AI for Business with BCN

AI That Delivers for Business: Building your Agentic Workforce

• BCN

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 27:14

In this episode we look at why 2026 becomes the tipping point for AI in information work and how leaders can turn capability into measurable outcomes. We break down  AI agents, governance, data readiness, and practical first steps that prove value without risking your most critical assets.

• Defining information workers and where disruption lands first
• Closing the knowledge gap with safe, governed tools
• Shifting from tasks to outcome-oriented orchestration
• Agents as digital colleagues with identity, policy, and skills
• Scaling benefits: cost, consistency, and 24/7 coverage
• First steps for SMBs using Microsoft Copilot and policy controls
• Lower barriers to proof-of-value and rapid prototyping
• Governance, compliance, and model provenance
• Data maturity aligned to decisions and outcomes
• Measuring ROI with partial automation and human oversight

If you're enjoying our content, make sure you hit like and subscribe so you don't miss an episode.

For more information about BCN, visit bcn.co.uk

Thanks for listening!

Connect with us on LinkedIn or visit our website.

Sinéad Hammond:

Welcome to the BCM Podcast, keeping you up to date with Microsoft IT trends, AI innovation, and business technology. Always built around people and driven by expertise. We share practical insights and real stories to help you get the most from your tech so you can work smarter and achieve more. In this special episode, we're talking about the key technology trends for 2026. This episode will focus on what leaders should be doing differently for the new year and which risks and opportunities deserve your attention right now. I'm your host, Sinead Hammond, and I'm delighted to welcome Mart Rotherham and Fraser Deer to the show. Thank you both for joining us. Can you just give us a little bit of an introduction as to who you are and what it is you do here?

Mark Rotheram:

Yeah, good morning. I'm BCN's chief technology officer, so I look after all things technology, how we bring innovation, both to ourselves internally, but more importantly to our customers, specifically focusing on what we can do with AI and what's happening and what to watch out for.

Fraser Dear:

My name is Fraser Deer. I head up uh BCN's approach to AI and data innovation. Much the same as Mark, my role can kind of tends to work alongside with our clients, unpicking outcomes to the activities that we would need to do internally to yield that outcome. And it's it's not quite the same as it used to be in this age of AI. So looking forward to unpicking that in this session. So Mark, um, why don't we start with the first question, which is you know, every year we've heard about AI being this is the year of AI. Why is 2026 different? What is it about 2026 that we think will be a turning point?

Mark Rotheram:

So I think the main change we're going to see is the capability that AI brings to us getting to a tipping point where it can finally truly disrupt the information workforce. So anyone that's an information worker currently is better than an AI doing their job. What we've seen over the last four weeks, never mind, you know, before that, is such a rapid improvement in the AIs that have been launched, ranging from Google with Gemini, Anthropic with Opus, um, and then in the most recently with GPT 5.2, the capability is increasing week on week at such a pace that we're basically getting to a point where that information worker challenge that the AI has struggled to capture up until now is overtaking what we can do as people. We're now seeing a lot of the traditional work able to be done with models right now ten times quicker at 1% of the cost. So we've got real disruption happening at an incredible pace. As we go into 2026, this is just gonna continue. The different providers of all of these artificial intelligence models are in a competition to get to the first AGI. You know, they're all looking to get to this point where they can proclaim we we've got a general intelligence that can just do stuff. And what we're seeing is this this rapid increase in capability week on week that really is groundbreaking. So I think you know the the technology advances that have happened in 2025 are going to continue, and we're only just now seeing the potential disruption to that that information workforce being squarely in the targets of um AI.

Fraser Dear:

And so um how would you define information worker as as opposed to somebody that's not an information worker? And why is that important for organizations?

Mark Rotheram:

So I think the the really easy way to define it is anyone that spends most of their time working on a computer. So people that work on computers over the next few years will see that being disrupted more and more. The manual uh workforce or people that do something more physical, they've got longer before the disruption comes, but again, the disruption wave is on its way there with all the advancements that AI are allowing robotics to have. So I think think about it from a if you are a digital worker, someone that works predominantly on a computer, a laptop, uh a Mac, or a tablet, those kind of things that you're doing, the AI is now starting to be able to do faster, cheaper, quicker, higher quality. And it's just a case of adoption now. The capability's there, it's just the timing of when it will be executed.

Fraser Dear:

So if um if it's here today and we think about organisations and businesses tomorrow, well what do you think the biggest gap is between having the capability and actually making it have an impact?

Mark Rotheram:

The main issue that we see across all of our customers and and is pretty much replicated in industry is knowledge. So ironically, AI is here to help us solve knowledge, but uh the biggest problem we've got is lack of knowledge of what it can do, how it can do it, and how it can impact a business. The main thing that companies and businesses can do is invest in the right levelling up to understand what AI can do and what it's going to look like for your workforce in the future. We're going from the uh world of doing the discrete tasks, so going into a spreadsheet and doing a thing, going into a PowerPoint and doing that, going into a CRM or a business management system and doing a thing, to describing what you want to have happened and letting the AI go off and do that for you. So there's a big mind shift and skill shift that's needed in order to take advantage of AI and to understand it. And I think the the biggest challenge most businesses are gonna have is that understanding of what it is, and then secondly is what's important to them, how how are they gonna leverage it and get their workforce in a position to use it. I suspect the the biggest challenge will be the lack of knowledge and lack of skilled workers that can actually influence and interact with the new systems of the future in a way that's that's gonna really make them come to life.

Fraser Dear:

And so, you know, let's pretend for a second that I'm a small to medium-sized business here in the UK. How do I start on this journey? How do I start enabling ourselves as an organization to be ready for AI?

Mark Rotheram:

Um we need to invest in adoption management. So it's all gonna be around how do we take the right steps on the journey to get your workforce more AI native, and that there's plenty of ways of doing that, but the easiest is to use safe and secure tools that are available to you today and get the adoption of those propagating around your workforce. It's not an overnight thing, it's a steady, let's start educating, let's get people using these tools and getting familiar with what they are and how they behave, and then it's starting to look at how we can map the challenges that a business has into AI delivered outcomes. What I mean by that is it is starting to find the low-hanging fruit where AI and the current capabilities, you know, let's not talk about future too much for a moment, the current capabilities can make a massive impact. They can take an existing process or an outcome that we want from an existing process and really transform that into something meaningful, impactful, but also demonstrating a return on investment. And I think a mix of getting the broad adoption of AI skills into the SMB and the SMC kind of businesses, and then demonstrating its value, are the are the kind of two things that we should be seeing every one of the businesses across the UK look at. Because if they don't do that, then they will be left behind on this kind of AI journey that will disrupt pretty much every business where they do have some form of information workforce.

Fraser Dear:

Well, uh let's move on a little bit and start to think about this concept of agents, because you know, you and I had a really interesting conversation last week about what is the definition of an agent and this concept of a genetic workforce. You know, how are we defining digital colleagues if they're going to be engaging with a human or not autonomous agents? These are big buzzwords, you know, break it down for me.

Mark Rotheram:

Yeah, and I think you're right. That there is so many fun words out there, generative AI, large language models, agents, a genetic workforce, that to kind of make it real, we should put it into kind of a human context. When we're talking about an agent, we're talking about a thing that has got an element of identity, so we can point at it, we can give it a policy, we can secure it, we can train it. But more importantly, we can give it a skill, we can give it the ability to do some actions. So effectively what we're talking about is taking an element of a role and we're supercharging it by putting it in this kind of construct of an agent, and we're giving it that ability to be trained, to have governance and control. You can't do this, you can't do that. We're securing it, we are giving it the ability to make decisions and to perform actions. It's very much the same thing we do to anybody that joins a business. We we give them training, we onboard them, we give them accounts, we secure them, we give them training, we do all these things, and then we start letting them do activities. And the way that we see the kind of agentic workforce of the future is very much like that, with the main difference being the intelligence that's powering it, and the fact that we're going from AIs, if we go back a year or two ago, that you know we would talk about them in the terms of the the maybe graduate level or secondary school level. We're now able to get AIs and agents that are the best in class, the best in the industry, people that you would not be able to hire doing certain jobs. So we've got this combination of super intelligent or hyper-intelligent agents that are being treated like this digital employee. You need you need to keep an eye on, you need to continuously train and make sure it's governed and make sure that the actions it's doing are correct. But leveraging the fact that for most of our clients and the businesses around the UK, they would probably never have been able to hire that talent because that talent doesn't exist for the the price that you would pay a real person to do it. It's just not obtainable. So yeah, I think it's visualising it in that construct as almost an entity that we're managing, we're bringing in, we're training, we're we're looking after, but leveraging the fact that it's very cheap and can be hyper-intelligent and can be quite closed from a set of actions we allowed it to do, with a view of growing that over time. Okay.

Fraser Dear:

So if we think about these new you know, if we call them the digital colleagues or these agents, it sounds like what you're saying is that leaders kind of really do need to ensure that we treat governance and security and training just the same as we do with any other member of staff. But if that's the case, then what is the business impact? What what is the value that the businesses are going to get from using these tools?

Mark Rotheram:

There's lots of different ways of talking about value. Um when we look at outcomes in ROI, it can be time, it can be quality, it can be cost, it can be all of those things and others as well. But if we think about just pure scale, yeah, let's use scale as a really good example. Today we see lots of human-centric processes, which are fine, you know, there's nothing specifically wrong with them. But in order to scale them you need more people, and more people does scale, but it introduces lots more cost. The benefit of providing an agent that has been trained, that has got the right actions and the right capabilities and governance and training, is that instantly you've got a twenty-four by seven autonomous capability that scales as you need it. You've reduced the cost dramatically from your scaling perspective, and you've increased things that are typically hard to get from a consistency perspective out of a human team. Now we're not saying everything is going to go there straight away, but there are ripe opportunities where there is boring, labour-intensive work that effectively is highly repeatable, where we've got people that spend a lot of time doing that that work, and you know, on a Monday morning they might come in and be a bit grumpy and not do as much as you would like. On a Friday afternoon, they might leave a bit early or think they've done for the day. Whereas you bring an AI in, and as long as you're maintaining it, you're gonna have that quality, you're gonna have that speech, you're gonna have the cost control, you're gonna have that governance, you're gonna have the repeatability, and all of a sudden you're scaling through technology and it's literally volume-based, rather than having to worry about where your next hires are gonna come from for a a job that may be boring and something that people don't necessarily want to do. And the real goal here is we've seen many times over history, is to free those people up to do something more interesting as you scale. We we see the AI impact starting at the dull and boring work and then maturing over time, but as businesses scale, that should free up those people from doing that dull and boring work and allow them to get into more higher value uh work that the AI hasn't got to yet because of all the things you talk about, training, governance, skills, you know, it will evolve and continue, but that's that's kind of how we see it playing out.

Fraser Dear:

Nice. So if you think about how an organization would start to take that journey, because many organizations have kind of had a little dip in the water, they've kind of maybe using kind of web-based AI tools and components. What is the kind of the very first step that an organization should take?

Mark Rotheram:

So the very first step is an interesting one. So we recommend all of our clients start with a governed, secure, free AI capability and expose that to their workforce with the right level of adoption. So we're talking technology, it's things like Microsoft Copilot Chat, where if you're in the Microsoft ecosystem, it's there, it's free, but it's secured into your tenant. So you there's no data leakage if you put data in there. So it's a really nice safe place to start. And from that point, we start to see really interesting things. We can trend the consumption and the usage of that, we can see which are the potential power users in AI going forward, we can start to see which teams are making advantage of it and which ones aren't. And you can focus your attention because really this first stage is just about getting familiar with the language of AI and the capabilities. Now we dovetail that with the security posture. So it's all great, well and good saying, use this one, but we all know that there are hundreds, if not thousands, of different ways of accessing AI for free. So having a policy that allows you to govern what data and what AIs to use is really important. So bringing in a capability that allows us to see, you know, who's using Deep Seek, who's using uh Grok, who's using uh Claude, all the different things out there. And then having, you know, backing up with a bit of policy that says, well, actually, we're happy with that in these ways, but we're not happy with it in these ways, and being able to manage that. So the first stages is really adoption, it's enablement, but it's also some governance and control to make sure that you're not using the wrong AI and you're not having data leakage or anything negative that could happen, and it probably is happening today.

Fraser Dear:

So, Mark, you've just mentioned a number of the different kind of AI providers. And you know, I can remember a couple of years ago thinking, oh yeah, we want to do this really smart AI project, and it was hundreds of thousands of pounds before you even got started. You know, with things like Copilot inside all of your business applications and custom applications being built in you know days, I can see that there's a quite a big shift happening now between that kind of AI enterprise program of work to making really big impacts with a very small team of business users, transforming that small team into something that that's unrecognizable to what it was a couple years ago. How does that happen?

Mark Rotheram:

Yeah, so I mean th the first bit is recognizing that AI can have an impact, and that's all around the education piece. But you're absolutely right, the the barrier to entry for transformative capability has never been lower. Uh we're at the point where prototyping and coming out with proof of value, you know, something that just demonstrates the art of the possible can be done in hours, if not days. You know, we we can get to that point of demonstrating AI having an impact very quickly. The trick now is really understanding what the key outcomes are and what's needed to get there. So many businesses run on fragmented, old, out of shape processes that aren't well documented. So in a lot of ways the barrier is actually understanding what and how we're going to execute on the thing, not building the thing that we're going to do. So the barrier is lower. I think it's business understanding and logic that needs to be kind of gathered to frame the right potential outputs. Now we can do things really quickly, so it doesn't need to be perfect first time. But I think it's now joining the art of the possible that AI's given us with the right approach of, you know, we we call it rapid prototyping. Having a go at things has never been easier. So with the right construct, we can demonstrate that shift extremely quickly. And that that's I think what we're we're seeing an increase of are clients that are aware of AI, they've seen it, they've they've had a play around with it, and now starting to think about how it would impact their business. And they're the early adopters of this new wave, you know, we've been doing this for many years, like you say, Fraser, but over the last few years it's been quite a hurdle. You know, we're talking quite an investment to do a full end-to-end business process, and it takes time. With AI, we can accelerate that, which means we can accelerate the time to value and do more, you know. And I think what we're seeing is the ability to do more quicker and have a bigger impact is pretty much where we're at right now. Yeah, it's it's a really important point that that barrier that stopped most SMBs and to a degree SMCs from doing transformation over the years has dropped. And now it's a case of having the right conversations with the right people about the right topics and then turning that into awesome outcomes very quickly, and then iterating. You know, we don't need to build a hundred percent of a thing in one go. We can build twenty percent and then iterate going forward. And what we're finding as we do this is that we can actually change the business process for the better, because we can start to think about the outcome rather than trying to automate every bit of an existing process.

Fraser Dear:

So it sounds like we need AI enabled people, or at least an understanding of what AI is. But from what you're saying now, it means that ultimately the tech isn't the problem anymore. But once you've got that tech solution, it's then around governance and assurity that the AI tool is. Still doing what it was designed to do.

Mark Rotheram:

I i exactly. The next layer is compliance, governance, control, hosting, certification. You know, we're seeing an awful lot of um new things appearing, like the latest ISO regulation for AI. And what what that's doing is making sure that we are not doing unethical or wrong things with AI. So what we're actually finding is the technology bit of AI is fast. Yeah, we can do things at a pace quickly. And then it's the considerations of right, how do we make it robust? How do we make sure it's secure, governed, how do you make sure that that you understand the training models, you know, talk talking tech for a moment, have come from a place that you're okay to use. Yeah? Um and that there is no risk of malware being injected because you're using an open source one that was originated in maybe an unfriendly country. So it's that governance and control and security that really is the thing that people need to be aware of as we go forwards and staying compliant um over time, as with anything, uh, you know, just look at security in general, posture changes, policies change, risks change. It's it's how we manage and maintain that that's going to be really important.

Fraser Dear:

Okay. So uh an organizational board is being told that they need to do AI. That all sounds lovely, and people get excited about that. Some people maybe don't get so excited about that, but is there still this kind of missing piece? Because sometimes when I talk to organizations, this kind of concept of a data asset is sometimes missing. You know, what what kind of data do we really need for AI to work?

Mark Rotheram:

Yeah, it's it's really interesting. And and thinking about data. We've got two types of data, we've got structured and unstructured. The AI and the outcome that we want from that could be dependent on one or both, or it could depend on brand new data that you've not actually got in your organization. So I think it's understanding what outcomes we want to drive, and then looking at the maturity of the data that we're dealing with and making sure that it is in a usable state. We do have a lot of businesses that want to do AI, but they haven't invested in all of the data platforms and the the security and the governance of data. That's kind of a prerequisite. Now that doesn't mean that they can't do AI, it just means that there's different starting points. We can do AI recognising that the data maturity is low, as long as we're working in the structured way with data that it's okay. Yeah, it doesn't all need to be perfect. Some of our customers we need to take on a bit of a maturity journey with the data first, because their outcomes rely on that. As with all things, data is key, but outcome of what we want to do with it is the most important. Uh there's no point in having the best data platform in the world with all of the structure and the capability if we're not making good decisions off the back of it, and we're not having good outcomes from any of our processes off the back of it as well. So it's thinking about it from a what is the key decision, the outcome that we want from this data, and then working backwards into right, how do we make sure that that's in a position for us to leverage.

Fraser Dear:

Okay. So organizations don't need to have all of the data all sorted, all tickety-boo before you can start your AI journey. But you definitely need to have some structure or at least some semblance of organized data sets if you're going to then apply AI over the top of it. And I think that kind of comes towards the last part of the conversation today around kind of the balance between the cost of implementation, the value to the business, and the economics of technology going forward. Because, you know, AI for AI's sake is awesome. The tech nerds in the room will be over the moon, but the reality is that's not how business works. And so, you know, can you give us kind of a little bit of a conversation around the return of investment when we come to do some transformation using AI?

Mark Rotheram:

Yeah, and i it's critical that we partner with our clients in the right way so that they've got a really clear understanding of what the outcome is that we're going to get. And I talk about outcomes a lot because what we want to do is provide an outcome that ticks all the boxes. Now, the value of AI could be quality, and and what we do there is discover and identify where quality is not good, and then we map that in as a clear outcome for any process that we put together that involves AI so we can demonstrate it going forwards. The most obvious one that we get asked about is cost. You know, it's it's it's fairly straightforward maths that we try and work out here is cost of existing process and cost to scale it, cost of things going wrong, cost of redoing work or the quality being bad, and then pairing that out with a solution where maybe we're not guaranteeing 100% accuracy, but what we will be able to do is provide maybe an 80% automation with a human oversight that can validate and check and train going forwards. So we work everything into some form of ROI-based outcome where we can be clear on what target we're going for, how we're going to get there, and what it's going to look like on the other side. And equally paint the pictures of the risks. You know, if we started the journey two years ago on an AI transformation, the risks would be, well, actually, what we're doing now might get overtaken by AI before we've done it. What we're seeing now is we're at such an accelerated point in time where even if AI got twice as good overnight, we we're already in that position of being able to get the value straight away. So it's first and foremost in our kind of discussions to have those, why are we doing a thing, what are the clear outcomes, how are we going to measure success, and then we paint the picture of well, this is how much effort and cost it's going to be to design, build, deploy, and maintain to demonstrate that value that the customer or the client is interested in.

Fraser Dear:

Brilliant. Well, look, um if I try and pull this all together into kind of some sort of summary, 2026 sounds like it's going to be the year of the turning point for AI and automation simply because the technology is now there to do it. The challenge is actually in the human aspects, where A, we need to be able to understand what capabilities AI has, but B, we need to be able to deploy that and engage all of our business users with the outcome of that AI.

Sinéad Hammond:

Wow, what an episode. Thank you so much, Mark and Fraser, for all the insights you've shared today.

Mark Rotheram:

Thank you very much.

Sinéad Hammond:

What a big year for AI and automation 2026 is going to be. Thank you to our listeners. I hope you took away a few key actions to kickstart the new year. As always, if you have any questions or want to work with us on any of your business projects, you can get in touch at bcn.co.uk. See you again soon.