What Businesses Get Wrong About AI with Alan Leigh

What does AI actually look like inside real businesses today? In this episode, we sit down with Alan Leigh, CEO of Experieco, to unpack how AI is showing up across complex organisations in New Zealand and Australia. Drawing on hands-on experience, Alan shares a perspective that cuts through the noise and challenges how many businesses are thinking about AI right now.

Alan explores the gap between expectation and reality, and why many organisations aren’t seeing the outcomes they anticipated. It also touches on how the rapid evolution of AI is reshaping what’s possible, particularly for businesses looking to build, scale, and compete in new ways.

For more B2B insights, ideas and opportunities, head to www.blueoceanagency.co.nz

Subscribe
When you subscribe to B2B Inspired, you're playing a key role in growing and supporting New Zealand's B2B Marketing Community.

Share Your Feedback
Got something to say? We're all ears. Your voice is what powers this community – it can't grow without you. 

Connect with Us
LinkedIn: https://www.linkedin.com/company/blueocean-agency/ 
Website: https://www.blueoceanagency.co.nz/podcast/
YouTube: https://www.youtube.com/@WeDoB2B

Let’s roll up our sleeves and take on tomorrow together.

Transcript

Welcome To B2B Inspired

SPEAKER_00 0:03

Yola and welcome back to B2B Inspired, Cuts by Blue Ocean, where we unpick the ins out ups and downs of all things business to business here in Nintendo and New Zealand. From emoting trends and thinking to the inspiring real-world stories and experiences of inspiring good people from within our ecosystem, we're here to help the New Zealand B2B community to become one of the best, boldest, and brightest anywhere in the world. Now, if like me you live and breathe all things business to business and you're looking for a place to connect, learn and be inspired, you have come to the right place. Now let's join Henrik Allen over in the studio.

SPEAKER_01 0:42

Hello and welcome back to Blue Ocean, the B2B Inspired Podcast. My name's Henrik Allen, and I'm here today with Adam Lee, CEO from Expericolf, one of the leading custom technology companies in New Zealand, who's going to talk about uh AI in New Zealand from an insider perspective. They've had some really great success stories already. Uh, they know where it works and where it doesn't work and where it can go wrong and where it can go terrible wrong, but also where it can be a success and really make a big difference for businesses. One of their big success stories was for debt work, which achieved a 50% productivity increase within three months, and a 35% cost in operating expenses. So, Alan, welcome to the podcast. Thanks. That was a reduction in cost, right? Yes, yes. How did I not say that? Reduction in cost, 35%. Very, very impressive. Yeah. So tell us a little bit about AI uh in New Zealand uh today. Where do you see it really working uh and being a success?

SPEAKER_02 1:44

Well it's a really big topic, right? So uh AI is showing up in all sorts of places, showing up in coding, generative uh AI for coding. It's showing up in workflows, agentic AI, uh and the intelligence to improve workflows. Uh and um and it's showing up in everyday work and the productivity tools that we use. So uh we'll talk about some specifics, um, but what perhaps is an umbrella comment is it's maturing all the time. So i irrespective of those three scenarios, you're you're always getting improvements at a very rapid rate at the moment. Where we are today versus where we were six months ago or twelve months ago is very different. And uh it's it's improving uh to be a really useful tool in all of those three scenarios.

SPEAKER_01 2:30

And you work with a lot of businesses that are quite complex. Um with, you know, some uh heavy technology demands, uh integrations, complex systems. Yeah.

SPEAKER_02 2:41

Um is it worth me sort of talking to that in just to give context? Yes. So we specialize as a business in well, we specialize in ugly is what we say, which is the back-end systems that businesses need to run effectively. And we're talking about y typically very large businesses. So an example would be New Zealand Automobile Association or a Downers or um Over the Ditch if you go into Sydney, uh NRMA motoring, where they're massive systems that are critical for them to run their business and uh and usually hundreds, if not thousands, of users and uh and and very critical for the success of those businesses. So that's what we specialize in. We don't specialize in building a mobile app. Uh you know, we we specialize in the bigger, more uh um complex systems that are necessary for these larger, medium to large businesses to run.

SPEAKER_01 3:36

So I would imagine AI is quite different uh within those businesses and those workflows, uh and different from what you know most people they s know AI from LLM models. Uh yeah. Uh so just talk a little bit what what an LLM model is and how you know where you see AI and where you apply that within these businesses going to test my knowledge, right? Yes, I am.

Why Most AI Pilots Stall

SPEAKER_02 4:00

Um I think for the nature of the context of what I gave you earlier, these are complex businesses. So they're going to be a lot more cautious, and we're talking about systems that are necessary for them to run. So the where we're seeing um AI used in their systems is very different to where we're seeing it used, perhaps where it's less critical in their day-to-day productivity. But um we'll start with the systems. So the the systems, you know, the w what we're seeing is we're seeing a very large up well last year and until now we've seen a very large uptake in pilots and POCs, where people were testing out what could these technologies do. That example you gave with uh networks at the start, that was um there was a six-month pilot for that where people were testing and monitoring to see if AI could improve some of their workflows. And uh I'll I'll talk to uh the specifics of that later, but that you know, th there's a lot of caution when a business takes on those technologies and we're seeing certainly for the last uh twelve um months there's been a huge uptake. It would be around 95% of businesses that have dabbled in to some degree of using AI to improve processes and their back-end systems. Most of that hasn't necessarily gone into production, meaning they're actually using it in anger to deliver better services or operate more efficiently. So it's all really just a hype? They're investing. Yeah, it's a toe in the water.

SPEAKER_01 5:31

And they're playing around, but they're not yet seeing any productivity gains.

SPEAKER_02 5:35

Well, a lot of them have seen it um put a toe in the water and then that toe in the water is measuring has this improved anything? So there's their their their benchmark is what they're doing today, status quo. And they're running a parallel process or a or a smaller team actually putting it into uh practice just to sort of test and validate is it is it doing something wildly different?

SPEAKER_01 5:56

So that sounds like there's a very low level of trust in caution.

SPEAKER_02 6:02

Yeah, caution. So I mean the l where we're seeing it um for for adoption and systems, you know, like you say, that these are very critical systems for businesses to operate. So they're they're showing a lot of caution from the perspective that they can't just put it into action in hope. They've got to make sure that they um they have a measurable understanding of how it's improved the business system. And in a lot of instances that hasn't happened. So and uh you know a lot of these businesses have written it that off as well as hype. It's not actually delivering the value I thought it would. But one of the challenges we'll get to later is um is quite often they're using AI wrong. It's not necessarily a technology problem, it's an organizational problem.

SPEAKER_01 6:48

Oh well, talk a little bit more about that. What does that mean? What do you mean by by by that? That's that's how they're using the AI.

SPEAKER_02 6:55

Like uh often um a lot of businesses have used it in a way where they're just using it to augment an existing human process or hu or automated process to see if they can enhance it.

SPEAKER_01 7:07

So they're trying to make AI work as a human or replace a human. Is that steps in a process, yeah.

SPEAKER_02 7:13

But what they haven't explored, and this is again caution, it's not a lack of trust, but uh you know jumping to the place where they're fully or orchestrating or engineering a process completely differently to what they have today to l fully harness AI is a leap some of some businesses have not been prepared to take. So they're not y you're not in return getting or yielding the the benefits of potentially doing that. And is that the way uh the the direction they should go uh in the Yeah, yeah, some some should. Um it's all with caution and like anything, no one none of these businesses are gonna put anything into into their business where it's a leap of faith. So um but in parallel you've got to realise a lot of this technology is is evolving rapidly, so the knowledge of what how to use it is evolving as well, and sometimes not as rapidly as the technology. So there's definitely those that are getting their heads around it and leading early. Um and then there's those businesses uh and individuals within those businesses that are a bit more cautious.

A Debt Collection AI Win

SPEAKER_01 8:20

So there's a lot of caution. But talk to me a little bit about where where you have seen it uh AI implemented today in New Zealand's uh businesses where it's worked well and where it's made a difference. Uh what what are kind of like the parameters around that and how do they go around that making it a success?

SPEAKER_02 8:38

Yeah, I think um uh this is AI in in as a uh in a process, not a um I've talked to AI and coding separately.

SPEAKER_01 8:47

Yep.

Deterministic Vs Probabilistic Work

SPEAKER_02 8:48

But AI in a process and in systems. Um the example you gave with um uh we had deployed uh well probably fifteen months ago now in in uh deadworks. That was using AI to do effectively better categorization and p p part of uh of a workflow, categorization that a human had done. So they were effectively uh debtworks, so they were taking on a lot of um uh information I have to be careful here that I don't sort of breach any confidentiality uh and reveal their special source. But th they were effectively dealing with a lot of data points, and with that data um humans were interpreting it and taking action. Some of those actions or responses might have been an email or uh an a trigger to do some other action. And when you were uh in their instance, uh a debt collection company, they were taking on clients that you know might have 30,000 customers that they have to follow up. There is a l vast amount of information and data points that they've got to interpret and take action. And what AI, we were able to build a system that enabled them to do take a lot of that heavy lift out and remove the uh um as much manual as possible, but still keep the human in the loop where it was critical and really important. And you know, example of that, where they were able to take on um you know 50% more clients without adding one more service agent to their call centre. Oh, that's impressive. Yeah. So, you know, that that is and you know that translates uh for anyone anyone who's not a business person, translates to good bottom line growth. It certainly does, yeah. 50% increase in productivity. Yeah, yeah. So, you know, that's that is a great use of AI uh in in a workflow that where they didn't remove any cost from um headcount, but removed the cost of needing additional headcount to get to s to uh support that growth. So that was a fantastic use case. Other use cases we've seen it um really good uh I was talking with you earlier around you know deterministic versus probabilistic. Yeah, so just talk to me about that. What does that mean? What are the differences here? Deterministic is where if you have a absolute outcome that you need. A good example is uh in a finance system, your um profit and loss for the end of the month. That is a um a determin the the variables that make that up are a deterministic. You don't want it to make up the numbers. If it's a hundred dollars, it's a hundred dollars. You don't roll the dice. Yeah, yeah, yeah. So that's deterministic. Probabilistic is where there could be some you know some variance in the like the data in and the data out.

SPEAKER_01 11:31

Yes. So a bit like marketing. So uh uh B2B marketing, uh where most of our audience are setting, they know a lot about that. Yes.

SPEAKER_02 11:39

So for deterministic systems, you're not necessarily getting the big uptake of of AI, or if it is used, it's augmenting a s very structured, rigid guardrails of of of code. Yes. And or humans in the loop. Where you have a process which is more problem probabilistic, that's you could go jump in fully, you know, and and uh uh use use AI for that, uh no problem. The consequences aren't the same. So and this is what we're seeing at the moment. That will likely change more and more as people's confidence to go up where if for the deterministic scenarios where people want a definitive outcome. They don't want hallucinations or variations in you know asking the same question of AO and it gives you three different answers.

SPEAKER_01 12:27

Yeah.

SPEAKER_02 12:28

But those systems you can't have that.

SPEAKER_01 12:29

No.

SPEAKER_02 12:30

So But as that matures and the the way uh um um um the use cases are are better understood and how that can augment humans and code, that will be used more and more in those systems as well. ERP. You know, a good example where in a um a finance system, an ERP system. But to today if you do a customization, it's it's coding. Yes. It can be configuration as well in the system. But you could use AI in the future where it's within the system and it's helping you you're prompting it to make some changes and configurations and it's presenting those to you and you have to approve them. That's a good use case of the Trevor Burrus.

SPEAKER_01 13:09

So it been more comes more of a conversation instead of uh technical knowledge, perhaps. Yeah. Yeah.

SPEAKER_02 13:17

So um so there's a whole lot of nuance to how this could be adopted and used. But we are seeing more and more there's certainly lots of people using the technologies. But the most important part is success cases. And the success cases are you know haven't been there's not a lot of poster childs. You know, there's not a lot of people showing uh um you know really great use cases. I've been to lots of conferences around these technologies, and uh that's probably the m no most notable thing is you go, some of the massive global companies even that you're uh that may be talking at these events, and these are global technology events, you'd expect them to reel off dozens of use cases where they've made massive, significant improvements for a business, and that was notably absent.

SPEAKER_01 14:03

Okay. That's very interesting. So so everyone's investing but not really seeing any RI yet. Is the yet there or where are we at?

SPEAKER_02 14:13

That that that comes back to what I was saying, but I don't think it's a technology problem. I think it's an organizational problem. Because the people that are implementing some of these solutions are doing it in the same way that they thought yesterday about solving the solution. And the the caution to uh use AI fully uh in the right processes has has been evident. So I think that will be a learning curve where people will start to test and validate them more uh than they have. So a lot of the pilots and test cases that uh um companies have been using, they were augmenting what they did uh um and maybe trying to use AI to automate a process in the same way that they may have done it if they'd coded it. Um it's all very nuanced. But they hadn't seen great improvements. And and a couple of things are happening. It's all moving at once. The technology itself has been improving.

SPEAKER_01 15:07

Yeah.

SPEAKER_02 15:07

As well as people's knowledge on how to apply it. So you're now starting to see there there was always some use cases like the one I, you know, we've referenced that we did some time ago with Deadworks. There's you're now suddenly seeing more and more of those as people get their heads around and go, ah, this is how we could, you know, this is how we could harness it. And we are expecting, certainly in the next 12 months, to hear more and more use cases where there has been successes.

SPEAKER_01 15:34

Okay.

SPEAKER_02 15:35

So uh and and the development of AI uh is is that accelerating or don't expect uh it's um without getting too nerdy here, uh it's it's not they're not going to try um I think we're we will hit a ceiling where the amount of training the um the I was gonna say LLMs, but uh um some of the AI tools, the chatbots that sit uh that you can use, things like OpenAI or Claude, if you've heard of those names. Um those tools, there's only so much training and data sets to train them even more and more. Uh we'll be hitting a ceiling in those. Um and then it will become about how do you use that technology. And I think we're kind of starting to see that now. So we we still expect improvements, but the percentage improvements of training is not as going to be as much as what we've seen to date.

SPEAKER_01 16:32

Oh, that's very interesting. So are you kind of seeing AI as as reading an upper barrier for how smart it can become? Well the data sets, yeah, yeah.

SPEAKER_02 16:41

So there is there's always discussion around, and this is getting really nerdy, but I know, yeah. Is that you get what they call synthetic data, which is the the the AI itself is creating new data sets to learn from. But um uh at the moment that's I think some way away. And uh and you know whether that is a thing or if that gets regulated or if there's any controls, I don't uh you know, that's all to play out. But right now I think most the the most sort of real-world observations are the learning uh uh AI learning more and more from the data that's available is probably nearing its ceiling. Um and there's only so much data it can consume to learn more, and the incremental amount that it's learning is not much. But y you know, the biggest uh um uh uh perhaps value that we could harness to implement uh is implementing AI in a different way to get better results of what we have today or what we might have in the near future.

SPEAKER_01 17:41

So from your knowledge and working with New Zealand businesses, uh if I was a New Zealand business and um there's all this hype and and everyone's asking for AI and I want AI, what should be my first step? What should I if if let's say I run a reasonable size New Zealand business with uh some complex back-end software, uh, and I come to you and say, okay, uh what should I look at and how should I approach uh AI in and what would I get out of that? Would I get productivity gain? Would I get cost reductions?

SPEAKER_02 18:13

Well, yeah, so let's just start with AI. AI is that umbrella term, right? So that could just be tools for the staff to do um improve their productivity. An example of that is you know write a document better, presentation better, work with their colleagues. So they can do that today. That's not something we focus on as a business. Um and I think everyone's using AI in that context.

SPEAKER_01 18:36

I think that's pretty run-of-the-mill these days. Yes. I was at a conference yesterday and that question was asked how many of you use AI on a daily basis? And everyone had their hands up. But I think they're just using these LLM models, uh large language models, which is really just helping people, writing an email, writing a report, uh all of those things, which I think is pretty standard now. But that's not really what we're talking about when we're talking about what for us.

AI Coding Breakthroughs And Claude Code

SPEAKER_02 19:00

Yeah. For us it's it's AI in the use of coding and AI, or they call agentic AI, which is in the use of workflow and um and and and logic in a in a process. And uh those two are you know incredibly potent. So on a coding side, um there's a actually what one of the uh you and I were talking about this the other day too, which is there's these there's a um from a social media perspective, you've got a lot of hype. And with the hype, you've got a lot of commentators you know making outlandish statements about how much AI is impacting areas and and things like coding or or workflow in businesses. And uh you know, you've got to remember the source, you know, who where is where is the commentary coming from and what's the how do they benefit from that. If the benefit is clicks, then they could be making statements that are you know a little bit far-fetched or outlandish or extreme. If if it's a vendor, and I the obvious benefit is they're they're the beneficiary of monetizing what they're they're trying to drive. So you always have that in mind. What we're seeing though is is uh AI for coding has improved dramatically in the last month. And it's been around for uh you know well, since 22, 2022. And uh the in the last month the increase in its accuracy and its performance, in particular clawed code, has just skyrocketed. Wow, okay.

SPEAKER_01 20:31

And so this is really February 2026. Yeah.

SPEAKER_02 20:36

Yeah, i it really was. And it was uh i for the type of systems and the c complexity of what we're dealing with, you you need to have um you know you you you can crack code, but if it's uh what they refer to if it's useless, it's AI slop. And if you're creating AI slop, you've just got to have a human coming in and fixing it up. And there was uh for for large complex systems we were seeing productivity gains of uh ten, fifteen percent. And uh um well that's uh with uh the c uh the latest version. So uh prior to that we were seeing five to ten percent. You know, it wasn't massive. If it's a big system, uh that can still equate to some some serious dollars in time but that if you're building a big complex custom system, but it wasn't anything you get out of bed and go, waha, you know, that's that's um massive. But for smaller applications, what we call greenfield applications, something that's brand new. It's not modernizing an older system, it's just a new idea. Uh we were saying that's where you were getting m much better results. Prior to February this year, it was okay. You know, you could create something it was impressive, it looked magical, but it needed a lot of work to to still to to to make it a user app. Production ready production ready app. Um now with Claude Code have having come out and fit. February, we're seeing results there that for a lot of the smaller to medium sized apps, you can do some wonderful things and it's it's workable, you know, it's tidy, tidy code. It still needs work. It doesn't you're not just pushing a button and asking one prompt and out comes this fantastic app. Um what we what we are seeing though is the quality of what it's producing has has certainly gone up. And that means you can spend a lot of time more on the creativity of like the problem you're trying to solve and you know really sort of focused on engineering a good solution, prompt engineering, uh uh and less about the coding. Okay. And we've seen some really good uh uh results in that space. The big thing for businesses though, that's actually changing the market. It's it's making what was yesterday unaffordable in a custom app affordable. So uh you know, you you could have something that was a hundred thousand dollars for an equivalent a year ago, two years ago, and now, you know, that that same application enhanced, so even more uh better than what uh what you'd first envisaged, you could have for you know ten, twenty thousand dollars.

SPEAKER_01 23:12

Wow. Yeah. So about one third. It's not a thousand dollars. No, no, no. But it it's one third the price, and I would imagine also it comes with a shorter time frame.

SPEAKER_02 23:21

Uh that's not a general rule, because it will apply to you know the type of app and the complexity you're building. Yes. But um, yeah, it's impressive. So what we've seen there, you know, there's there's a lot of applications that businesses have wanted, and they've either gone and got a SaaS tool because the cost of building the the app they really want, a custom app, just they couldn't afford. Or they they couldn't justify from a return on investment. And um with that dynamic it's changing that completely. Wow.

SPEAKER_01 23:51

And these are production ready apps, uh security, that cost is coming takes into account all of that.

SPEAKER_02 23:58

Yeah. Rather than just the time, that's a really good point. It's not just the engineer prompting and developing an app, it's uh it's putting the polish on it so you could use it in production. Yes.

SPEAKER_01 24:09

And you have documentation for support and further development and changes uh and ongoing support.

SPEAKER_02 24:16

And it's tested for security and and the code is maintainable. Yep.

SPEAKER_03 24:21

Yeah.

SPEAKER_02 24:21

So doing all that. And sometimes you know you need a um design changes. Uh so all of that takes into that into account. W what we're kind of seeing there though, is that's making when it was just custom applications in the traditional sense without AI coding, you the price was prohibitive. That meant that was only a certain part of the market, you know, very large businesses. Typically we'd say over a hundred thousand turn a hundred million turnover that could afford that.

SPEAKER_03 24:50

Mm-hmm.

SPEAKER_02 24:50

Where their problems were big enough to warrant that investment. With these changes, we're seeing that uh businesses is, you know, with ten, twenty million turnover. You it could be lower, but for the area that we specialize in back-end systems, that's you know, we're seeing businesses of that size where it's now affordable. And uh, you know, they can they they can and they're throwing out the sort of weak SaaS solution they may have had and uh and replacing it with uh AI coded custom application, the thing that they really wanted. So and that's that's not a um hype, that's happening today. Wow. Yeah. So we're at that end of the market we're seeing some remarkable it's opened up a new market. Uh it's certainly our visibility is New Zealand and Australia, and we're seeing um businesses ten million turnover and above, you know, removing SaaS solutions out of their business and coding, you know, getting new solutions in. We're seeing them do things that they never they didn't have an application for, or they operated on spreadsheets and wanted a proper application for. All of that is quite exciting because they're they're starting to do more and more and more.

SPEAKER_01 26:02

So the market is growing exponentially. Aaron Powell, Jr. Wow. And these are custom developed at using AI that effectively the businesses own, and is their custom IP that can give them a competitive advantage either for internal use, benefit the staff, or or possibly even, you know, APIs uh and integrations. All of that.

Cheaper Custom Apps Change The Market

SPEAKER_02 26:22

Well, uh this is where you would come to a business like ours, is uh you know the I actually I I I've seen businesses do this. The the complexity of coding for an application like the applications I'm talking about there is perceptually gone a w away. You know, people feel like they can prompt uh in their coding in English, because they're just asking in a prompt.

SPEAKER_01 26:46

So I can sit down and do my own little prompt. What if I did that for for my own business and uh did a little prompt and built my own little app? But what uh uh what could you foresee any problems with it?

SPEAKER_02 26:54

Yeah, yeah, yeah, exactly. So typically that you you prompt something and out, you know, you create an app. First problem you've got is um well, where is it going to sit? And there's a laptop. Yeah, exactly. On your laptop. So, you know, um for a production system in a real business that's going to use that in in a secure way and wanting a robust system, it's not gonna sit on anyone's laptop, right? So it's it's about how do I get that to a server? Um, how do I test that it's secure? How do I test that it's actually robust in its logic? And you the the the tools enable you to code faster and create something, but often people don't have the expertise to do the rest that make it robust. So we've seen scenarios already where you've got clients that are creating applications that might be about stock inventory, and they haven't test they don't know how to stress test the system in its entirety, uh, and they've got carried away with coding, and they suddenly realize, oh, okay, it's full of holes. And the logic isn't right. Uh uh so they haven't tested and an example I came across the other day was we had a um an application a client had developed themselves and it was to uh do an inventory check in real time around the um um how much stock was in the warehouse. And there was one particular item we focused on and said, well, how many have you got in stock? And they he said 1200 units. And we punched into the system with like fourteen hundred units of that, and immediately the the application said, Yep, no problem. That's no problem. I'll take your money. Yeah. And and there wasn't. There was only the twelve hundred. So it that's just a really light example, but that's what we see happening as well, where the new problem I think businesses will face in that end of town where they're creating their uh apps is how do we make sure they're robust? How do we make sure we're not just creating other problems in the business in my customer experiences? You know, we're we're trying to create these apps to deliver really good customer experiences in that instance to their supply chain. But it could be end clients and uh and they they fail. And then then and then unaware of how to look back through the code and fix the problem. So uh they can prompt it, but often they don't actually understand what the problem is to begin with. That that was a really simple example. They can be a lot more complex and they'll be a lot more obscure. Trevor Burrus, Jr.

SPEAKER_01 29:11

So you still need the right human to do the right prompts and you still need to follow through the documentation, security checks, integration, and all of that, and and and enable uh these are not small businesses, but to have that continuity so that they can evolve and develop that app over time.

DIY App Risks Security And Robustness

SPEAKER_02 29:26

Yeah, that and what we're kind of we're expecting is that that's just gonna go boom. You know, like they're they're going to get a lot of apps they're building and uh to solve all sorts of problems because the the marginal cost of creating the app is lower. But the new problem is I yesterday I had zero apps, now I've got ten, twenty. And you know, how what do I how do I make sure they're safe, they're secure, you know, um then open to hacking, how do I make sure that they work as they should? Uh suddenly the level of complexity has gone from this was just a manual process to people not actually understanding what the process is that sits behind the code in the system. So that's where we see things will start to shift. What what we're already helping some uh clients with is it's not so much the coding at the um at the smaller end of the market, it's knowing what to code.

SPEAKER_01 30:20

Oh yeah. So the discovery, understanding the needs, understanding the flow, how it fits in.

SPEAKER_02 30:25

Yeah. So if you if you weren't if you didn't think uh like uh somebody who designed and built apps yesterday and suddenly you know you're prompting a few things, you might get uh the shell of a good system out. But um what we've found is we've had to do what we call SaaS rescue, uh or or not SaaS rescue. A new product vibe. Yeah, yeah. Yeah, which has come in and and and help clients you know just understand that in part that what they built was was robust and good. But the bit there was uh perhaps a number of bits that didn't quite work as they should. And uh uh you know we can see if you're building more and more apps, that's just going to be more of that. So um already we're talking to a few clients around just a continuous development where you for a monthly fee just help them now they might be creating an app, but we're helping them make sure that's robust, it's hardened and ready to be used in pr production.

SPEAKER_01 31:21

Okay, so they kind of get to play around with to kind of build their own little add-on to and changes, and then you come in uh and ensure that it's fully documented, secure, and it it works with the workflow and you can't business system. You can't order fourteen hundred dollars off the product that's already twelve hundred dollars off uh uh uh in in stock.

SPEAKER_02 31:42

Yeah.

SPEAKER_01 31:42

Yeah, yeah, yeah.

SPEAKER_02 31:43

So yeah, we can help them out with that. That gives them uh you you know, we absolutely see a scenario where some clients will vehemently want to do their own and create um create their own apps. Other clients will come to a place where they they go, you know, why don't we just give this to the specialists? It's not about creating the code. It's about what what are we designing and what's the output going to really be. You know, this is this is a really good system, sorry, a critical system. Do we want our best efforts and and winging it, or do we want a really nicely designed system?

SPEAKER_01 32:16

Yeah. Yeah, so you talked about this, these are the Greenfield apps, so the little add-on apps that just sits on top, uh relatively uh isolated, uh started from scratch, and clawed code really makes a big difference there. So that all happened in February 2026. Yes. Um how are the AI uh software development tools and generally evolving this year? Uh what are you what are you seeing there? Over the next 12 months.

SPEAKER_02 32:42

Um we'd expect more improvements in uh productivity. So if you were to talk in sort of general terms, if you could say for a Greenfield app, um you might be getting amazing productivity gains. It could be 70, 80 percent productivity gains. Part of that is because it's creating code that um and there's there's no other complexities. You you you don't have other complex integrations or some complex unique logic or some complex um user d interface designs. You know, so uh you can get enormous productivity gains and feel like it's just magic. Like Greenfield, yes. But if you're looking at some of the more complex uh as soon as you go up on that scale of complexity, yeah, that's when a human is still really good in the mix. So what you know, AI doesn't solve that. So uh critical thinking, creative thinking, uh those skills are very human. And the AI is not as anywhere near as strong as a human in that. So what we're finding is for smaller systems, we do it t-shirt sizes, so a small t-shirt and a greenfield app. And small t-shirt meaning not uh that complex. Uh an example would be for somebody who's n not into building systems, so an example would be a forms server, a survey sort of tool, or a a basic portal um that that didn't do anything complex. Or uh um an app that might be used for some very simple use cases within your business. Uh and those sorts of apps you can build really well low cost and and uh AI does the majority of the heavy lift. If you go up to the next t-shirt size, a medium-sized t-shirt, m my size. Um you you know, you you quickly get into there's generally complexities. And the complexities might be there's you know elements of unique design required that uh need a human critical thinking and uh to design. It could be some complex integrations with some other systems and and outside of that application that have to be considered. Uh I'm just giving examples, it's not limited to these things. But there are def elements that effectively start to introduce complexity. And rather than that 70-80% productivity that you get for the little smaller apps, they don't have to be small, but the the analogy. Um the uh the medium-sized ones will start to, you know, you might drop to 40% productivity gains. These are all guidelines, they're not hard to do. Well that's still better than the five to ten. Yeah, yeah, yeah. Yes.

SPEAKER_01 35:17

40% is is significant.

SPEAKER_02 35:19

Yeah, yeah, yeah. So it's very material as well. Again, we see that opening up a market that didn't exist yesterday. So that's applications where no um where where businesses, medium and large, typically might have bought a platform or a product and tried to customize it.

SPEAKER_03 35:33

Yeah.

SPEAKER_02 35:34

And you suddenly say, Why? Why would you do that? Because you're trying, you know, you're you're creating a product that's uh or trying to bend a product to your will, uh, and there's always limitations to that. Whereas creating a custom application there's none.

SPEAKER_01 35:50

You get what you wanted. So you're saying those two uh prices are getting closer together?

SPEAKER_02 35:56

All of this is pushing the price down for the average marginal cost of building a custom application. So the the small t-shirt, that that really didn't exist. You know, there was some Okay. So now we cater to the little people. Yep. Yeah. So that's opened up a market. The medium-sized t-shirt stuffy there was a lot of workarounds to create medium applications, or th the the return on investment had to be substantial to warrant you building it.

SPEAKER_03 36:22

Yeah.

Productivity By App Complexity

SPEAKER_02 36:22

And um and both those markets have fundamentally changed because AI's impact is it's growing it's lowered the cost of building those apps, therefore it's growing the market. Uh both in bigger businesses where yesterday triage and return on investment meant they could afford it, but triage and return on investment meant that they didn't. You know, there was always something more important and that warranted their their their fund, their capital. And for smaller businesses they couldn't afford these and now they can. The where there's not the same impact is the much larger, more complex systems. Um so uh and this is where you, you know, I always again when we were talking that the other day, I was saying we need to be careful about the hype in the social media commentary around AI on all of like it's a ubiquitous improvement across anything and everything you do. Because it's there's definitely improvements, but it's depending on the t-shirt size, how big and complex the system is means how much improvement there is. And for big, complex legacy systems, big in my reference, is where there's lots of integrations across the business, there's lots of complex logic to perform tasks, there's lots of um, you know, really usually these are systems of what we call systems of differentiation. So they're critical systems to deliver their services and stand out from their competitors. And in those systems, yeah, humans, the critical thinking and the architecture of them, the critical thinking and the unique uh processes to create different experiences for end clients, all of that humans are really good at. And uh AI is not, excuse me. So uh not as good as. And um we're seeing productivity gains on the coding for those systems of that 10, 15. Exceptional was could be 20 or 25 percent. Um we're expecting this, your question about what's happening in the next 12 months, yeah. We're expecting that to lift, all of these to lift a little. So uh if the top end of town, the really big complex systems are getting say 15 to 25 percent today, we'd expect perhaps in the next 12 months that goes up another 10%. In the medium-sized T-shirt complexity, you we'd expect those systems to go perhaps from a uh a guideline of say 40 to 50 to maybe sixty, even seventy. And uh the smaller systems uh they're already at 70-80 percent productivity. Um, we just that we think that won't fundamentally that might improve a little bit more, but it's not as material.

Are NZ And Australia Keeping Up

SPEAKER_01 38:55

Well, those are some massive changes uh within the next 12 months and what we've already seen in the last 12 months. Yeah. So and really what we can probably look out of that is that you know, most businesses that are not looking in this area and looking at improving the flow in their businesses, and a lot of that is determined by the software they're using and the app they make available and interfaces they make available both internally as well as APIs externally and to customers, they're going to be left behind. So so your view, so you work both with New Zealand and Australian businesses, uh obviously quite a large range from you know uh mid-sized to large businesses uh and some of the smaller ones obviously as well. Um Is is New Zealand Australia uh are we up there in in regards to uh uh where we're sitting and and are we keeping up with uh trends internationally? Are we keeping up with American businesses in in in developing software?

SPEAKER_02 39:53

I I think so. Um you know there will always be uh post-to-child use cases that you hear of some child uh some business in America that's done s or Europe that's done something amazing. But putting aside you know the outliers, I think as a general rule, we're we're not that far behind. Um we are a little because there there's you know you've heard the term number eight wire, you know, which uh in New Zealand. That that's actually a good thing to some extent. There's a lot of experimentation, particularly at the smaller end uh of use cases, I'll say, because it's not necessarily small customers. It could be a big customer but a small use case and they're experimenting how they could use these technologies. We're certainly seeing a lot of that. Uh and in New Zealand businesses, we yeah, we're we're certainly seeing people aren't afraid to play around and and and do that. Um and we're definitely seeing that uh at the smaller and mid market sized clients as well. So I don't think we're behind in any way. I mean you hear all sorts of commentaries and one day it could be we are, the next we're not. But um I I do think though our uh most of our you know you've got to keep in mind most of New Zealand business is small and medium enterprise compared to the global size businesses. I know. A big business in New Zealand is probably uh slightly smaller than anyways. And and a lot of those businesses don't have the layers and complexity of global giants. Um they don't have the resources either. But I think that's actually in this instance a good thing because the friction to experiment and create is actually probably lower.

Line Managers Building Faster Pilots

SPEAKER_01 41:31

Oh yeah, and and this I mean the the cost is going down. Yeah. Uh so the barriers to entry uh is being reduced, which is great for the smaller businesses because it enables them to uh better compete with the bigger ones, which used to be able to afford this cost of uh software and technologies. So what you're really saying is that New Zealand businesses, uh leaders within those should look at coming up with the idea of what they want to do. So the opportunities are there now. The cost is going down, they should really look at well, how can I take advantage of this and really just challenge both their teams and somebody like you in saying, oh, this is my vision, yeah, I want to do this.

SPEAKER_02 42:11

Do you know what I think of um um and I can see it happening uh not as much as it should, but yesterday it was coders um that were creating applications and IT departments. And um even though it was to really fulfill on an um line of business requirements, so you might have you know managers sitting uh and various managers sitting in the in a business requ requesting it, it was a very formalized process. The ability to create ideas doing a pilot or get a something just small that brings an idea to life is is really low cost now. So what you're um seeing more of, but not as much as I'd like, that I can see New Zealand being really well positioned for this is line of business managers, marketers, um operations, head heads of operations, um even a CFO, you know, in businesses where they're not following such a formal process where it's still within a gov uh uh an agreed governance process, so they're not you know just creating things like skunk works. But it to create something that's a pilot in a test is you know low friction. And I'm hoping uh that um we certainly have offerings out to our clients um and to some defined offerings, um, which is enabling those managers to in a frictionless sort of process to ideate, come up with ideas, but coupled with people who know how to build apps to do more.

SPEAKER_01 43:45

So really just get these people to think what they want to see, write a prompt for it, get something up that kind of looks something like what they want. It may not work, but it gives an idea, uh, and they can actually take that in and say, Yeah, come to somebody like you or go. upwards in the organization and saying, look, this is what I need for me and my teams to succeed, it'll make a huge difference for for you know the value we can deliver to our customers.

SPEAKER_02 44:09

Well yesterday, because it was so expensive and bigger businesses in particular, your your ability to ideate, you know, come up with ideas and tests stress test them and and and it was just difficult because if you were a marketer and you have these ideas, oh if we had a system that did this and I could offer this to our clients, that was a really expensive process to follow. And and it was heavy in documentation and most people don't s can't interpret all the documentation, go, oh yes, I've got all the scope. I I can visualize the system.

SPEAKER_01 44:37

They just don't. Oh I don't know what would happen. The first thing you would meet as a developer they would be like, oh no, that's difficult.

A New AI Discovery Tool Example

SPEAKER_02 44:44

Well I know this to be true because I'm a marketer, not a not a developer. So you know I um and I you know know that a lot of people are visual, you know, so the this change in being up to at a lower cost create applications so the people that are visual, line of business managers can see the application actually enhances it because as soon as they see it they go, oh what if it could do this now or that? Whereas if yesterday's model it was a on a piece of paper some lines of scope, functional requirements document. And they the a lot of people just couldn't digest and interpret that and visualize it. I can see back to New Zealand business how we run with it is the ability to create things that we can then just enhance, enhance, enhance will actually give us you know the that same opportunity exists for you know every business across the globe. But I think New Zealand being um innovators and and being a largely m small and medium enterprise market, same with Australia, is that we should really run with that. And I'm hoping that in this next twelve months uh what we're starting to see is more line of business managers engaging businesses like ours uh or their own internal people if they've got that capability to ideate. You know, let's let's stress test this and you know let's create things. Uh we were talking about some of the innovative things we've seen you know being developed. Uh and uh they're things that you just wouldn't have done yesterday. An example w I was giving you is you know we've created a tool and out one of the key processes we have when we talk to clients is what we call pre-sales discovery. And that's when you're talking to a client they're going, oh this is a problem I've got and I wish I knew how to solve it. Here's the implications, here's my you know desired state, here's what I'd really like and that's a to to be useful that can't just be a an uh an arbitrary discussion. You've got to it's land the plane a little and and get that the detail. And we've built a a a application with AI embedded in it as well and we used we built the application using AI. And we built that in probably three weeks. And uh we've embedded AI to help in prompting questions along with a human to effectively extract out, well what's what's your problem, what's your desired state? And this will then produce schematics of what they've got uh in the system process design and indicative price range for yeah. And so now that is something that if we'd done that in the traditional development way only years before you wouldn't have done it. It it just it it would have been too expensive to solve the problem and you end up using just humans. And that's a that's a a use case we can sort of see you know just changing. Everyone go, oh but for those in the know like us, well you can look at that and go, well that's two, three weeks worth of work. Let's let's create the app. But there's only there's large percentages of the of the market only just catching up and understanding. And though there's lots of people that have heard the you know the um all the language of AI, you know, LLMs, agentic AI and and and in reality a lot don't know what that means yet. So the next twelve months we'll s they'll start to see what it means and experience it with people who do know. And I'd encourage anyone listen to this if you don't know that's what you want you know rather than go and study the the the terms and acronyms and it's meaningless. What you really want to understand is what can it do for me? And that's you know go to the people that know what they're doing and get an experience that you'll find that wow that just opens up the doors to what we could do.

SPEAKER_01 48:33

This sounds so exciting. I mean I'm so excited to hear all of this so really what you're saying is that the barriers to entry are being lowered. Oh hugely it's about creative thinking think outside the square uh some some great Newsulan uh Kiwi ingenuity number eight wire come up with the idea draw it on a napkin draw it on a whiteboard and you now have the opportunity to actually get that developed and and get that uh applied into your business. That's exciting.

SPEAKER_02 49:03

Yeah it is the one comment I would say to that is that a it AI creates code fast. It doesn't make a necessarily a great app if you don't ask it the right questions and putting it in the right.

SPEAKER_01 49:16

No but I think that's why we need somebody like you and and your business may be involved to take this further. But but what I'm really loving to hear just is that cost is coming down and it's becoming more accessible for more New Zealand businesses and they really need to start just thinking more about how they can apply technology because it's more available now. Whether they you know prompt it themselves and get something up and running as a minimum viable product that may have some security issues and need some ironing out later down the track. Who knows? But certainly go in that direction because it'll enable them to effectively compete with much bigger businesses who for years have been developing these more custom apps.

SPEAKER_02 49:56

Well they're in a perfect position because this will enable you know I the larger businesses have different problems. You know they've got to transform these big legacy systems that they've got to change their shoes whilst running. So they they can't stop those systems and just put a new system in. So they've got to um you know that they've got to do that with caution and and be careful. But the a lot of smaller businesses and sell new services, widen their proposition really you know and are fast at innovating and change.

Closing Thoughts And Community CTA

SPEAKER_01 50:25

A lot more agile uh able to adjust and come up with new offerings in the in the market. And that's what the technology's enabling you know so it it's that's exciting. Very exciting. Well with that thank you so much Alan uh thanks for uh sharing your knowledge today uh it's very very interesting to hear.

SPEAKER_00 50:40

Good lovely thank you all thanks cheers that's that thanks for listening to We Do B2B by BlueOcean. Now Race for CTAs if you want to join and grow the community make sure to subscribe wherever your eyes and ears absorb information. Don't forget to switch on notifications so you know when the latest episodes drop and for more B2B goodness be sure to follow BlueOcean the B2B agency on LinkedIn. Hello you know how this next piece works the more reviews we get the faster this thing grows. So please do for us what you hope your customers would do for you. Leave a review and share your thoughts. Let's stay connected and keep the B2B marketing conversation going