Latest / Speaking Your Language / Episode 6 - Data, People and AI: What successful transformation really takes

Transcript

Episode 6 - Data, People and AI: What successful transformation really takes · Speaking Your Language

• 0:00
  SPEAKER_01: Welcome to Speaking Your Language, the AI podcast that is focused on how you really implement AI into your business.
• 0:07
  SPEAKER_01: No hype, no noise, no jargon.
• 0:13
  SPEAKER_01: So our guest today is somebody that I've known for 25 years, believe it or not, and somehow ended up working in the same space, AI.
• 0:22
  SPEAKER_01: He's been working in the space as long as I have.
• 0:24
  SPEAKER_01: I'm gonna let him introduce himself and then we're gonna have a good old chat.
• 0:27
  SPEAKER_01: Over to you.
• 0:28
  SPEAKER_00: Thanks very much, Marco.
• 0:29
  SPEAKER_00: Good to be here.
• 0:30
  SPEAKER_00: So yeah, my name's Oliver, Oliver Heyman.
• 0:33
  SPEAKER_00: Um I'm director of partnerships at Configure.
• 0:35
  SPEAKER_00: And welcome.
• 0:36
  SPEAKER_00: Thank you very much.
• 0:37
  SPEAKER_01: It's been six months now, but welcome anyway.
• 0:38
  SPEAKER_00: Yeah, thank you.
• 0:39
  SPEAKER_00: Um yeah, and my career has been um done a done a few different things.
• 0:44
  SPEAKER_00: So started off as a commodity commodity market analyst in the oil and gas industry.
• 0:47
  SPEAKER_00: So working with massive data sets, lots of data analysis, uh, then moved into into technology.
• 0:53
  SPEAKER_00: So started um working with a technology startup based in Cardiff in 2018, working on um what now feels like quite primitive AI solutions, machine learning models.
• 1:03
  SPEAKER_00: Um, did that for almost eight years and then joined Configure to basically bring your ex our excellent product suite to uh to a large audience.
• 1:11
  SPEAKER_01: Nice of you to say.
• 1:12
  SPEAKER_01: I think one of the I think that's a really important point to start on, though, is you've been in the data space a long time.
• 1:17
  SPEAKER_01: You know how important data is, and that's one of our core pillars, right?
• 1:21
  SPEAKER_01: And the reason we're having this chat today is we were thinking, well, who do we bring on to talk about data?
• 1:26
  SPEAKER_01: And who better than someone that's coming over to the company in the last six months?
• 1:30
  SPEAKER_01: But let's address the elephant in the room before we get into any of the data stuff.
• 1:33
  SPEAKER_01: We always said we couldn't work together.
• 1:35
  SPEAKER_01: Yeah.
• 1:36
  SPEAKER_01: It's it's a it's a challenge, right?
• 1:38
  SPEAKER_01: If you think you've been friends for a long time, can you work together?
• 1:41
  SPEAKER_01: How's it going, Al?
• 1:42
  SPEAKER_00: Uh if I wasn't live on a podcast, maybe I could be honest.
• 1:46
  SPEAKER_00: No, no, no.
• 1:46
  SPEAKER_00: It's been it's been great.
• 1:48
  SPEAKER_00: Um, I think you do take some liberties around things that don't necessarily appear in the job description.
• 1:53
  SPEAKER_00: Um, like come to this event, do this presentation, record a podcast, uh, stand in at very, very early networking meetings for me.
• 2:02
  SPEAKER_01: So, you know, it's just all part of the course.
• 2:03
  SPEAKER_01: It is, it is.
• 2:04
  SPEAKER_01: And you do that as a friend as well.
• 2:06
  SPEAKER_01: It's always good.
• 2:07
  SPEAKER_01: But let's get on to the data side.
• 2:09
  SPEAKER_01: Why do you think, from your perspective, data's so important in an AI transformation journey?
• 2:15
  SPEAKER_00: Um, because data is effec effectively the the lifeblood for AI.
• 2:20
  SPEAKER_00: So AI without data is just uh is just a big model spitting out gibberish.
• 2:26
  SPEAKER_00: So it's a really key thing for organizations to understand not only the importance of their data, and obviously there's lots of things around getting the data in the right shape and making it usable.
• 2:37
  SPEAKER_00: And that they they, I'm sure we'll get onto those discussions later, but also recognizing the value that is inherent in the data that organizations have.
• 2:45
  SPEAKER_00: So, yes, we're gonna talk and we can talk, and and we do this with our customers, build really impressive processes powered by AI.
• 2:54
  SPEAKER_00: The technology is so advanced now that it can do really powerful things, really take the burden off organizations and teams and individuals within them.
• 3:03
  SPEAKER_00: But the data is core to all of that working effectively and fitting together.
• 3:07
  SPEAKER_00: So it's a fundamental pillar of any transformation program and all of our transformation programs.
• 3:14
  SPEAKER_00: Um, and yeah, getting organizations to understand the value and actually that they have a, they have an asset in their organization data.
• 3:21
  SPEAKER_00: Okay, it might live in different places, it might live in different systems, there might be, you know, different governance structures around it, there'll be different owners for that data.
• 3:29
  SPEAKER_00: But really, where the powerful, you know, where it becomes very powerful is when you bring all that together and then have that as a really solid foundation upon which you can start layering in then more advanced processes, like AI-driven processes.
• 3:41
  SPEAKER_00: And that's when you can really get amazing results.
• 3:44
  SPEAKER_01: Yeah.
• 3:44
  SPEAKER_01: And I I think AI has changed the landscape of what people class as data as well.
• 3:49
  SPEAKER_01: Like when we talk about data, typically we always talk structured.
• 3:51
  SPEAKER_00: Yeah, exactly.
• 3:52
  SPEAKER_01: And like you say, a neural network, an AI model has no value unless it has the data.
• 3:56
  SPEAKER_01: And the data could be from the web, it could be from the things it's trolled, but the real power is leveraging it with your own data because it makes the model better.
• 4:05
  SPEAKER_00: Yeah, exactly.
• 4:06
  SPEAKER_01: And and what do you think is the thing that people need to really realize with the models themselves and how they operate and how they work?
• 4:14
  SPEAKER_01: Because a lot of people are throwing AI at data and thinking it's going to solve the problem, but that isn't the case, is it?
• 4:20
  SPEAKER_01: What what what do you think?
• 4:22
  SPEAKER_00: Yeah, exactly.
• 4:22
  SPEAKER_00: I think one of the most common problems I think we see is in organizations is people using lots of tools, people experimenting, but it's a very kind of messy process.
• 4:32
  SPEAKER_00: It's very ad hoc.
• 4:33
  SPEAKER_00: There's not necessarily any structure around what it is that they're putting in, um, around prompting, and therefore the outputs of whatever that process can be very variable and unpredictable.
• 4:46
  SPEAKER_00: And in a business process, unpredictability is not a good characteristic.
• 4:50
  SPEAKER_00: Um, you're absolutely right that the you know data, many people think of ones and zeros in a very structured database.
• 4:56
  SPEAKER_00: And of course, that is a very valid, you know, lots of data exists in that way.
• 5:00
  SPEAKER_00: Um, and that structured format can be um a useful kind of foundation for AI processes.
• 5:07
  SPEAKER_00: But with the advancement in large language models and how the technologies move forward, we're able to kind of gobble up much more varied sources of data and bring them into the into play as well.
• 5:18
  SPEAKER_00: So, you know, emails, uh large unstructured documents, um, you know, notes from sales calls, meeting transcripts combined with perhaps the structured data.
• 5:29
  SPEAKER_00: So you've got that combination of structured and unstructured.
• 5:32
  SPEAKER_00: And effectively what we're doing is mirroring how people within organizations would use and process that data, but with an AI system.
• 5:39
  SPEAKER_00: So, you know, any person working in a sales team, for example, would join a sales call, would run that sales call and store all the information in some maybe some handwritten notes and in their own memory.
• 5:50
  SPEAKER_00: Combine that then with the structured information that they might have on their product or service or previous customers in that space, and then produce a series of outputs that combine those two things.
• 5:58
  SPEAKER_00: Effectively, what we're able to do now with the technology is have AI do a lot of that stuff for you.
• 6:03
  SPEAKER_00: And that's just obviously one small area where it can be applied.
• 6:05
  SPEAKER_01: It's a really good example because we take sales as a general example.
• 6:09
  SPEAKER_01: It it was quantitative data previously.
• 6:11
  SPEAKER_01: How many calls did you do?
• 6:13
  SPEAKER_00: Yeah, exactly.
• 6:13
  SPEAKER_01: And it wouldn't be about the qualitative side, but now with models, you're gonna apply things like sentiment and say, okay, how was the call?
• 6:20
  SPEAKER_01: How did it come across?
• 6:22
  SPEAKER_01: And you can take that from a transcript, you can take that from a video, you can match that with the quantitative and the qualitative.
• 6:28
  SPEAKER_01: So all of a sudden, your conversion rate isn't just your conversion rate, it's a you have a reason for it without having to dig around and things like that.
• 6:35
  SPEAKER_01: And that was, I know a lot of the work you were doing previously, uh, because obviously you were Amplify before, was around looking at data that typically was untapped previously or wasn't available, right?
• 6:44
  SPEAKER_01: So, do you want to talk a bit more about that?
• 6:46
  SPEAKER_01: Because I think that's a really good idea.
• 6:47
  SPEAKER_00: Yeah, it was Amplify was all about unstructured data from the internet and all of its variety.
• 6:53
  SPEAKER_00: So anything from you know, news articles, uh company, you know, company reports, company media statements, marketing materials, patents, academic research, literature, uh blog posts, social media in some cases, you know, hugely diverse forms of data, some very, very short, some very, very long form, some very, very technical, some very, very casual, some with lots of, you know, colloquialisms and and you know things that that like innuendo and sarcasm.
• 7:23
  SPEAKER_00: And the challenge really was okay, how do you produce a system that can translate all of that into something meaningful, into some, you know, finding the signal in the noise, if you like?
• 7:32
  SPEAKER_00: Um, so that the the importance of getting that data right and also understanding what in that large mess of data, because if you point, you know, a model at the internet en masse, well, that's not very helpful.
• 7:43
  SPEAKER_00: Like, how do you isolate the bits of that internet, so the bits of content that you think have a greater chance at answering your question or giving you the insight that you need.
• 7:53
  SPEAKER_00: It's a very similar thing, I think, with business data inside organizations.
• 7:57
  SPEAKER_00: Um, you have to try and unpick what you think is the right data for the right use case and the right data for the right process.
• 8:06
  SPEAKER_00: And finding the nuance in that data is actually where you get better results.
• 8:10
  SPEAKER_00: And I think one of the things that's really happened more recently with the technology is actually rather than just getting kind of I think a lot of people experimenting with AI in the past two years would complain about generic outputs.
• 8:22
  SPEAKER_00: So it sounds impressive, it sounds logical.
• 8:25
  SPEAKER_00: There's nothing that you would necessarily say was incorrect in an output, but it's not specific enough to that conversation that I had or that that that customer need.
• 8:33
  SPEAKER_00: And I think that's one of the the the you know, if you can structure and process that data in the right way, combine it with the right prompt or series of prompts or agentic processes if you've got a if you've got a process like that running.
• 8:47
  SPEAKER_00: Now the power of the large language models combined with those elements can actually really get to the nub of that information that you want.
• 8:52
  SPEAKER_00: And it can make each output genuinely unique and genuinely kind of relevant and valuable, therefore, to the to the to the task that you're trying to complete.
• 9:00
  SPEAKER_01: I think that's the key word is value, isn't it?
• 9:02
  SPEAKER_01: It's find the value in what you're trying to achieve, whether that's value for you, whether that's value for your client.
• 9:07
  SPEAKER_01: And you know, as we was, as you were saying then, you can point it at all this information.
• 9:12
  SPEAKER_01: And one of the biggest things that we find is that when you don't centralize data, you don't have a golden record, you don't have that one point of truth, you can throw an AI model at it.
• 9:23
  SPEAKER_01: Doesn't mean they're all going to tell you the same answer.
• 9:25
  SPEAKER_01: Yeah.
• 9:25
  SPEAKER_01: So it's one of those things of saying, okay, I know I have an answer somewhere.
• 9:30
  SPEAKER_01: Yeah.
• 9:31
  SPEAKER_01: But what do I trust and what don't I trust?
• 9:32
  SPEAKER_01: And that's probably just as important now, right, as having the information in the first place.
• 9:37
  SPEAKER_01: Is it being captured right?
• 9:38
  SPEAKER_01: How is it being captured?
• 9:39
  SPEAKER_01: Is it easy to capture so that we can actually make sure we've got it?
• 9:42
  SPEAKER_00: Yeah, exactly.
• 9:42
  SPEAKER_00: And I think one of the ways that you can think about this is deterministic versus probabilistic outputs.
• 9:48
  SPEAKER_00: Large language models are inherently probabilistic.
• 9:50
  SPEAKER_00: That's why they're so great at creative tasks and writing new content and whatever it might be.
• 9:57
  SPEAKER_00: When you combine that and embed it in a business process or embed it as part of your operating model, you need it to be a little bit more deterministic.
• 10:06
  SPEAKER_00: So you want ideally to have greater consistency and a bit more predictability around what the outputs should be.
• 10:13
  SPEAKER_00: And you can control for some of that in the prompts.
• 10:15
  SPEAKER_00: So you can place constraints or restrictions and do this and don't do that.
• 10:19
  SPEAKER_00: But actually, fundamentally, as you say, having a single source of high quality data that is unambiguous, that doesn't have lots of messy nuance is also a really important factor in getting that predictable and therefore valuable in the business context output.
• 10:35
  SPEAKER_00: Yeah.
• 10:36
  SPEAKER_01: Yeah, is I think you're right.
• 10:38
  SPEAKER_01: If you can remove data quality issues, you can remove the silos and try and make them more combined.
• 10:46
  SPEAKER_01: All of a sudden, instead of analysing your marketing and then analysing your sales and analysing your finance, you're analysing it all at once.
• 10:52
  SPEAKER_00: Yeah.
• 10:53
  SPEAKER_01: And that's where the actual real value is, is understanding how it all comes together in a business because we don't operate our business as separate teams, right?
• 11:00
  SPEAKER_00: Yeah, absolutely.
• 11:01
  SPEAKER_01: They are the teams are still delivering to a core goal.
• 11:04
  SPEAKER_01: We've just talked about it on previous podcasts around having a core vision for the business.
• 11:07
  SPEAKER_01: Yeah.
• 11:08
  SPEAKER_01: But we've always treated systems differently or processes differently.
• 11:11
  SPEAKER_00: Exactly.
• 11:11
  SPEAKER_00: So businesses do operate generally, well, there should be.
• 11:15
  SPEAKER_00: The best businesses operate with a, you know, with a single set of goals that every department's striving towards.
• 11:20
  SPEAKER_00: And of course, all departments are interconnected and need to work together and communicate.
• 11:25
  SPEAKER_00: The systems that serve those departments often don't reflect that.
• 11:28
  SPEAKER_00: They're very fragmented.
• 11:30
  SPEAKER_00: Marketing data doesn't talk to sales data, doesn't have any consideration of product or roadmap or engineering.
• 11:35
  SPEAKER_00: You know, uh trying to measure people's capacity or output doesn't sync up with the sales pipeline that's coming down the track.
• 11:43
  SPEAKER_00: All of that stuff is a common problem around businesses.
• 11:46
  SPEAKER_00: So having kind of point-in-time AI analysis in one of your 15 SaaS tools on one fifteenth of your data asset is not a, you know, you might get some small wins from that, but it's not fundamentally or systematically changing the way that your operating model functions.
• 12:03
  SPEAKER_00: It's not delivering value that you, that you know, that you need and that business leaders want to see.
• 12:07
  SPEAKER_00: I think one of the kind of critical things that we hear from business leaders all the time is, you know, how do they create the time and space to step back and actually look at their organization as a whole, look at their operating model as a whole and figure out where AI fits and where where there can be some value derived.
• 12:22
  SPEAKER_00: Because, you know, just adding, like I say, an AI model to one of your 15 systems is probably not going to cut it.
• 12:28
  SPEAKER_00: And I think that's when people can start to get disillusioned, you know, not to call out any CRM operator, but okay, maybe there's a an AI feature that's been plugged into your CRM all of a sudden in your last upgrade.
• 12:39
  SPEAKER_00: Okay, great.
• 12:40
  SPEAKER_00: But it's, you know, if if my marketing data isn't in there and my other data that I need to inform the sales team's activity is not in there, then you're getting a very, very partial view.
• 12:49
  SPEAKER_00: And that's where that's where the limitations are.
• 12:51
  SPEAKER_01: Yeah, and you're right.
• 12:52
  SPEAKER_01: I think that's what we've started to see is at the moment there is a very big focus rightly, because there's not enough time in this world for for people.
• 13:02
  SPEAKER_01: So operationally, we're looking at efficiencies.
• 13:04
  SPEAKER_00: Yeah.
• 13:04
  SPEAKER_01: Insights, we're looking at efficiencies.
• 13:07
  SPEAKER_01: But those two combined give freedom for innovation and new possibilities and new capabilities.
• 13:16
  SPEAKER_01: And when you're looking at the businesses we work with, and obviously it's brought, every day is different, right, for you.
• 13:24
  SPEAKER_01: It's you might have marketing, you might have a retail client, you might have a manufacturing client, you might have a construction client the next day.
• 13:30
  SPEAKER_01: But there's always common themes in what they're trying to achieve.
• 13:33
  SPEAKER_01: What are the most common that you're starting to see that you think, yeah, that that's pretty much what most people are trying to do now?
• 13:39
  SPEAKER_00: I think the thing that most people start with is efficiency.
• 13:42
  SPEAKER_00: So, what are, you know, how can we optimize processes?
• 13:44
  SPEAKER_00: How can we make people more efficient?
• 13:45
  SPEAKER_00: How can we do more with less?
• 13:47
  SPEAKER_00: Um, there's obviously lots of cost pressure just generally in businesses and in organizations today.
• 13:52
  SPEAKER_00: Um, and everybody's looking for a ways to maintain that margin.
• 13:55
  SPEAKER_00: So that's that's often what business leaders come to us and that's say what they're looking for.
• 13:59
  SPEAKER_00: I think the question that we try and play back to them is okay, we we, you know, we can we can achieve some efficiencies in these areas.
• 14:05
  SPEAKER_00: Um, what do you want to do with that time?
• 14:08
  SPEAKER_00: What does that efficiency unlock for you?
• 14:10
  SPEAKER_00: What does that enable your team to do?
• 14:12
  SPEAKER_00: What does that enable your leadership to do?
• 14:14
  SPEAKER_00: What does that enable you to do more of as a business?
• 14:17
  SPEAKER_00: Um, because, you know, and that's I think when it gets really exciting, especially for the teams that we work with, they can start to kind of have a little bit more control over what their ideal week might look like.
• 14:27
  SPEAKER_00: And if we can save them 20, 40% in these kind of admin tasks or, or, you know, maybe more advanced cognitive tasks that we're that that they're currently spending lots of time on, then that's when you can see that they're kind of whole demeanor changes.
• 14:39
  SPEAKER_00: And if there is maybe some skepticism or some nervousness about running these AI transformation projects, a lot of that melts away when you start to play back, okay, think about now what you can do with all of this time.
• 14:50
  SPEAKER_00: Like what bits of your, it could be personal, professional development that they that they can get more time on.
• 14:54
  SPEAKER_00: It might be they need to get out of the office more and, you know, go and meet clients and spend more time, you know, out in the field.
• 15:00
  SPEAKER_00: Um, all of these kinds of things become a genuine possibility on quite short time frames.
• 15:05
  SPEAKER_00: And, you know, business leaders, well, for the team, that's obviously great because they're getting to do more of the things they love, they get into build their skills and capability, they get to grow their network, whatever it might be that is the ultimate aim.
• 15:16
  SPEAKER_00: Um for those team leaders, they've got happier people, they've got people less chained to their desks, you know, they're not, they're not getting frustrated by processes and moving pixels around a screen and data from one place to another.
• 15:28
  SPEAKER_00: Um, and then the leadership, could you know, ultimately the company leadership is driving towards doing more of what they want the organization to do.
• 15:35
  SPEAKER_00: So it's it's it's it's a win at kind of every level.
• 15:38
  SPEAKER_01: Yeah, I I think it promotes creativity as well.
• 15:41
  SPEAKER_01: Like you have more time to think, you have time to think outside of the constraints of what you're having to do day to day just to get your job done.
• 15:49
  SPEAKER_01: Yeah.
• 15:49
  SPEAKER_01: And all of a sudden, people who might have thought their business was heading in one direction suddenly end up going another direction, or having another arm to that business and they've got they unleash these new capabilities.
• 16:02
  SPEAKER_01: And what do you think might be?
• 16:06
  SPEAKER_01: I like to try and cover good and bad.
• 16:08
  SPEAKER_01: What do you think might be a pitfall of getting distracted by AI?
• 16:11
  SPEAKER_01: Because that's another side.
• 16:12
  SPEAKER_01: You can go, I can be creative, I can do these things, I can go and do this.
• 16:15
  SPEAKER_01: Yeah.
• 16:15
  SPEAKER_01: But what what potential problems do you have if that happens?
• 16:18
  SPEAKER_00: Yeah, I think there's just having lots of people not acting in a coordinated way.
• 16:23
  SPEAKER_00: I think, you know, you talk to IT leaders as we do, obviously.
• 16:27
  SPEAKER_00: And, you know, three to five years ago, nobody's interested, particularly in the technology that where they were using in their business.
• 16:34
  SPEAKER_00: Nobody had an opinion on, you know, which was the right cloud services provider to use.
• 16:39
  SPEAKER_00: They they were just, you know, that was the remit of the IT team and they they got what they were given.
• 16:44
  SPEAKER_00: Now everybody's got an opinion.
• 16:46
  SPEAKER_00: Now anyone can log onto LinkedIn or TikTok and find a, you know, the new latest and greatest use case for a model or something fun and interesting that somebody else has done and that they're sharing.
• 16:56
  SPEAKER_00: So you've got lots of people now who are much more engaged generally, I think, with technology.
• 17:02
  SPEAKER_00: And the good side is that that can be creative and they can get ideas, but the bad side is that can get out of hand pretty quickly.
• 17:07
  SPEAKER_00: And actually, you can get lots of shadow usage where people are not conforming to best practices, they're not conforming to um, you know, their company policies around data security, GDPR, um, just general cybersecurity, you know, you can find yourself quickly in contradiction to a lot of those things if you don't have some control around this.
• 17:24
  SPEAKER_00: And I think, you know, for the IT leaders listening to this, hopefully, you know, I think they'll recognize that now anyone can come to them with a new product request or a new idea for a feature.
• 17:35
  SPEAKER_00: And while that, I don't think anyone wants to discourage that kind of engagement from their general staff, that can get difficult to handle, I expect.
• 17:42
  SPEAKER_00: And, you know, not every idea is going to be the right one.
• 17:45
  SPEAKER_00: And not every idea, you know, it might work for one specific instance or for that what person's one view of the organization, but it's not going to work for the organization as a whole.
• 17:53
  SPEAKER_00: So I think just kind of a challenge I think for IT leaders in particular is not dampening down that enthusiasm and the engagement that they're seeing from teams, but also, you know, not drowning in those requests or action in every single idea that somebody has come to them because it, you know, you do need to take a step back and think of the organization as a whole.
• 18:11
  SPEAKER_01: I think when it comes to innovation, it comes with potential risk and potential distraction.
• 18:17
  SPEAKER_01: Yeah.
• 18:18
  SPEAKER_01: It's both.
• 18:18
  SPEAKER_01: Yeah.
• 18:19
  SPEAKER_01: It can drive, it can drive you in a bad direction as quickly as it can drive you in a good direction.
• 18:23
  SPEAKER_01: But it shouldn't be starvable, but it should be controlled.
• 18:25
  SPEAKER_01: Definitely.
• 18:25
  SPEAKER_00: Exactly.
• 18:26
  SPEAKER_00: I think one of the most powerful things you can do is encourage people.
• 18:28
  SPEAKER_00: So one of the best ways I think to learn about what a tool can do and what it can't do, or like a system or a process, is apply to something that you know really well.
• 18:37
  SPEAKER_00: So most of us in our jobs will have lots of areas that we know really well, and that's fine.
• 18:40
  SPEAKER_00: But a kind of lower risk way of doing it without falling into the pitfalls that we just discussed are doing it in for your personal life.
• 18:47
  SPEAKER_00: So if you're you know, love going on holiday and you can book amazing holidays, experiment with building an AI process that basically does holiday research for you or whatever it is that you're interested in, trainers or cars or watching, you know, whatever it might be.
• 18:59
  SPEAKER_00: And if you're a kind of a bit of an expert or amateur expert in that area or you know what you really like, you'll really, really quickly see where that process works really well and where it doesn't work really well.
• 19:09
  SPEAKER_00: But again, it's in a kind of safe environment because it's just related to your personal interests.
• 19:13
  SPEAKER_00: That kind of thing can be really powerful.
• 19:15
  SPEAKER_00: You can learn very quickly, kind of, you know, what works well, what doesn't, and then how you optimize and refine that kind of process into something.
• 19:22
  SPEAKER_00: Um, but yeah, I think the the experimenting and innovation should be encouraged.
• 19:28
  SPEAKER_00: As you say, with any new technology and with any new kind of technological change, there's always good bits and bad elements.
• 19:35
  SPEAKER_00: The trick is just to try and minimize specifically the the larger and more kind of existential risks, um, but without kind of killing that innovation spirit, without killing the kind of enthusiasm that people have to experiment and adopt these new technologies.
• 19:48
  SPEAKER_01: Yeah, I think that's a great point.
• 19:49
  SPEAKER_01: It it's one of those things as well where there's a mentality of one shot.
• 19:55
  SPEAKER_01: So let me just put this into a prompt that it's gonna do everything I can.
• 19:58
  SPEAKER_01: If you work on something you know, you are gonna do it step by step because naturally you know the order and the process.
• 20:04
  SPEAKER_01: And that's the way to think about even building anything with AI.
• 20:07
  SPEAKER_01: Don't try and get it to do it all for you.
• 20:09
  SPEAKER_01: Yeah.
• 20:09
  SPEAKER_01: Because it's gonna do it in the way that it thinks is best practice, but that might not be your best practice and it might not have the actual value that you have, which is the knowledge of how to do things well, right?
• 20:18
  SPEAKER_00: Yeah, exactly exactly.
• 20:19
  SPEAKER_00: And and it's also much harder to kind of troubleshoot.
• 20:22
  SPEAKER_00: So if you're getting it to run effectively a multi-step process in one shot, how do you know which bit worked well and which bit didn't?
• 20:28
  SPEAKER_00: You don't.
• 20:29
  SPEAKER_00: If you can break it down and say, okay, that first bit that we did was really on the money, the great output there, you know, then you don't have to spend time looking at that part of the process and refining it.
• 20:38
  SPEAKER_00: You can essentially more readily surface its weaknesses or surface its limitations.
• 20:43
  SPEAKER_00: And you can probably much quicker to troubleshoot it as a result of that because you say, okay, I can see what it did there, misinterpreted, or I didn't explain myself correctly, or I didn't give it the right data in the right format in order to make that, you know, it produce that output or that artifact that is then going to follow in the next start of the process.
• 21:00
  SPEAKER_00: So yeah, doing those things is those those elements are really important.
• 21:03
  SPEAKER_00: And I think that's what we try and do a lot at Configure is try and educate people around because we've done, you know, so many implementations across so many organizations, across so many different sectors with lots of different people with lots of different objectives.
• 21:15
  SPEAKER_00: We can bring all of that experience to bear and and and kind of help help guide the that that that kind of way of approaching it.
• 21:22
  SPEAKER_00: And like I say, a lot of them can be really small tweaks and adjustments, can have a really big impact.
• 21:27
  SPEAKER_00: And like, you know, the the key is always driving towards value.
• 21:30
  SPEAKER_01: Yeah.
• 21:31
  SPEAKER_01: And people hear terms, and we try and simplify any terminologies and things, but you'll hear terms like agentic engineering.
• 21:37
  SPEAKER_01: That in a nutshell is agentic engineering as the simplest form.
• 21:41
  SPEAKER_01: It's yeah, do it as a step-by-step agent-based process that isn't trying to do everything at once.
• 21:47
  SPEAKER_01: Yeah.
• 21:47
  SPEAKER_01: Or having agents run autonomously and do all these things.
• 21:51
  SPEAKER_01: So that's really interesting.
• 21:52
  SPEAKER_01: Thank you, Oliver.
• 21:53
  SPEAKER_01: That is brilliant.
• 21:54
  SPEAKER_01: And we ask everyone who comes onto the podcast to either recommend a book, a podcast, you can do either or both.
• 22:02
  SPEAKER_01: One tool and one piece of advice.
• 22:06
  SPEAKER_00: Um a book I'll recommend.
• 22:10
  SPEAKER_00: I mean, it's very much in keeping with this kind of conversation, I think, because um we talked about it just before we we started started recording.
• 22:18
  SPEAKER_00: It's an old one, but a good one.
• 22:20
  SPEAKER_00: And it's the seven habits of successful people, which is a I think a timeless kind of blueprint for, you know, how well, as the title suggests, how to how to be successful.
• 22:31
  SPEAKER_00: But it is the interesting thing is it focuses on individual behaviors and individual attitudes.
• 22:36
  SPEAKER_00: And I think, yeah, it's a pretty it's a pretty timeless book.
• 22:39
  SPEAKER_00: Um podcast, what am I listening to a lot at the moment?
• 22:44
  SPEAKER_00: Uh oh, I'm actually listening.
• 22:45
  SPEAKER_00: Uh what podcast I recommend is um Amol Rajan um from the BBC, and it's called Radical, and lots of very, very interesting um kind of AI and technology talk on this.
• 22:58
  SPEAKER_00: So if you if you like into you know hearing kind of thoughtful interviews from real leaders in the space, then it's it's a it's a really good one for that.
• 23:06
  SPEAKER_00: So that's an excellent podcast.
• 23:08
  SPEAKER_00: What are the other things?
• 23:09
  SPEAKER_01: Uh so then we have uh one tool to start there.
• 23:13
  SPEAKER_01: What would be one tool?
• 23:15
  SPEAKER_00: Um probably very, very as simple a tool as you can get uh a checklist.
• 23:20
  SPEAKER_00: So just having the things that you need to do in a day written down so you can tick them off is one of the most underrated and satisfying as well.
• 23:29
  SPEAKER_00: Yes, yeah, massively, yeah.
• 23:30
  SPEAKER_00: I I yeah, I I wouldn't I wouldn't know where to start if I didn't have a checklist.
• 23:34
  SPEAKER_01: I like that.
• 23:34
  SPEAKER_01: Um, one piece of advice for anyone right now.
• 23:37
  SPEAKER_00: Uh I'll keep it related to AI.
• 23:40
  SPEAKER_00: Um, don't be overawed by the kind of scale of change.
• 23:45
  SPEAKER_00: Don't be overawed by all of the hype and all of the noise that you hear in the media.
• 23:49
  SPEAKER_00: Um it's a technology that can be very, very transformational and can be highly valuable to all of us as individuals, organizations, countries, economies.
• 24:00
  SPEAKER_00: Um but yeah, just try and listen to people who are who are knowledgeable about the space and and try and cut out all of the additional kind of external hype good and bad.
• 24:14
  SPEAKER_00: Um and yeah, sounds good.
• 24:18
  SPEAKER_00: Practical advice to staying calm in the AI case.
• 24:21
  SPEAKER_01: Of that.
• 24:21
  SPEAKER_01: And I've got you an early wedding present because you're getting married this year, obviously.
• 24:25
  SPEAKER_01: I am.
• 24:25
  SPEAKER_01: And everyone who comes on the podcast has a book.
• 24:28
  SPEAKER_01: And in line with what you mentioned earlier, doing a lot more networking now.
• 24:34
  SPEAKER_01: So I've got you a body language, it's what you say, what you don't say that matters.
• 24:40
  SPEAKER_01: Oh, excellent.
• 24:40
  SPEAKER_01: Okay.
• 24:41
  SPEAKER_01: So I thought this would be helpful in your new networking era that you've been invited to.
• 24:45
  SPEAKER_01: Thank you.
• 24:46
  SPEAKER_01: And uh thank you very much for having uh thank you very much for attending, Oliver.
• 24:50
  SPEAKER_01: And thank you very much for listening.
• 24:51
  SPEAKER_01: See to you all soon.