tasks. So once you've so say I'm using
- 10:18the example of I could use an agent to help me with my
- 10:20research. I would develop train this as a
- 10:25virtual Co worker to be like my research assistant.
- 10:28So it would start to overtime learn what I do, what I need,
- 10:32how I do it. So you can start to interact
- 10:34with this like a virtual Co worker, like a research
- 10:37assistant, for example. And you can leverage this to,
- 10:40you know, create whole new organizational paradigms.
- 10:43I'm not joking. In 12 months time you can say,
- 10:46hey, they feel I need to have a meeting with you next week to
- 10:49talk about XYZI can literally have you talk to my agent like
- 10:54who will look at my diary and coordinate what I need in time
- 10:57and maybe ask particular questions.
- 10:58So we are training virtual Co workers to do the jobs that we
- 11:04either used to do ourselves or someone else did.
- 11:07And we can start to get into real examples of this.
- 11:10But the challenge is, you know, going to your staff in another
- 11:14company and asking them to almost recreate their jobs into
- 11:20software, which is very different than saying train up
- 11:23another human being and transfer tasks from yourself to another
- 11:27human. We're now expecting people,
- 11:29including ourselves, to transfer human work tasks into software
- 11:36That technically frees us up to do other things.
- 11:38Oh, let's be honest, could make us redundant, right?
- 11:41We're not needed anymore. We can actually leverage agents
- 11:45to do the jobs of humans. And we're now seeing enterprises
- 11:49who are really trying to have an AI first mindset.
- 11:53They're now insisting before you hire any new staff, you need to
- 11:57show that this work can't be done by AI.
- 12:00We've reaching that point quite quickly.
- 12:03This wasn't like maybe 20 years ago.
- 12:04People used to say, hey, if you hire new staff, can we see if
- 12:08that work can be done offshore in the Philippines or India or
- 12:11something? Now C-Suite directives are, can
- 12:14we not do this with AI? So the whole point of agents a
- 12:17really this ability for companies to grow and scale in a
- 12:22way that you don't need to keep adding more and more people.
- 12:24You do a lot more with the people you have.
- 12:28And I think that's the positive way to think about this.
- 12:30It's, you know, I want to run a marketing campaign.
- 12:33Can I develop a planning agent who can coordinate and breakdown
- 12:38the campaign requests into specific tasks?
- 12:41Can I create a research agent that gathers market
- 12:44intelligence? Can I create a creative agent
- 12:48that develops, you know, creative assets and messaging?
- 12:52Can I develop a strategy agent that optimizes my campaign and
- 12:57engagement across marketing channels?
- 12:59Right. You can go on and on about
- 13:01almost every new staff member you need.
- 13:04You can create an agent for like it could be, hey, I need
- 13:07somebody to manage social media and I need to do automated
- 13:11LinkedIn updates, that sort of thing.
- 13:13Or it could be, you know, even a campaign coordination agent to
- 13:18synthesise inputs from all agents into a cohesive campaign.
- 13:22So it's it's creating people into software.
- 13:26So, you know, we call this thing, you probably heard of
- 13:29services as software, but this is what's happening in the
- 13:32services industry right now is companies are starting to think
- 13:37about how can I replicate the services I'm receiving from an
- 13:41IBM or an Infosys or one of these companies and receive this
- 13:45using agent agentic software versus why do I keep having to
- 13:50add more people all the time. So that's the real nub of
- 13:53Agentic and I think why it's causing, you know, excitement
- 13:57and friction at the same time. Go to cxotalk.com, subscribe to
- 14:01our newsletter and join us. Join our community and join our
- 14:05live shows. So we have a a very interesting
- 14:09question coming from Twitter from Anthony Scrifignano, who is
- 14:14the former Chief data scientist of Dun and Bradstreet and has
- 14:19been a guest on CXO Talk a number of times.
- 14:23And he's asking about the unintended risks or unintended
- 14:30harms that can emerge. Can you talk about that aspect
- 14:36of it? I think lots of people are are
- 14:38concerned about the impact of AI agents on the workforce.
- 14:43You spoke about the positive aspects, but what about the
- 14:46unintended risks? First thing is you've got the
- 14:49more general risks of AI. So when you read something sent
- 14:53to you now a lot of people are thinking was this written by AI
- 14:57or human being right. That's that's a big problem and
- 15:01I see that as an opportunity for a research firm like us, because
- 15:03it's people need real more than ever.
- 15:06And do you trust information that's all produced using agents
- 15:12and genetic software? Is it is it truly trustable?
- 15:15Is it reliable where the source is coming from and that sort of
- 15:20thing. And I think a bigger risk right
- 15:22now is can you trust information from people?
- 15:28The other issues, obviously with interactions and hallucinations
- 15:32and these types of things, all the types of teething problems
- 15:34you'll have with honestly any type of technology.
- 15:38You know, you, I remember when we were getting into more
- 15:41sophisticated accounting applications 20 years ago when
- 15:44people worried then about software malfunctioning and
- 15:47producing, you know, incorrect calculations and stuff like
- 15:50that. So a lot of it is trusting the
- 15:52software, trusting the security of that software as well, and
- 15:56understanding how to navigate your way around this climate
- 16:01because it's only going to get more confusing and more
- 16:03worrying. And, you know, three times, you
- 16:06know, we're getting more sophisticated spamming phishing
- 16:11stuff or, you know, we're getting them everyday and texts
- 16:13and emails and quite convincing ones.
- 16:15Sometimes it's like, you know, I get I get my own staff coming to
- 16:20me saying, hey, Phil, did you send me this text about getting
- 16:23Amazon vouchers, things like that.
- 16:25So we can talk about this for a very long time.
- 16:29All the different risks, all the different worries, all the
- 16:31different concerns. And the other thing is, you
- 16:36know, where do you want it to impact?
- 16:38So you know, I sat in a room with 10 senior level AI decision
- 16:45makers in banking just a couple of weeks ago and they had a lot
- 16:50of common issues, which was we really want to leverage Agantech
- 16:54to improve our customer experience function.
- 16:56But you know, it's all about the customer experience with using
- 16:59banking apps and technology and that sort of thing.
- 17:01And a lot of their customers, they still want to talk to a
- 17:04human being, right? Especially when you're getting
- 17:06into your finances and, and, and, and loans and borrowing and
- 17:10that sort of thing is, you know, how far do you go before you can
- 17:14truly trust the technology versus versus the people?
- 17:18And and I think, I don't think we have a full answer for that
- 17:20just yet. We have a question from Arsalan
- 17:23Khan, who says agentic AI requires the correct data at the
- 17:29right time with the right human and systems integration,
- 17:33eventually these agents become autonomous.
- 17:37What happens to humans then? So he's asking about this
- 17:40boundary between human work and autonomous agent work.
- 17:46This isn't just about enterprises trying to cut costs
- 17:50from the place people with bots. This is about us as human
- 17:54beings. We're all, we're all threatened
- 17:56by this and we all have opportunities with this.
- 17:59And if you're in a job where you can be effectively replicated
- 18:04and replaced, you kind of you kind of know that and you need
- 18:08to figure out how do I continue to add value in an enterprise?
- 18:12And I think the value comes from collaboration.
- 18:15It comes from people skills, it comes from empathy.
- 18:18And if you can become a great person everybody likes to work
- 18:22with and you become very thoughtful about what you do and
- 18:25you start to collaborate beyond your existing area, you become
- 18:28very valuable to your company. And you know, I can, I can go
- 18:32through many examples of this. You know, I have a, a guy
- 18:36running my IT systems who actually was a procurement guy
- 18:39just a couple, about 3 or 4 years ago, But he, he broadened
- 18:43his knowledge into understanding how to manage HubSpot and
- 18:49accounting software. He manages our stack of social
- 18:52and Grammarly and all this sort of stuff.
- 18:54And as part of his job, he started to get to know all his
- 18:57colleagues in different departments in the company, like
- 19:00analysts and, and, and finance and HR and all this sort of
- 19:03stuff. And then you start to develop
- 19:04real value to deliver across your organization.
- 19:08And I think no matter what role you're in, if you're on a sales
- 19:11role, you're on a delivery role, you're on a tech role, you need
- 19:14to become broader and more aligned to your business to add
- 19:19value there. Because if you if you become
- 19:23just a replicatable solo task driven professional, you do run
- 19:29a risk. So you know, you wouldn't
- 19:32believe some of the conversations I have with CIOs
- 19:35right now who are under immense pressure to wipe out costs
- 19:39because of code. One major organization I spoke
- 19:43to produces half a billion lines of code a year to keep that
- 19:46organization functioning. And they've been tasked with
- 19:50eradicating 90% of the effort because you don't need to have
- 19:54armies of legacy coders anymore. A lot of these code code can be
- 19:59rewritten using Gen. AI and other types of AI
- 20:04software now. So, you know, we're, we're just
- 20:07all facing the challenge of how relevant are we now?
- 20:10I think you can't replace the humanity and the human ability
- 20:15to be empathetic. To collaborate, to energize
- 20:19people and to curate content that is real.
- 20:22I still believe and I think more than ever, we're going to be hit
- 20:26with so much AI. Fake information, or could be
- 20:30real information, but it's written by AI.
- 20:32We want to read stuff written by people.
- 20:35AI can help us as humans get much better what we do.
- 20:39It can help us become better communicators, maybe want more
- 20:42productive, get more done, like I told you earlier.
- 20:45And I find I'm becoming way more productive as an analyst because
- 20:49I've now got, you know, some AI tools which can develop charts,
- 20:55synthesize data, get me some bits I want so I can, I can
- 21:01answer my questions. Be specific on how these agents
- 21:06help you and your job. Tell us the tools you're using,
- 21:08and then we're going to go back and get some more questions.
- 21:10Questions are coming in. If you want to develop real
- 21:13value within your own organization, you have to run
- 21:15boot camps with your own colleagues to present to each
- 21:19other how you're using these tools to be more effective at
- 21:22your job. We've even, we've even hired an
- 21:24AI expert who's a full time employee within our company.
- 21:28She's probably listening to this that who's actually working
- 21:32across our operations people, our analysts, she's working with
- 21:36Amazon and, and, and a company called Lizier, for example, to
- 21:39identify how we deliver our research to our clients.
- 21:43So while yes, I, I can go on about the personal Productivity
- 21:46Tools I use, we're using agentic to transform our whole business
- 21:50because we're in the information business and we have set up a
- 21:56fairly complex system. We're using an agentic solution
- 21:59called Lizier, which is a, it's a start up, but it's, it's in a
- 22:03pretty mature phase. They're very popular and they're
- 22:06powered by AWS to produce at scale the ability for we have
- 22:12like 150,000 subscribers to go in and create their own research
- 22:18support agents to help them leverage, get the most out of
- 22:22HFS. So that's how we're using it
- 22:24from a corporate standpoint, from a personal standpoint,
- 22:27right now I use, I'm using chat CPT Pro.
- 22:31So I paid the extra money. I'm not sure I need the $200 a
- 22:35month package, but I'm loving it right now because it gives me a
- 22:38lot of query time. It, the computing power is a
- 22:42little challenging. Sometimes it takes a bit of time
- 22:44to produce everything I need. I'm finding that effective.
- 22:48I'm using deep research from Perplexity, which is pretty good
- 22:53as well. And I've also been experimenting
- 22:55rather tools like Claude, which is the anthropic tool, And I've
- 23:01also looked at some other tools that can be fairly effective,
- 23:05like Gemini, I'm still not completely convinced by, but
- 23:08other people love it. So a lot of this is, you know,
- 23:11people finding technologies that they think are better than
- 23:13others and they like the way they're interacting with these
- 23:16tools. But the new, the new suite from
- 23:18ChatGPT Pro is excellent. You've got the image creation,
- 23:21you've got the operations piece, you've got the deep research
- 23:24piece. What I'm seeing right now, this
- 23:27thing is pretty good and we're going to get to a stage fairly
- 23:30quickly where we're going to be whittled down to maybe 3 or 4
- 23:33powerhouses in this space who are going to be dominating the
- 23:37progression here. I use so many different LLMS,
- 23:41I'm always experimenting to see which one is better.
- 23:43Here is a question from Wes Andrews who says you jokingly
- 23:51referenced Terminator earlier, but given the struggles that
- 23:56that AI and other sectors are having with establishing
- 24:01frameworks, guardrail standards such as NIST and GDPR, what do
- 24:07you suggest? And I'll just mention also to
- 24:10folks that last week we had two members of the House of Lords
- 24:14from the UK discussing these issues.
- 24:16So if you care about these issues, listen to our last show
- 24:20and you can get the transcript on our site.
- 24:21But Phil, what what about this this framework and guardrails
- 24:25set of issues? We look deeply into this because
- 24:28we cover Global Services a lot with an HFS and every different
- 24:33region has slightly different attitudes towards AI.
- 24:37So obviously you mentioned GDPR is, is huge in the UK and
- 24:41Europe. India is a little bit more of a
- 24:45free fall right now with how they're accepting AI based
- 24:49solutions and US, you know, this could be the second coming with
- 24:54the tech Bros driving a lot of policy here.
- 24:58So I think we're still waiting to see how a lot of this shapes
- 25:00up. EU has typically been the most
- 25:04closed from a framework perspective and demanding in
- 25:08terms of compliance. And anyone running a business
- 25:11knows how challenging running GDPR practices has been in
- 25:16recent years to get to, to the other side.
- 25:21But I, I do think that as this continues to evolve, the need
- 25:26for common frameworks is going to become more and more
- 25:29paramount and the need for cooperation is, is going to
- 25:34continue to proliferate. They're really doing, look
- 25:37what's going on politically across the world right now in,
- 25:42in many ways, this is going to actually bring I think a lot of
- 25:45regions close to the governments and regions close together and
- 25:48which may actually drive better cooperation with AI.
- 25:52So for example, I was hearing today about a strong movement to
- 25:58create the China less supply chain, right?
- 26:01So how can countries start to group together to manufacture
- 26:05goods outside of China to avoid these potential tariffs, right?
- 26:10And in that case, you need to sort of build a supply chain
- 26:13competency that sensors and responds, that manages
- 26:17inventory, that brings cooperation together and these
- 26:20types of things. So I, I think the need to build
- 26:24and supply chain standards, trading standards, you know,
- 26:27around AI, I, I, I think this is just going to, it's only just
- 26:32beginning and we're going to see a lot more of this emerge in the
- 26:35next couple of years. What about enterprise adoption?
- 26:39Where are we today? AI agents are still relatively
- 26:45new. There's lots of promise, but in
- 26:48terms of actual usage and enterprise adoption?
- 26:52I can share the latest and greatest that we've been working
- 26:55with. We spoke to over 1000 major
- 26:59enterprises looking at the adoption of of Gen.
- 27:04AI and the Gen. tech and 45% of them are either worried about
- 27:13job loss or they're resistant to change and adoption is I'd say
- 27:18fairly diminished. The other at the end of the
- 27:21spectrum, only 15% of AI leaders are generally positive about AI
- 27:29adoption and they have fairly integrated views of where
- 27:34they're going. They have a strong culture of
- 27:36support and they're they're embracing this.
- 27:39And then in the middle, you've got about 40% of enterprises
- 27:43where they're still in that sort of pilot purgatory phase.
- 27:48Their culture is becoming more adaptive.
- 27:50They're recognizing the benefits of AI, but they're not there
- 27:53yet. So in terms of actual adoption,
- 27:56you've only got about 15%, maybe a little more, who are getting
- 28:01to the point where they have a real clear vision and
- 28:04understanding of where they're going.
- 28:07One thing that is crystal clear is we're seeing immense pressure
- 28:10coming from the board level people and also C-Suite leaders
- 28:16in organizations to drive AI adoption a lot faster.
- 28:20There's real pressure coming right from the top to really
- 28:24embrace and become more effective as you know, AI first
- 28:28cultures. So, but the reality is we're
- 28:30still at early days, You know, we, we, we've been talking about
- 28:34this. It's, you know, for a long time,
- 28:37but the reality is ChatGPT 35 only came in not even 2 1/2 to
- 28:42three years ago. So we're playing catch up, but
- 28:47what's happening is the technology is staring it on our
- 28:49face. It is really here we've got big
- 28:54firms really trying to get on top of it.
- 28:56You've got the big software companies like Salesforce in
- 28:59particular with their Agentforce roll out and service.
- 29:02Now somebody's business is really trying to muscle in on a
- 29:06gentic because they see that as an opportunity to take market
- 29:09share away from services firms. And at the other flip side,
- 29:12you've got services companies like Accenture really trying to
- 29:17become more dominant in the services of software realm as
- 29:20well. So adoption is alone is the is
- 29:23the real answer to this, but the pressure is there and and it's
- 29:26on like never before. I just want to invite everybody
- 29:30listening to join the CXO Talk community.
- 29:35Go to cxotalk.com and subscribe to our newsletter so we can
- 29:40notify you of upcoming conversations.
- 29:42We do this every week and you guys who are listening, you are
- 29:47the cream of the crop. So join, subscribe to our
- 29:51newsletter and join these conversations and add your
- 29:55points of view and your questions.
- 29:57So we have an important question from Arsalan Khan who says how
- 30:02do you convince the C-Suite that agentic AI is not just a fancy
- 30:08chatbot before they move on to the next shiny object?
- 30:12What are the challenges and the opportunities associated with
- 30:17this we're. Past that point where C-Suite
- 30:21can keep denying that this is just another fancy chat bot.
- 30:24I think if, if you're leading ACX function in particular, you
- 30:32know, you, you, if you, if you're not familiar with how
- 30:36easily replicable call centre Asians are with Jane, you know,
- 30:41with, with smart agents, you know, right now you shouldn't be
- 30:44in a job anymore. To be quite harsh about it.
- 30:49I can tell you, you know, just an example of an organization I
- 30:53spoke to with 50,000 onshore, onshore staff responding to
- 30:57healthcare inquiry calls and the, the leader basically said,
- 31:00look, the bottom line is, is there's the same 6 questions
- 31:04being asked over and over and over again.
- 31:06We've already run the analysis. We can literally replace half
- 31:08these people with, with intelligent bots and they call
- 31:12them, they call them empathy bots very, very quickly.
- 31:16We're not going to do it straight away, but we know the
- 31:18possibility is there. And I think this is this is a
- 31:22typical case across a lot of companies is they're very aware
- 31:25of what they can do, but they they're still yet to have that
- 31:28burning trigger platform to go do it.
- 31:31My concern is if we plunge into, you know, a, a deep recession,
- 31:37you're going to see some organizations literally come out
- 31:39and say, we're just going to start relying a lot more on AI
- 31:42and we're going to let people go.
- 31:44So my vain hope is we don't fall into recession so we can have a
- 31:49more positive view of people and technology.
- 31:52But there is that risk there that a negative economy can
- 31:56drive a lot more weaponized AI where companies would say, look,
- 32:01well, let's just replace these people, we don't need them
- 32:03anymore. So I, I don't think companies I
- 32:07speak to are not aware of this. It's more how advanced they are
- 32:11with embracing this. Are they prepared to do
- 32:13anything? And my concern is I do talk to a
- 32:16lot of enterprises. We have a lot of summits and
- 32:18round tables on top of our research where, you know, people
- 32:21want to talk. But when it comes down to what
- 32:23are you actually doing, They're not doing, They're not doing a
- 32:26lot. And I think what I just said to
- 32:27about the 15%, that's not a big #15% are kind of on the path.
- 32:33The rest are either still figuring it out or they're not
- 32:36on the path. And and this is just going to
- 32:39become more pronounced as we go through the next few months of
- 32:43macroeconomic turbulence. You just made wild comment,
- 32:48which is, and I don't want to put words in your mouth, but it
- 32:51seems like you just said that the technology is becoming so
- 32:56good at sufficient number of use cases that an economic downturn
- 33:06can push many companies to replace many workers because
- 33:12those use cases and the effectiveness are so broadly
- 33:17dispersed even even today or if not today soon enough.
- 33:21The technology is available. It's there.
- 33:24I think companies are aware of it.
- 33:26I do believe as well most enterprises don't tread lightly
- 33:30on the fact that, hey, let's go replace 5000 people with, you
- 33:35know, 1000 or 500, you know, or get augmented consultants who
- 33:44can manage a team of bots. But one of the things that has
- 33:48been looked at in industry is why do you need 500 people in
- 33:54India, for example, running a bunch of coding or app support,
- 33:59that sort of thing, when you can potentially replace them with a
- 34:02team of maybe 25 people who are local and onshore who are
- 34:08supported and augmented by Gentic technology.
- 34:10So it's this ability to reduce the scale of people numbers that
- 34:16you have and they'll augment higher value, you know, people
- 34:21with, with the Gentic to support them.
- 34:24And you know, we, we put out some research recently around,
- 34:28you know, the impact of tariffs, for example, that could have a
- 34:32real impact on what we call it could drive the whole services
- 34:39and software adoption curve, right?
- 34:42Because suddenly it's like if it becomes really difficult to
- 34:46manage a disparate global workforce, manufacturing goods
- 34:50and all over the world, you need to bring stuff back home.
- 34:54You know, suddenly, hey, I can actually do what I need in the
- 34:59US with a smaller number of staff.
- 35:02They might be more expensive, but I don't need as many.
- 35:05And they're supported by this technology.
- 35:07So we are at a point where companies are starting to make
- 35:10much more radical assumptions on what they can do.
- 35:15You may have seen a recent announcement from the bank Citi,
- 35:18Citibank, who have decided to reduce their 144 service
- 35:24provider relationships down to 50.
- 35:26And they've actually increased the numbers of staff that they
- 35:29have on shore in the United States and some other regions
- 35:34who are directly within the company.
- 35:36Because they what they want to do is they want to spend less on
- 35:39the legacy and more on the new. So I'm not trying to say
- 35:43companies are just going to fire everybody or replace them with
- 35:45bots, but I think a lot of smart businesses are thinking, how do
- 35:49we stop spending billions of dollars on maintaining legacy
- 35:54applications and legacy systems when we really want to reinvest
- 35:59that money in modernized thinking, modernized agentic
- 36:03technology, that sort of thing. So what some companies are
- 36:06doing, and I use the example of Citibank is they're trying to
- 36:09stop the cost of the old so then they can bring you back in house
- 36:14and then start to think about how do we reinvest in and the
- 36:19technology they need to take them to a different place.
- 36:21So I don't think companies are thinking right now about how do
- 36:23we just get rid of people. They're actually thinking about
- 36:25how do we break from the past. I did a great podcast with Jason
- 36:30Aberbrook, who's one of the leading minds in HR technology.
- 36:35You know, he, he's a Mercer these days and he talks there's
- 36:37like this CHROS across all the big global 50 companies.
- 36:41And he, he actually came out and said these companies have so
- 36:46much data. They, they don't want to do with
- 36:48it. They can't join it up.
- 36:49They can't make decisions on it. It's got to the point where he's
- 36:53got clients who are literally thinking, oh, just just get,
- 36:56let's just trash this old systems and rebuild, rebuild
- 37:01with the new. And I think this is where some
- 37:03of these conversations are happening with a Gentech, which
- 37:05is how do we start to really build out the new and and make a
- 37:09break from from the legacy that's been holding us back for
- 37:11so long. Anthony Scrifignano makes a
- 37:15comment on Twitter directly addressing this point that you
- 37:19were just discussing. He says it's equally likely that
- 37:23the C-Suite is being taken to task for not adopting more to
- 37:29drive down cost. He says the KPIs need to be more
- 37:33than just cost savings. What new problems are being
- 37:37addressed that were unaddressed before being enabled by this
- 37:43technology? And it sounds like you're saying
- 37:46the same thing, that cost savings is a part of it, but
- 37:50there's also a whole set of new opportunities.
- 37:53I would agree that the same fundamental issues have remained
- 37:58for a very long time in terms of changing we, we call it, you
- 38:05know, paying off your debts, your technical debts, your
- 38:08people debt, your process debt, your data debt within companies
- 38:12and, and, and, and this inertia of companies refusing to change.
- 38:15And there's so many managers and leaders within enterprise who,
- 38:19let's be honest, have got away with not having to do much
- 38:22different for the last 20-30 years.
- 38:24I mean, we still have companies operating the processes that
- 38:27were designed before the Second World War, some even the
- 38:30industrial revolution. So what is different this time?
- 38:33I think what's different is the technology is much more
- 38:39pronounced. It's much more ready, it's much
- 38:42more scalable. And there's a final exhaustion
- 38:47where you talk to CIOs off record, they'll all tell you one
- 38:51thing. They are fed up spending 10% a
- 38:54year on their services firms and then 10% increase is every year
- 38:57on their software license hikes. SAS is becoming a legacy
- 39:01paradigm and services people just don't want to keep paying
- 39:04more and more and more. You can't keep going up this
- 39:07exponential cost curve. Eventually the chickens come
- 39:10home to roost and, and, and I think C-Suite executives are
- 39:14really being held to task now. Can you deliver an AI first
- 39:19organization where a culture has to change within the company?
- 39:23And I think that's the problem we've got with a lot of these
- 39:25businesses is they haven't got the right cultures to shift,
- 39:30shift forward and really embrace.
- 39:31And you know, while I would agree, I don't think the
- 39:34fundamental issues have changed all that much.
- 39:37What is changing is the onus on AI that's coming right from the
- 39:41top. Because when RPA came in, in
- 39:442012, the reason why one of the reasons it failed was the CIO.
- 39:48It would get dumped on the CI OS docket and it would eventually
- 39:53get dumbed down two or three layers into the what we call the
- 39:55frozen middle within the organization.
- 39:57And that's when technology solutions go to die.
- 40:01That's not happening so much with Agentic.
- 40:04But Phil, I remember those RPA days and I remember software
- 40:10companies describing RPA just like you're talking about
- 40:14agentic AI right now, which is it's going to save us money.
- 40:19We're not going to need as many employees.
- 40:23This is going to, it's going to be great.
- 40:27But the promise was never fulfilled.
- 40:29So what's different and how do we see our way through the hype?
- 40:34I think what is different is that 15% of high performers and
- 40:39I think the following 15% behind are organizations where the
- 40:44leadership have realized they can no longer keep painting lip
- 40:47service to not fixing their underlying problems with data
- 40:52technology and legacy. And all our PA was really doing
- 40:57was it was like a Band-Aid tech that stitched together old
- 41:01systems to get them functioning more effectively.
- 41:03It was very useful. But in terms of could you use
- 41:06RPA to replace thousands of people unless it was a very high
- 41:10throughput process, very repeatable, very predictable.
- 41:13Of course you couldn't. No, this was like a patchwork
- 41:16technology. You, you know, if you want to,
- 41:18if you want to say, hey, we have 1000 people answering calls in
- 41:24the Philippines for our consumer products that we're selling,
- 41:27right? That's people on mass at scale
- 41:31where you need technology that can actually have some empathy
- 41:35with clients, that can replicate CX behaviour, that can actually
- 41:40do the job. And I think that's the big
- 41:42difference right now is agentic is much, much closer to doing
- 41:45the job of human beings than RPA was, which it really wasn't.
- 41:49It was a patchwork back office break fix technology that was
- 41:53great if you wanted to keep, you know, your old kicks mainframe
- 41:57working with a cobalt system, working with ASAP system, for
- 42:01example. But now it's much more, you
- 42:04know, you can see where this is all shifting.
- 42:07And I think there's a, there's a real exhaustion with companies
- 42:10having to keep maintaining real, you know, creaking old systems
- 42:16in a world where if, you know, competition is much more
- 42:18cutthroat and you got to be really slick and on the ball if
- 42:22you're going to be effective in this economy.
- 42:25Phil, I get the sense that what you're really also saying is
- 42:30that the the difference between RPA and agentic AI is that that
- 42:3715% of early adopters of agentic AI have demonstrated that in
- 42:44fact it really does work. It really can bring these kinds
- 42:48of benefits and savings that you've been describing.
- 42:52Yeah, you can actually create a virtual Co worker to complete
- 42:59end to end processes. It's proven it works.
- 43:02We've all seen the demos, we've worked with companies who are
- 43:05piloting it, we have done it on ourselves.
- 43:07And a lot of enterprises, more advanced ones in particular are
- 43:11at least working with single agents and some move into multi
- 43:14agent models. So they're on the path.
- 43:17And it, it's a different type of technology that removes the need
- 43:22for constant human oversight of complex processes.
- 43:25It's a transformational tool rather than a task automation
- 43:29tool, which RPA was. That was about tasks.
- 43:31This is about human oversight, support and real process
- 43:37capability and the fact that you can build these Co workers.
- 43:42You can, you can engage with these things, you can talk to
- 43:44them, right? I don't even want to get into
- 43:47some disturbing things about teenage boys building
- 43:51relationships with AI girlfriends and things like
- 43:54that. I don't know if you've been
- 43:55reading about some of these things, but this, this stuff is
- 43:58real. People are building
- 44:00relationships with their software.
- 44:02The software, you know, you can ask the question, if you ring up
- 44:05customer service today, do you care that you're dealing with a
- 44:08computer or dealing with a human being?
- 44:10When you're go checking into your airline, do you really want
- 44:14to talk to the gate agent? No, of course you don't.
- 44:17You just want to use your app and get on the plane, you know
- 44:19what I mean? So we're getting to this whole
- 44:22next layer of, you know, technology becoming part of our
- 44:26daily lives much more than ever before, to the point where we're
- 44:29actually engaging with technology in a much more
- 44:32humanistic real way. And I don't know if you saw the
- 44:38CEO of Google Deepmine the other day, He was, I dig this out, I'd
- 44:42saw it on X this morning. But he was saying how the new
- 44:46way to develop code is inviting creators and people with
- 44:52heuristic creative skills to develop the code.
- 44:54Rather than in the old days you were going to like a computer
- 44:58science engineer and having to kind of explain in a very clunky
- 45:00way, this is what we need. We're getting to the point where
- 45:03we can create code and we can create applications without
- 45:07being technically proficient. And one of the things where you
- 45:10still want to talk about the difference between agentic and
- 45:12RPA is agentic is the first time ever really that we can take
- 45:19non-technical C-Suite or leaders within enterprises and have them
- 45:23dictate what they want from their technology.
- 45:25But we are seeing technology that is pivoted towards the
- 45:29business professional and we're already in a situation where I
- 45:32think 46% of IT decisions are made outside of the CI OS walls
- 45:37of their offices. This is the age where the CFO,
- 45:42the head of supply chains, the head of marketing, these people
- 45:45are making their own technology choices because they can start
- 45:48to build technology that is very much answering the needs of the
- 45:53business versus something that you're having to be dictated to
- 45:59by engineers, that sort of thing.
- 46:01It is extraordinary the level of research support that, for
- 46:07example, that these tools can provide.
- 46:11I had a networking issue of my own here in our studio and doing
- 46:18a little bit of research I was able to figure out a fairly
- 46:21complex question having to do with routing.
- 46:24Rather than need to call an IT person and bring a consultant
- 46:27in. It is amazing, but we have
- 46:29we're, we're almost out of time and we have a number of
- 46:32questions that are left. So I'm going to ask you, Phil,
- 46:35to answer these questions pretty quickly, pretty concisely.
- 46:40First one is from Prem Kumar Aparanji and he says when LLMS
- 46:46powering the AI agents aren't reliable or predictable, how do
- 46:51we rely on them to automate unpredictable scenarios that
- 46:56need to work? You have to train the model to
- 47:00work is is my answer. So if there's something wrong
- 47:03with the LLMS, then you need to really have a look at the
- 47:09underlying technology that you're using and find the right
- 47:12LLMS that can deliver the scenarios you want.
- 47:16So I think there's a lot more technology based conversations
- 47:21we got to have to get this, you know, and really enterprise
- 47:25grade ready. And what I would say is, you
- 47:28know, I know from a lot of friends in the industry that
- 47:30like open AI, for example, is very, very obsessed with
- 47:34becoming enterprise ready. Like, you know, you know, the
- 47:38leadership within that company are spending all their time with
- 47:42the C-Suite within Fortune to 500.
- 47:45They're trying to figure it out. So I would say it's great
- 47:48question and there are a lot of faults in the system right now.
- 47:52And a lot of this is honing the models and training the models
- 47:56until you get them working. I mean, as I said, we're doing
- 47:59our own model, we're putting our whole business into into an
- 48:02agentic solution. And it's it's taken us three
- 48:05years to get to a point where we still haven't gone live with the
- 48:07new system yet. But you've got to learn
- 48:10yourself, you've got to learn your business.
- 48:11You've got to learn these models.
- 48:12You've got to try it and try it and try until you know what
- 48:15works and what doesn't. And I remember when shortly
- 48:19after ChatGPT came out, I remember that your company HFS
- 48:24Research was one of the first analyst firms that I was aware
- 48:27of that was making that attempt to put your research online into
- 48:32an LLM. So you, you truly are an early
- 48:36adopter at this. We have a question very quickly
- 48:39now because we're just going to run out of time from Elizabeth
- 48:41Shaw, who says CE OS and boards are driving the use of AI
- 48:47agentic and beyond. There's serious implications for
- 48:50worker and social society impacts.
- 48:54What concerns do CE OS and other senior business leaders have
- 49:00with these concerns? And very quickly, please,
- 49:03please. I know it's a complicated
- 49:04question. Data privacy and cyber are the
- 49:07number one problems and biggest concerns by country mile, to be
- 49:12honest with you. And after that, it's, you know,
- 49:16it, it's, it's other areas around transformation and
- 49:19replacing process and compliance.
- 49:21But cyber is by far and away I think the biggest headache as
- 49:26companies look at shifting to these models is, is maintaining
- 49:29a secure infrastructure. Arslan Com comes back and says
- 49:34who gains the most value from magentic AI, small companies or
- 49:37large companies? I would say at the moment small
- 49:40companies, it allows I, I hate using my own example, but it
- 49:46allows mid sized businesses to really punch above their weight
- 49:49because you can scale fast, you can act nimbly and you often
- 49:54don't have as much legacy within the business to change.
- 49:57You don't have as many people resisting change.
- 49:59And I think large companies can also be truly beneficial in
- 50:03terms of how they can leverage this.
- 50:05But I just found with a lot of large businesses, it's harder
- 50:09for them to beset their legacy. You know, look at the technical
- 50:12debt they have the the lock in they've got with legacy software
- 50:16providers, you know, that sort of thing.
- 50:17So I think it's harder for large companies to change because
- 50:20there's a huge amount of training and, and, and cultural
- 50:25change and shifting that needs to happen.
- 50:27And I think SM ES a better place to pivot and and I see a lot of
- 50:32people I know wanting to go and work for smaller businesses
- 50:34'cause they are more nimble and you got to be nimble in this
- 50:37market. What advice do you have for
- 50:40enterprise, technology and business leaders when it comes
- 50:45to how they should be relating to this agentic AI world today
- 50:52and very quickly, please? Get on top of it, learn it,
- 50:54understand it, experiment with it, do boot camps with it.
- 50:59You've got to educate yourself. The days of BS ING around
- 51:03technology are over. You've got to be much more
- 51:06proficient at knowing what is possible and engaging and
- 51:11building relationships with the whole emerging AI ecosystem
- 51:14around you, you know? And with that, a huge thank you
- 51:18to Phil. First, he's the CEO of HFS
- 51:22Research. Phil, thank you so much for
- 51:24being here. I'm just so grateful to you.
- 51:26Yeah, a pleasure, Pleasure, Michael.
- 51:29Now I went quickly, enjoyed it very much and I look forward to
- 51:33more interactions. And a huge thank you to
- 51:35everybody who is watching today. You guys are an awesome
- 51:39audience. You're so intelligent, so smart.
- 51:42Go to cxotalk.com, subscribe to our newsletter and join us.
- 51:46Join our community and join our live shows.
- 51:50We have one next week and the week after that, so check it
- 51:53out. Thanks so much everybody, and I
- 51:55hope you have a great day and we'll see you next time.