The Real AI Opportunity Is Hiding Inside Your Business

AI is advancing at an extraordinary pace and for most businesses the challenge is no longer access to the technology, it’s knowing how to apply it. In this episode, we’re joined by Mark Rotherham (CTO at BCN)  to explore the latest news and developments in AI and why the focus is shifting from model capabilities to business execution.

We discuss the race between frontier AI labs, the short-lived release of Fable, and how platforms like Microsoft Copilot are evolving from assistants into systems that can take action on behalf of users. Mark explains why organisations should be spending less time chasing the newest models and more time identifying where AI can deliver measurable business value today.

The conversation then turns to agentic AI and how businesses can begin redesigning workflows around outcomes rather than tasks. From sales and research to content creation, we explore where early adopters are already gaining an advantage and why AI decisions are increasingly becoming strategic, board-level conversations.

Subscribe for more AI for business insights, share this with a leaders who are shaping their organisation’s AI strategy, and leave a review with your biggest question about adopting AI successfully.

Thanks for listening!

Connect with us on LinkedIn or visit our website.

Thanks for listening! 

Connect with us on LinkedIn or visit our website.

Transcript

Cold Open: The AI Adoption Divide

SPEAKER_01 0:01

There's two types of customers at the moment. Some get it. The huge and significant risk for those that don't is that the longer it takes you to get on the train, the harder it's gonna be.

SPEAKER_00 0:12

SpaceX, Elon Musk, a huge amount of resource behind them. Doesn't that give them a huge advantage moving forward?

SPEAKER_01 0:19

From a financial and a resource perspective, absolutely. They have got more money now than they could imagine.

Why Execution Matters More Now

SPEAKER_00 0:29

Welcome to AI for Business, BCN's monthly briefing for business leaders who want to stay ahead of the curve without getting lost in the noise. The biggest shift in AI right now is that the technology is no longer the constraint, as we learned a couple of episodes ago. It's all about execution. That's the potential obstacle. Models are improving at pace, but the real question for leadership teams in businesses is how do these new extraordinary capabilities translate into meaningful change across cost, productivity, and competitive advantage? So, in this episode, we're going to break down exactly what's changed in AI and what those changes mean for your business and what practical actions leaders can take to stay ahead of the curve. So, here to do all that is Mark Rotherham, Chief Technology Officer at BCN, an all-round AI for business guru. Hello, Mark.

SPEAKER_01 1:20

Hello, Ed. Thank you for having me on again.

SPEAKER_00 1:22

No, thanks very much for joining us. So, Mark, before we get into the nitty-gritty of what businesses should be doing to stay ahead in terms of those AI trends, talk me through some of the big news stories in AI this month.

SPEAKER_01 1:35

Yeah, so it's been a very interesting month as it always is with AI. Yeah. There's probably three or four key moments that have defined it so far. I say so far because things appear to be happening almost daily at the moment.

Anthropic Fable And The Shutdown

SPEAKER_01 1:53

So if we go with you know one of the biggest AI releases that's happened ever, uh, with Fable coming out from Anthropic a few weeks ago. Um, we saw probably the most intelligent model available to the public come out on a Thursday. It's based on the the Mythos model, which they wouldn't release because they said it was too powerful and it was too good at breaking into systems. Yes. Um and it was great. So Fable came out and it was like ha going from having a really intelligent person working with you to Fable, which was almost like having a super intelligent team that would just crack on and do anything you wanted with a much greater level of intelligence. Within 48 hours, though, of it coming out and us all having fun seeing what it could do, the US government pulled the plug and decided that it was too powerful. They'd been made aware that um the security that Anthropic had put into Fable, which was to stop it from doing dangerous things like talking about creating bioweapons or creating uh cyber attacks, those guardrails could be bypassed. So effectively it was deemed too powerful to let out into the public domain. They told Anthropic they would only allow it to be used by uh Native uh American citizens. And Anthropic had no way of doing that, so they turned it off and it's been off ever since.

SPEAKER_00 3:14

Was there pushback from Anthropic on that? Did they did they want to keep Fable in the public domain?

SPEAKER_01 3:19

Yeah, they they they had um really good argument that said, well, the the jailbreak, as it was known, applied to other vendors and other models as well. So OpenAI with GPT 5.5 had the same kind of rules and the same ability to be jailbroken. But I think the the point that the US government were making was, but you've released something different, it's more intelligent, it's more powerful. Jailbreaking this model was significantly more dangerous. So um I think it will come back, but it'll probably come back with you need to have registered and they need to know exactly who you are before you can access it and uh hopefully um put some better guardrails in.

SPEAKER_00 3:56

Presumably there's an argument that people that are anxious about the ever-evolving capacity of AI, that there are some guardrails in place. That's slightly reassuring, certainly from my perspective.

SPEAKER_01 4:07

Yeah, it it it is, and you know, with great power, they they need to have some responsibility. Um next, I guess, significant thing is with another frontier lab.

SpaceX, XAI, Cursor, And Compute Power

SPEAKER_01 4:19

So if we wind back time a little bit, because there's been a lot going on here, we have XAI with Grok, which was another frontier model. XAI and SpaceX merged, creating I think it was the fifth or sixth biggest company in the world for a while. Now, a few months earlier, SpaceX had done a deal with Cursor, and Cursor is one of the biggest tools that leverages all the different models to actually write software. So it's a it's a very, very popular, one of the best in-class tools. It's model agnostic, frontier lab agnostic, but is used globally to write software. They've done a deal to work with Cursor to improve their models and to potentially buy them. And the dust had barely settled on SpaceX and XAI joining through the IPO when they announced actually they'd done the deal and bought Cursor. A lot of really interesting things in there. Cursor do have their own models which are now being powered by all the compute that SpaceX have got. SpaceX have got more distributed compute than many other hyperscalers at the moment. Part of the news over the last month was that Anthropic are paying an awful lot of money to SpaceX to access that compute. Google are also paying SpaceX to access that compute. And this is the big shift from um SpaceX to actually look at Grok and say, well, actually, maybe it's not uh competing where they wanted it to. Bring in Cursor and look at the models they've got and say, right, we're gonna double down on this, we're gonna buy the platform that is used by most of the top Fortune 100 companies globally that's got access to all the other models but has their own models, and we're gonna power it with our computes and really start to think about how they can get back into the game and start competing with Anthropic and OpenAI.

SPEAKER_00 6:10

Two questions. First up, presumably, SpaceX, Elon Musk, huge amount of resource behind them. Doesn't that give them a huge advantage moving forward?

SPEAKER_01 6:20

From a financial and a resource perspective, absolutely, they they they have got more money now than they could imagine, you know, with the IPO. But the number one thing that they've not got is the top talent.

SPEAKER_00 6:33

So, second question on that. You've mentioned the term frontier labs uh a few times. Can you define that for us just generally?

SPEAKER_01 6:40

The best way to describe it is these are the guys that are driving the best global models and being ahead of the curve. So Frontier means they're they're the early adopters, they are the guys pushing ahead. We've then got the people that are strong followers that learn from those. Okay. And they're releasing models that that might be nearly as good but a lot cheaper. Um we've got a handful of frontier labs and then an awful lot of smaller labs that are kind of coming up from that like copycats, but kind of doing it their own way.

SPEAKER_00 7:14

So we we've talked about Anthropic, OpenAI, SpaceX there. What about Microsoft? What what are they doing?

SPEAKER_01 7:20

So Microsoft have been really interested in the AI space. They started their AI bust with Copilot, which was awesome, Copilot 365. That led the way. You know, they got some really good traction there. And they and to do that, they partnered with OpenAI. Yeah. Then most recently, you know, we go back six months, they partnered with Anthropic to bring Anthropic through as one of the models in in Copilot. But despite that, they were being disrupted by what Anthropic were doing with Claude. Uh, and we saw a lot of our customers actually jump on the Claude Bangwap because it was demonstrated more mature than what you could get out of Copilot at the time. So since then, Microsoft have continued that partnership with both OpenAI and Anthropic. So we're getting the best models through and available from Microsoft, but they've also um started upping their game with their tooling. And I think this is where things start to get interesting.

Microsoft Cowork And Action-Taking AI

SPEAKER_01 8:13

It's not necessarily all about the model anymore, it's about how you use it. And what Microsoft have done is release in the last few days Cowork. This is Microsoft's frontier capability, it's kind of standing on the shoulders of Copilot 365. But effectively, Cowork leverages those frontier models, and you can give it tasks to go do. So one of the first fun things you can do is get it to go and sort out your uh inbox. Now, Copilot had some features where it could go and look at your emails and tell you what was going on. Cowork does actions. It went into my inbox, it decided what was junk, what I needed to look at, and it got rid of tons of emails. It just plowed through the last seven days, got rid of everything that was spam junk. So it started reading my emails, and a customer had emailed me saying, Hey Mark, we need to have a quick pre-meet for the workshop we're having on Thursday. Can we meet on um Wednesday morning? And it went and checked my calendar. It saw that I was already busy on Wednesday morning, and it wrote an email back to the customer saying, I'm sorry, I'm really busy on uh Wednesday morning. Could we do it Wednesday afternoon? And it saved that as a draft for me where I could just go in, read it, and click send.

SPEAKER_00 9:23

Yeah.

SPEAKER_01 9:23

So it it worked through and thought about the challenge that had been put in that email. Again, it's something that is very powerful, but also within the ecosystem of Microsoft, and that's where they're playing now.

SPEAKER_00 9:37

They're looking at the ecosystem play. Um, so that I feel like we've covered the biggest news stories in AI over the last month. Let's make it a bit more about general shifts and how they might be applied in business.

Building Agentic Workflows At The Edge

SPEAKER_00 9:49

What are the biggest shifts you're seeing in AI right now, and I guess about how it's applied?

SPEAKER_01 9:55

Yeah, and it's it's probably an ongoing theme. Um, we are seeing an incessant need to do something. Yeah. And the do something is becoming easier as we go through and get more maturity. Okay. The majority of our customers have already started on their journey with end-user um individual-powered AI, so a co-pilot license or adoption of the free version of co-pilot. So most of our customers are and the the industry is starting to get to grips with that. Now, not everyone's fully there yet, so there's still quite a lot of work to do just to get AI native, but where the real value is and where we're seeing that the customers that that are kind of realizing the opportunity is starting to build out the edge version of their business that is agentic through and through. And what that looks like is choosing one or two workflows and starting to think about right, how do we put these into an agentic way of delivering them? So it's really important that they don't just copy what they do today, it's thinking about the outcome and how do we get to that outcome the most efficiently. And it's starting to build out almost the 2.0 version of the business from an agentic point of view, with a view to that's the direction of travel. So we we've been um customer zero for a bunch of this within BCN, and what we're starting to see now is that demand and that recognition that actually companies need to edge towards what does the agentic version look like and how do they do it? So the questions we're getting asked is to how do we go from an individual having a co-pilot or an AI assistant that helps them do their job through to reimagining the outcome and getting an agentic through and through approach? How do we do that in a safe, secure, scalable way that allows us to over time put more and more of the business process into that new world and shrink the old one? So it's it's really interesting. That that's that's something that will add significant value and savings and opportunity to the businesses that are able to do that. You know, the first movers doing that will reap a lot of rewards. For a long time we said, wait until AI is ready to do stuff, it's been ready to do this for a while now. And now it's a case of right, how do we execute on that? How do we leverage these really clever AI models in a way that helps us accelerate the business? So that's for me the most exciting thing at the moment is how do we bring that kind of concept of it's on the edge, but it's going to grow, and over time it's gonna kind of absorb more and more to the point where it is the 2.0 version of an organization or a business that is agentic first.

Best Starting Points For Leaders

SPEAKER_00 12:55

Okay, so on that, say I'm a business leader, I want to dip my toe into this new world of agentic AI. What areas of the business are the most common or perhaps the easiest areas to try out agentic AI?

SPEAKER_01 13:11

If I'm honest, it's not what we'd expected because the simple things are cross-functional. Yeah. And for a long time we've done things like mailbox management where you've got a shared mailbox and we can wrap AI over the top of that. And that tends to help, but it doesn't really transform the process, it just kind of solves challenges there. Where we're seeing um a lot of the opportunity at the moment is where we've got teams of people that are spending an awful lot of time not doing the the work that brings in value to the business. Yeah. So the best example I'll give is sales teams. So we've got an awful lot of sales teams across our customer base that spend more time not selling than selling. And one of the most interesting things we've we can help with there is not necessarily making all of the stuff they're doing that's not selling go away quickly, because that quite frankly can be a mess.

SPEAKER_00 14:11

Yeah.

SPEAKER_01 14:12

But what we can do is use AI to make them more powerful and more efficient sellers to give them better assistance, better research, better output, better qualification and than they've ever had before. So we're seeing it come out in quite interesting ways because the obvious thing would be, well, let's go and sort out the things that's wasting their time. And what we find is those are really valid, but they're quite hard. There's reasons why things are quite hard there. Whereas when you look at a value chain and you say, Well, actually, what can we do to that team to add value? Can we take away some of the stuff they don't like doing, or can we make them super powerful for the stuff that we really want them to do? And that that seems to be where the AI can really help power. And that's what we're seeing from our initial customers that are in this kind of sales consultant type of role. That it doesn't take away the stuff they don't want to do, but it really powers them to the stuff they do want to do and gets more bang for the book, uh levels them up, and gives them more capacity back in that direction. The value that we are bringing at the moment is in these kind of making a team better at what they are designed to do and being very focused on that. Um, so sales is one example, um, but there's others like researching and creating content, um, taking lots of manual creation work away and automating that.

SPEAKER_00 15:36

Have you seen a change, Mark, in how businesses are getting themselves into the AI world, how they're buying AI?

Buying AI Platforms And Picking Models

SPEAKER_01 15:45

I think the word I'll use there is emerging. So there's some stuff that's really straightforward. So buying a license for an end user has been relatively straightforward. We've seen a lot of people buy Claude licenses that would have traditionally just bought a 365 copilot license. And I I think that will shift from one to the other as we see the tooling that controls the models be the most important thing as we go forward rather than the intelligence of the models.

SPEAKER_00 16:15

Yeah.

SPEAKER_01 16:16

So that that's almost straightforward. Historically, a lot of the things that we've sold have been quite point in time because the intelligence was fairly straightforward. As we go into this new emerging kind of agentic workflow world, it's almost buying into a new platform. And you know, we we have lots of interesting conversations about customers around what is it, yeah, because they're very used to buying a Salesforce or another tool. And when we talk to them about well, we're buying into a platform, what would you like it to do? It's a different conversation. Yeah, when they see what it can do, they get it. So it's it is really interesting, if I'm honest. Um, we're trying to pave the way of what that looks like with um being customer zero of our own agency workflow platform and our early adopters that that we're kind of taking on that journey, learning what it looks like to make it work fully commercially. But it does look like a bit of a journey, you know. If there's elements of we've got to discover, we've got to experiment, uh, and then we've got to iterate and we iterate in a safe, secure, controlled place where the data is in a good um safe position, and we're we're showing value until we we've got to a point where right, we've shown enough value there, let's move on to the next thing. I I can see it potentially turning into subscription where you subscribe to value over time and you subscribe to outcomes, but that's you're going from a 30 pound co-pilot license to subscribing to value. It's quite an interesting gulf of of kind of um commerciality there.

SPEAKER_00 18:01

Yeah, I and I guess I mean it's that's just reflective of the extraordinary rate of technological advancement over the last few years that continues to get faster and faster. And presumably there's some, and I think you've touched on this, but there's some sort of reticence in businesses to dive in because of the fact that the technology is going to change again rapidly. So maybe that subscription model that you mentioned covers that off.

SPEAKER_01 18:23

Yeah, exactly. And and it's moving at such a fast pace from a frontier model perspective, like with the example of Fable coming and going, and it'll come again, you know. That that model is insane, it's awesome. But if I'm honest, the models that we use are the ones that came out in January.

SPEAKER_00 18:42

Yeah.

SPEAKER_01 18:43

And they're perfect for all of our agenda workflows. The models we're putting into our first customers are the models that came out around January, February. Right. Yeah. So the capability of the models is there, and people do get distracted by, you know, I get asked by customers, do we need to go to this? What can we put fable in? Will it make it better? And I'm like, well, actually, it'll cost you more, but you don't need the level of intelligence that's coming out now to do the business workflows, uh, the agency workflows that we're targeting. So that argument is is still there, but when you explain it and say, actually, you don't need a frontier model. We we use a lot of haiku sonnet and opus 4.6, you know, they're the anthropic uh models. Um Opus 4.8 is out. Um we don't use that in any of our agentic workflows currently because 4.6 is just great. And in fact, we use uh Sonnet for quite a lot, which costs a fraction of the cost of Opus because it's the right model to choose for the job. And a lot of what we're doing is bringing the right model to the job, and more often than not, the right model has existed for a good six months. Therefore, people that are waiting for Fable and the GPT 5.678910, they'll be waiting for everybody because they will keep coming and they will keep getting better. But you don't need that to take advantage of AI right now. And the our early adopters are seeing that, they're witnessing it, and they're getting great results based on just using the models that have been around for a while and do it using them well.

AI Moves From IT To The Board

SPEAKER_00 20:11

Are AI decisions technical or are they becoming strategic? And I guess that question is linked to the shift from AI tools to AI agents. There's some sort of connective tissue there.

SPEAKER_01 20:24

Yeah. Um so the answer is it it's becoming less and less about technical from a need and a decision maker perspective. And great examples of that with one of our customers, the decision to bring in a launch pad and a bunch of sales-oriented use cases was made outside of the technical teams. It was made by the people that wanted it and needed it. It was made by the board and the sales leadership, right? The technical teams that are now charged with delivering some of that, they're almost playing catch-up to the technology that is now being exposed to them. What we're seeing is the conversation we have has radically changed. We used to have the conversation with the IT function. We talked to the the CTO or the head of IT or the people responsible for technology. And that's been historic. You know, MSPs have loved that relationship because they're the people that hold the pus strings for the servers, the desktops, the office licenses. But that's not where the value of technology is anymore. The value is across the whole business, and it's looking at it from a business lens of where will the capability of AI most benefit the business. And that doesn't come from a technology perspective. Most of the time, it comes from a business perspective, and it's looking down and saying, Look, my sales team, I need Them to sell more. My product team, I need them to be much more efficient at creating product. It's looking at it from that perspective. And our conversations are with the whole board rather than with just that technical silo now because of that, because that's really where the outcome-driven approach needs to sit. And to be honest, when you position AI, it should be a strategic direction. But let's try to get the low-hanging fruit in, because that gets the confidence in. And that low-hanging fruit could be anywhere. But generally, the board will be able to point at somewhere that says, look, this is the thing that if we can just nail, is going to get the ball rolling. Let's nail that, then let's nail the next one. And it should be like that. Even when we do a program of AI initiative, there's still a roadmap and a journey where we start with quick wins and that are going to set the direction and get the buy-in before we get on to things that are more complex and maybe higher value but take longer to kind of come through. So yeah, it's definitely shifted, and the shift has been really good, really powerful, and needed because it's got to be done at that business level.

SPEAKER_00 23:09

Yeah, I guess the technology is a given, isn't it, now in mid-2026. So, final question, and it's really a summary question. Can I get you to summarise that directional shift that you've kind of covered over the last 20 minutes for business leaders and what they need to understand?

SPEAKER_01 23:24

Yeah, so the best way to say it is don't get super excited about the next best model. Look at your business, what it does, and where the best application of an agentic approach would actually add value, and think about how that can be formed as a future version starting point of how the business operates. Got to think about giving a pathway to something different. Yeah, because this is something different. This is, you know, one of the biggest waves of technology change we will ever experience. The internet was quite a big wave. This is quite a big wave. So uh we've got to think about the strategy and how we get on that wave to get the most value out of it. So it does need that take a step back. This is different, it's here, the value is there, available today. How am I going to get on that journey and what does it look like? And what we're trying to do is facilitate a really efficient way of getting on that journey in a safe, secure way that will scale out and be what it needs to be for that business. We are demonstrating that ourselves. We're very focused on being customer zero. Great fun being customer zero. We learn a lot, but we also get a lot of value. So we want to bring all of that through into our customer base and really help them think about what their future business looks like with the agentic capabilities, maybe starting at the edge, but eventually becoming as big as a and as bold as it should be to transform them and to drive

Fear, Disruption Risk, And Next Steps

SPEAKER_01 25:02

them forward.

SPEAKER_00 25:02

So, and I have a lot of sympathy for people who might be in this position. What if people are sort of frozen by the fear of getting it wrong? Does that make sense? The idea that you're making this big shift, the uh consequences are gonna be significant, you don't want to get it wrong, and therefore you don't do anything. What about those people?

SPEAKER_01 25:21

Yeah, um, there's two types of customers at the moment. Some get it and they're they're working with us and working towards it. The the huge and significant risk for those that don't is that the longer it takes you to get on the train, the harder it's gonna be. What I mean by that is it takes a while to get your head around and to start adopting and start thinking about how to use AI. There's a massive risk of customers becoming overtaken by either their competition to embrace the sooner, or by that startup that might not have your heritage or your history, but can do the job at a fraction of the price, twice as fast with higher quality because they've been AI first. Uh, we talk an awful lot to customers about their moat, and um that moat shifts all the time. Um, data is emote, customer base is emote, but there will be people trying to do what you do using AI. The more they grow, the more they get stature, the more they they get their data, they will win. You know, they'll get customer based. And if you just look at some of the things that happen across the world, this isn't fiction, this is happening. The disruption, a lot of the disruption, if I'm fair, you can point at anthropic and clawed, but that's the thing to keep an eye on. Looking at how you cannot be disrupted, get on the bandwagon before you are overtaken by someone else that can just do it quicker, better, faster, stronger.

SPEAKER_00 26:44

Mark, that's brilliant. So much to think about. Thank you very much indeed. Thank you.

Closing And How To Reach Us

SPEAKER_00 26:48

Okay, that's it for this episode of AI for Business. We've covered the latest shifts in AI, the huge wave of change that we're seeing right now, and why this is moving from a technology conversation into the boardroom as a core business decision. Most importantly, we've shared some practical ways to help you get moving. Hopefully, you found it useful. Do get in touch if you have any questions. Thank you for listening, and we'll see you next time.