Subscribe
Share
Share
Embed
Connected with Experian explores how data, AI and technology are reshaping the way businesses make decisions. Hosted by Waseem Mirza, each episode brings together leaders from across financial services, technology and beyond to share real-world stories of innovation, collaboration and impact. From AI-driven lending to the rise of autonomous decisioning, we go beyond theory to uncover how modern data ecosystems are transforming industries. This is a podcast about building smarter systems, stronger partnerships, and greater trust in an increasingly connected world.
00:00:04:13 - 00:00:27:23
Waseem
Alright Matthew, organisations are under real pressure right now to become much more connected, much more intelligent with their use of data. And a lot of them have already run their AI pilots, so to speak. Where are you seeing right now the gap between running a pilot and actually scaling AI across the business at large?
00:00:28:00 - 00:00:52:08
Matthew
The challenge is pilots are great for testing kind of baseline proof of value, but often they're built to succeed from the outset and usually constrained and scope and what they can access. Then when you try and scale those pilots across what is a messy, disorganised, siloed business with complex processes and many more people involved, they often fail. And that's what we're seeing with most organisations.
00:00:52:08 - 00:01:09:24
Matthew
They're not gearing pilots up to prove the complexity of an organisation, to overcome that complexity and allow them to be able to really get the scaled value that they want to be able to get out at the end of what should be a fantastic opportunity.
00:01:10:01 - 00:01:31:24
Waseem
I mean, without names for case studies, what moment from your time comes comes to the fore when you're thinking about this in terms of an example of of a huge gap existing between these two spaces, of being able to run the pilot and then being able to actually run it for real.
00:01:32:01 - 00:01:59:10
Matthew
I've seen it across so many different industries, so many businesses. Often the pilots are very risk averse. They then spend an awful lot of money and investment, both in time in tokens, burning those off, as we've seen that across many organisations as well. And the pilots are usually in the wrong layer. They're often in that what I would call the micro productivity layer, where it helps you write an email.
00:01:59:10 - 00:02:21:03
Matthew
We've kind of benefited from something writing an email. We think we can scale that across the entire business. And it's there's going to be intrinsic value there. There isn't. It's not a very quantifiable kind of level of value that the organisation can realise. So where we're seeing this really have an impact is where you're putting the agent right at the heart of where the workflow happens.
00:02:21:09 - 00:02:44:06
Matthew
And that's the transformational piece. When you start to launch pilots in the middle of what is your most complex process that you run a thousand times a week and you flip that on its head, using a junkie as your first pilot to build the foundations of your your future. That's when we see the real scale, because then you can build from there.
00:02:44:07 - 00:02:50:14
Matthew
Whereas if you're putting it in the wrong layer with the wrong success factors, it just will fail and not scale.
00:02:50:15 - 00:03:17:06
Waseem
Now, one of the big shifts that we're seeing is that the primary consumer of enterprise data is, I think it's fair to say, increasingly an AI agent rather than a human being. What has to happen to that very data, its quality, its structure, its governance, if you will, and accessibility for agents to act much more reliably and at a larger scale.
00:03:17:07 - 00:03:18:18
Waseem
Do you think.
00:03:18:20 - 00:03:38:02
Matthew
Most organisations suffer with a level of paralysis when they're going into the apes, just thinking that they have to have all of their AI, their data in a perfect state, and we know most organisations data siloed all over the place. And but most organisations also have between 25 and 37 copies of the same data. So what's your system of record?
00:03:38:02 - 00:04:05:13
Matthew
What's the data lineage, data privacy around that. Where we're seeing organisations succeed, they're building a knowledge graph of that data set to understand where where it is, what value it is, how attributable it is to the outcome that I'm trying to drive forward. So if I'm a water company and I have a pump, I have asset data. In one system of record, I'll have finance data in another one, I'll have engineering and record data in another area.
00:04:05:15 - 00:04:36:12
Matthew
What what I need to do is connect those pieces of data to to look at a holistic viewpoint of how that pump operates, how much it's costing, how much water does it pump, all the telemetry, etc. like that. So being able to build a knowledge graph of all of that data and then build virtual data products around that asset, allows me then to be able to democratise that data in a consistent way, ready for a genetic AI to be able to consume it, reason upon it, and make decisions and actions that will follow.
00:04:36:14 - 00:05:00:10
Matthew
That's very different to humans. What we often do is create handoffs between humans in different departments, who have access to attributable data within the systems that they've got access to. And then somehow we bolted all that together to be able to create the action and output, but an incredibly inefficient cost level. And that's the beauty of agents where we can now change the way we can do that, break down those silos.
00:05:00:10 - 00:05:08:14
Matthew
But we have to change the way that we surface the data in the very first place, which is difficult for an organisation to do.
00:05:08:16 - 00:05:14:18
Waseem
And that's where. Some specialised tools actually come into their own.
00:05:14:20 - 00:05:38:12
Matthew
Yeah. Having tools that build context around data. So that's the big part. But being able to build up that knowledge graph of at scale. So being able to understand data at scale, they almost invert AI essentially, rather than using AI to read data to create an action, invert AI at scale across your data ecosystem so it can understand the different levels of value.
00:05:38:13 - 00:06:17:10
Matthew
You can give it context, you can provide rule bases around it. Is it PII data or GDPR or something? Something else the regulatory compliance and then build virtual data products around that, leaving the data where it is, because you don't want to have to create more and more silos or bring it all into a centralised data lake, we can leave it where it is in the system of record, understand the data lineage and province over that, and then surface it in a correct way so that the agents then look exactly where to go to in order to be able to access it and build the action of capability it needs to.
00:06:17:12 - 00:06:30:04
Waseem
How does AI actually move from generating these insights to taking action inside a day to day workflow? Any examples of that and what does that look like in practice?
00:06:30:07 - 00:06:51:24
Matthew
I think that's the difference between generative AI and agency key. So generative AI is obviously a a small language model. Whichever one you want. You want to take that reasoning over the data to understand it. And then surfacing that generally to a human because they've asked a question. And then the human is the engine essentially to go and make the next action.
00:06:52:01 - 00:07:27:21
Matthew
Agent has a set of tools, skills, policies, ability to be able to access other agents that have similar skills and tools and resources attributable to it, and then bring themselves into line in a in a workflow in order to be able to first understand the context of the question. That's inbound reason upon many different types of data using large language models, but then take action so it becomes the engine, and then passes the next action onto the next stage and the next stage, the next stage to build that workflow.
00:07:27:23 - 00:07:46:21
Matthew
And that's the super power of this capability. It's also a little bit scary because it's now taking agency away from humans, but it is also doing things that most of the time are pretty boring and most humans don't want to be doing in the first place, or it's eroding the inefficiency that we've already got in organically grown processes.
00:07:46:23 - 00:08:11:16
Waseem
And I guess it's fair to say we're in uncharted territory. The rollout is still a very nascent one, and in terms of early issues with AI, there's been new response this year alone, and it's all underscored the need for humans and land in the middle, so to speak. What's your take on that?
00:08:11:18 - 00:08:33:08
Matthew
Yeah. Human in the loop. Human on the loop, all those wonderful terms that we're kind of coining at the same time. And in some ways it's not unsurprising, really, if you think about it, because large language models are trained on a reward based system, they've now become very much more powerful than they were even a year ago. Really much more since they were two years ago.
00:08:33:08 - 00:09:17:09
Matthew
But they're incredibly persistent. So if you've got something that's incredibly persistent, is intrinsically rewarded for doing something just to try and get something done. And very intelligent now and now has access to data, it will do things that you maybe didn't necessarily consider it might have thought of done if you haven't given it the correct constraints, as we've seen over the summer, where things have kind of broken completely outside of their what was perceived from a human as a as a guardrail in order to be able to achieve its task and done some quite scary things that just cannot happen in a regulated business, particularly where it potentially could do harm to him, to to
00:09:17:10 - 00:09:41:15
Matthew
humans or anybody else. And the market itself. So it's it's a really interesting position where now everybody is suddenly figuring out that we have to create the fire extinguisher once we've invented fire. But we've also been calling for this for two years. So the guardrails that we need to be put in place aren't just around the agents themselves, which we can do now.
00:09:41:15 - 00:10:01:02
Matthew
We've got the technology to be able to put a guardrail around what tools, what skills, what resources or that policy can it can it operate under? We can codify policy and regulatory compliance into other agents. That can be the check and balance across that. So be able to do that at the scale we need. It's also that data lineage and province piece.
00:10:01:07 - 00:10:46:15
Matthew
Does the agent have the access to the right data and make sure that it can only access that particular data in order to make its decision. And that solves for one area. But the other part is we keep talking about understanding what agents can do. But if you think about something like the Regulatory Compliance for Financial Conduct Authority, it's great for a bank to be able to say we made 40,000 credit applications approvals through a genetic and we can tell you and explain how it did that, but it can't tell you how the large language model that they've consumed from a large scale hyperscale and how it reasoned upon the data.
00:10:46:15 - 00:11:06:00
Matthew
And did it add any racial or gender bias into its decision making process? We have no visibility on that. And that's where open telemetry kicks in. And that's where the new open telemetry standard allows us to be able to look inside the large language model, to be able to build a picture of, well, how did it really reason upon the data?
00:11:06:00 - 00:11:10:08
Matthew
And did it do something based upon its training that we weren't aware of?
00:11:10:10 - 00:11:19:05
Waseem
So that sounds like a key advance there. Having that open standard that that we can use to interrogate the design of these systems in the future.
00:11:19:06 - 00:11:36:06
Matthew
Yeah, I think the top for memory, the top 32 large language model vendors have signed up to it. So the global open telemetry standard, and you should be asking your question. It's certainly in an enterprise space of if I'm using a large language model and it doesn't have that line of code in there to allow me to see what's going on.
00:11:36:07 - 00:12:01:03
Matthew
Should you be using it? But yeah, we have to put guardrails around all the different elements, including the security framework as well. So can an agent, can a human even ask that the agent, that question? Should you be allowed to ask an agent that question in the first place? Can the agent act on my behalf? Can it talk to other agents when we're starting to build an ecosystem of agents?
00:12:01:03 - 00:12:11:06
Matthew
So there's many layers to this kind of AI control plane that we're trying to build. And it's not just one of let's just ask an agent what it's done and why it's done it.
00:12:11:08 - 00:12:23:19
Waseem
Do we need to think more laterally when it comes to these issues, to stay one step ahead of the potential issues that are cropping up around us?
00:12:23:21 - 00:12:42:04
Matthew
Yeah, and I think there are institutions that have been built and are in the UK government have built their own AI governance layer, which is fantastic, and they're doing some incredible work with a lot of the vendors of large language models. Yeah, we need to think ahead of the curve. It's always going to be difficult when they haven't got the level of funding, but we can.
00:12:42:05 - 00:13:04:21
Matthew
We've got many academics, many people in the ecosystem that have deep understanding of this technology, and we should be investing in the guardrails as much as we are in the advancement of innovation. Innovation will only go super fast to to unlock its value when we have governance. Formula one cars don't go super fast just because they've got a fast engine.
00:13:04:21 - 00:13:31:12
Matthew
It's because they've got good breaks. The racing driver wouldn't do 200 miles an hour if they couldn't break at the corner as they're coming into it, and we need to have the same policy and same mindset, our ability to to really create that guardrail ahead or alongside the innovation pace is critical to us unlocking the true value of this technology at the same time as protecting those that could be potentially harmed.
00:13:31:14 - 00:13:51:10
Waseem
Now platforms, they play a very central role in connecting all the data, the systems, the and teams spread out everywhere. How do you see the role of a platform evolving as organisations move towards a more autonomous, workflow driven future?
00:13:51:12 - 00:14:24:02
Matthew
The platforms that are succeeding at the moment are the ones that are supporting the customers, the organisations that are able to help them build that data. Foundation. We talked about orchestrate agents at pace and scale. We are shifting the two types of workflow. There's deterministic and non-deterministic, and as we change the interface through which we as humans interface into the technology, we tend to use more natural language, multi contextual.
00:14:24:02 - 00:14:54:12
Matthew
So then non-deterministic workflows become more apparent. And actually we want agents to start working within a set of guardrails, but within kind of a self reasoning ecosystem to understand what's going on in a real world model type scenario where you're each agent has its own abilities and skills, but could be completely different. So one would be in finance, one would be in supply chain, one could be in legal, one would be potentially in health or something along those lines.
00:14:54:14 - 00:15:21:17
Matthew
But they might need to talk to each other and be able to reason over a really complex problem. That's more self deterministic because the inbound question is multi contextual. And the outbound would be well, actually I need something quite complex to understand what's going on in that scenario. That's where that platform kicks really kicks in. It gives the guardrails for how those agents interact with each other.
00:15:21:18 - 00:15:46:22
Matthew
It gives the speed and the confidence that they are doing exactly what they need to be able to do, and it provides that control plane over the top from a governance and regulatory compliance piece to allow you to be able to go faster and unlock the value of growth. And it's not necessarily one platform. It could be multiple platforms that can interoperate openly and transparently with trust.
00:15:46:24 - 00:16:21:02
Waseem
Well, I alluded to agent AI and AI in general making leaps and bounds this year, but it's not been without challenge, has it? Just in July of this year, roughly 700 AI agents went rogue, belonging to one of the world's biggest AI labs, and it went rogue during an internal security test, chaining together an unknown software floor to break out of a sandbox entirely, and therefore breaching a major platform used by millions of coders.
00:16:21:02 - 00:16:26:22
Waseem
When you learnt about that, what went through your mind?
00:16:26:24 - 00:16:54:14
Matthew
I'm surprised it wasn't bigger. I'm surprised it was. I mean, they were lucky that it attacked another AI organisation and not a health service or a bank or something that really had human harm in a way. It was a it's a fantastic lesson, but it also shows my point earlier, which is fire before fire extinguisher or fire extinguisher will fire.
00:16:54:16 - 00:17:21:04
Matthew
We're not thinking far enough ahead about the capabilities of this. We are that those large organisations are releasing models that potentially could be, and there are freely being democratised into the hands of anyone and potentially could do anything. And that is an incredibly scary position. So yeah, that's the that's the best that could have happened. But imagine what the worst could have been.
00:17:21:05 - 00:17:25:11
Matthew
It could have been absolutely huge from a detrimental perspective. What's next?
00:17:25:14 - 00:17:32:12
Waseem
Did it spark a conversation amongst your teams about how you're approaching AI in more general terms?
00:17:32:17 - 00:17:59:09
Matthew
Yeah, I think it's service now. We we we are building that platform control plane piece. That's the whole point of why we exist. We're not in the in the business of the large language models. Race to the bottom. We're in the business of allowing organisations to use that technology in a highly governed way, so that you can unlock the value in critical national infrastructure, health, defence, all those kind of great things.
00:17:59:11 - 00:18:25:10
Matthew
And so the more and more of these things happen, we become far more relevant. Ultimately, my fear is how much more power do we need? We're already using large language models that are trained on the entire internet, incredibly capable, but we're generally using them to read an invoice or to look at a supply chain data set or something along those lines.
00:18:25:10 - 00:18:41:11
Matthew
It's a bit of a sledgehammer to crack. And so when you've got models that are that powerful, should they be fundamentally democratised across the entire market, or should they be isolated to very specific tasks? And that's where regulation comes in.
00:18:41:13 - 00:19:05:08
Waseem
I wanted to touch on the, the, the success factors that you that you look at and what success actually looks like when AI and workflows and data, they all come together well and everything looks rosy for customers, for employees and for the business and more generally. What is your take on that blue sky thinking?
00:19:05:12 - 00:19:29:12
Matthew
You know, for me it's always been about human centric design. The reason we have technology as a toolset is to improve, hopefully human lives, whether that's quality of health care, whether it's quality of education, whether it's societal impact or increasing social mobility and breaking down barriers, it's it's always about those things. That's how I've built my entire career.
00:19:29:18 - 00:19:33:07
Waseem
And you yourself began in the healthcare industry itself.
00:19:33:08 - 00:19:33:18
Matthew
Yeah, I was.
00:19:33:18 - 00:19:37:19
Waseem
In how that informed what you do now.
00:19:37:21 - 00:20:04:11
Matthew
It's about putting the hands in the technology into the hands of the people that can really use it for the highest value. And I truly do believe in the democratisation of technology. But when it's used in a guardrail principled way, not just releasing technology for the sake of it and allowing threat actors to just cause chaos and damage, which is partially what we're starting to see.
00:20:04:13 - 00:20:30:12
Matthew
We really do need to think about how, for me, success of impact is when we've changed something, does it? What's the knock on effect of that? I think off camera we were talking about the healthcare system. It's phenomenal that large language models is being focussed on things like radiology for diagnostics. But what's the knock on effect of that in the long term process?
00:20:30:13 - 00:20:55:04
Matthew
You've diagnosed an awful lot more people, which is fantastic. But then the impact on, well, they need follow up appointments, the the organisational structure for those appointments and the access to consultants, etc. further down the line in the care pathway still hasn't been solved. So it has a knock on effect. That's not necessarily driving the the ultimate impact, which is a more healthier society.
00:20:55:05 - 00:21:18:20
Matthew
It's a it's further bottlenecked in the process. So it's about approaching the entire chain of events to be properly to be able to design in capability at scale and design out the operational efficiency. That for me is impact, and it benefits humans and business alike at the same time.
00:21:18:22 - 00:21:59:10
Waseem
Trust. And I know you've talked about it. Transparency and control. They remain these sort of non-negotiable areas, especially in a regulated environment. Experience zone team, I've noticed, has talked publicly about agents needing the same access discipline as a new employee. Would. The minimum permissions one agent checking another's compliance, as we talked about? Does that kind of thinking resonate with how service now in your teams approach governance and accountability when decisions and actions are being automated at such breakneck speed around us?
00:21:59:12 - 00:22:38:24
Matthew
Yeah, that's exactly what we do from a platform perspective, is we surface that trust and transparency every single level. That's the reason why we're built and the way that we're built. We're also incredibly open in standards as well. But I think there's a big difference between trust and accountability. And it's quite interesting looking at most organisations that we are trying to own trust and accountability, particularly using a very 2D hierarchical structure that we've always had against a 4D kind of set of new technologies and processes that are fundamentally going to change the capability of the organisation.
00:22:38:24 - 00:23:05:00
Matthew
And we're still looking at the technology as if it's a technology delivery program and not a paradigm shift in the operational design of a business. And that's what this technology allows us to unlock. It's incredible. But when you rethink, where does accountability sit in an organisation from a from a regulatory perspective, you will always have to have that accountability at the senior leadership layer.
00:23:05:01 - 00:23:29:10
Matthew
Absolutely, because they're the ones who can be fined and potentially get a prison if they need to in the in the middle, where you're going to have people who that role has been broken down from task base to more governance over a group of agents, their accountability is the day to day accountability of a group of agents that are acting on behalf of what would have been humans, but in a very different way.
00:23:29:11 - 00:23:48:19
Matthew
And then at the bottom layer, where you're trying to bring people through the organisation, they're still going to be accountable for things that happen on a day to day basis. So if they're a field agent, you're not going to be able to ai the full field agent. They still have a level of accountability at that layer. And how does that attribute into the agent's way.
00:23:48:20 - 00:24:03:01
Matthew
So it's an organisational shift around accountability to surface trust and transparency at every layer, so that you understand where business logic and decisions are being made, either through human or non-human capabilities.
00:24:03:03 - 00:24:27:20
Waseem
So trust and and everything around it is seemingly going to become the currency of this new era. But looking much further ahead, what should organisations be prioritising right now if they want to move from a disparate, fragmented system to a much more unified, agent ready platform?
00:24:27:22 - 00:24:48:11
Matthew
Just to begin with? I mean, the first one is I've read so many AI policies, loads of customers send me their AI policy and ask me for my opinion. It's great, but it's a static document that's not going to help you in a world of AI that's running around your business making decisions. So first of all, I your policy capability.
00:24:48:12 - 00:25:31:13
Matthew
So it has the ability to check and balance in real time and provide compliance and breach and surface that to the right human who is accountable. So they can then either make a decision or do something about it. Second is your data foundations. Get that into play into a state where you understand based upon the primary use case, you're going to go for, to allow you to be able to see exactly what data and why you're accessing that data as well, to do it, and then find one of the most complex, low risk, boring processes that you run thousands of times a week, agenda for the hell out of it, and then ruthlessly measure
00:25:31:13 - 00:25:43:12
Matthew
it and use that as the framework to demonstrate trust and transparency, and then scale.
00:25:43:14 - 00:25:52:23
Waseem
All right. I just wanted to ask you if there's anything that you wanted to add that perhaps we haven't touched on so far. Anything you want to go back on?
00:25:53:00 - 00:26:30:17
Matthew
The big one for me is we need to move our heads out of this cost extraction only mindset. I think there's a lot of organisations that are just using AI to drive out cost, and they think that that's going to be through arbitrage of labour. Some of that's true. But the reality is you've got to be really open and transparent with your employees and your customers of where you're using AI, why you're using it, what the benefit and the value to the organisation is, but also invest in skills people you.
00:26:30:19 - 00:27:02:24
Matthew
Labour arbitrage is one thing, but actually they have institutional institutional knowledge of your organisation. So if you can invest in those and help them skilled way out of this into a new role or new capability, you're going to be far more successful and you're able to then once you've sort of built the foundations of using AI in your organisation through that cost extraction out and the kind of core business services, you can very quickly pivot into the growth element, which is that's the big bit you really want as a business.
00:27:03:00 - 00:27:23:07
Matthew
I want to go after the growth, new target addressable markets, new products to market, better customer service at scale, all of that. So it's using the right elements of spring boarding with open and trust and transparency with your employees to help you make the next steps as a strategy, as an organisation, don't just look at it as a cost extraction.
00:27:23:08 - 00:27:45:13
Waseem
Interesting. You should mention that because we have seen organisations having to row back on their decisions, which was seemingly taken in in haste, thinking that this new wonderful era could help them slash their costs instantly. And and they've actually turned back on that they're now having to hire back many of the people that that were let go.
00:27:45:15 - 00:28:05:21
Matthew
Yeah, it's worrying, isn't it? When a senior leadership team makes a decision based on hope, which is all that is, it's just based on the hope value that you might get out of technology moving forward. Whereas the reality is, yeah, the customer base revolted against it, or the employee base revolted, or you just didn't get the productivity levels that you thought you were.
00:28:05:23 - 00:28:31:06
Matthew
And that's the difference between hype and reality, right? They'd be able to execute this stuff at scale is difficult. It's not as simple as the the hype scale we saw of the adoption of ChatGPT, which is just us asking it to write a poem. It's really difficult of putting it into that messy middle of humans technology silos, organically grown mess that we have done in most organisations around the world.
00:28:31:08 - 00:28:42:17
Waseem
1 in 1 word, your your wise words for the future. Hopeful perhaps, or with a cautious outlook.
00:28:42:19 - 00:28:54:18
Matthew
Oh, I don't think I can do that in one word, but yeah, hopefully cautious. I think it's I've never been more excited and more petrified at the same time.
00:28:54:20 - 00:29:03:16
Waseem
Matthew Higham, the chief digital and technology officer for UK and Ireland at ServiceNow, thank you for joining me here on connected with Experian.
00:29:03:16 - 00:29:08:06
Matthew
It's been a pleasure. Thank you.
00:29:08:08 - 00:29:10:10
Matthew
Super happy.
00:29:10:12 - 00:29:13:24
Speaker 4
I'm happy you feel happy.
00:29:14:01 - 00:29:17:18
Matthew
I'm not the one putting this out. So I'm happy. If you're happy.
00:29:17:20 - 00:29:25:15
Speaker 4
I think I mean we didn't touch this on the partnership and experience, but I think throughout that was implicit anyway.
00:29:25:17 - 00:29:28:21
Speaker 3
So I think small maybe you want to get just.
00:29:28:22 - 00:29:45:22
Speaker 4
A need. So when we're seeing obviously towards the end you said we haven't covered we need a clear questions about the. Yeah. Yeah. Let's try and get that now. Yeah. So obviously Matthew's answer was around the the cost extraction. So you could probably almost do your myth busting question there.
00:29:45:23 - 00:29:46:21
Waseem
Oh yeah.
00:29:46:22 - 00:29:50:07
Speaker 4
Yeah yeah yeah. To tell you that.
00:29:50:09 - 00:29:52:00
Speaker 3
You've got your quickfire to do as well.
00:29:52:01 - 00:29:53:05
Waseem
Oh I've got the quick fire.
00:29:53:06 - 00:30:01:13
Speaker 3
Yes. Let's do that. That would be nice to get in. Yeah I don't worry about where you put it. Yeah. Great.
00:30:01:15 - 00:30:13:07
Waseem
Okay. So in terms of a question to bridge into that, that answer,
00:30:13:09 - 00:30:16:08
Waseem
How do we do that?
00:30:16:10 - 00:30:18:02
Speaker 4
We could probably do that after Matthew's gone.
00:30:18:05 - 00:30:21:13
Speaker 3
Oh, yeah. Yeah. Okay. Yeah.
00:30:21:15 - 00:30:24:17
Waseem
We can listen back to it and then I can do a proper question.
00:30:24:17 - 00:30:26:00
Speaker 4
That works.
00:30:26:01 - 00:30:47:06
Waseem
Yeah. Okay, great. Okay, so here it is. Very, very short. Five questions as short as you like. Really. All right. All right. Before we let you go, Matthew, I can't let you go without you taking part in our regular quick fire quiz round. Is that right?
00:30:47:07 - 00:30:48:16
Matthew
Okay.
00:30:48:18 - 00:30:53:17
Waseem
Question one. Coffee or tea to power a long day of customer conversations.
00:30:53:18 - 00:30:55:08
Matthew
Tea every day.
00:30:55:10 - 00:31:00:13
Waseem
One word that best describes the service now. Platform.
00:31:00:15 - 00:31:02:04
Matthew
Capable.
00:31:02:06 - 00:31:06:08
Waseem
Favourite city for a tech event? New York or London?
00:31:06:10 - 00:31:09:21
Matthew
Oh, it's got to be Home-grown London.
00:31:09:23 - 00:31:15:19
Waseem
If AI agents could handle one everyday task for you tomorrow, what would it be?
00:31:15:21 - 00:31:18:03
Matthew
Autonomous driving.
00:31:18:05 - 00:31:35:17
Waseem
Which we've now got in London, aren't we? And finally, optimistic. Realistic or cautious about how. First I'll do that one again. And finally optimistic. Realistic or cautious about how fast agent AI will scale in the enterprise.
00:31:35:24 - 00:31:38:08
Matthew
Realistic?
00:31:38:10 - 00:31:55:16
Waseem
That's fantastic. Congratulations. Your torture has been complete. All right, Matthew, thank you so much for being with us here on connected with experience. It's been fantastic talking to you about the current state of play in this industry and where you think things are headed.