AI vs public sector jobs

WEBVTT - AI vs public sector jobs Cura and welcome to the business of Tech. I'm Peter Griffin in Wellington. Right now, ministers are looking hard at the public sector wage bill. The plan on the table is to cut eight thousand, seven hundred rolls and carve two point four billion dollars out of operating costs with artificial intelligence.

Heavily signposted as part of the solution automate pretty much everything that can be contacts in his back office, processing, case management, the work that thousands of people currently do. It's been very controversial, but beyond the political soundbites, what would a serious nationwide AI transformation of government actually look like? My guest this week has had a cracket answering exactly that. Brandon Hutchison, as head of Quantum at Global at Services firm HSO, formerly Aware Group, which Brandon founded and which was acquired by HSO back in twenty twenty four.

In response to the government's cards, Brandon has spun up a website laying out one hundred and sixty concrete ways AI could reshape New Zealand's public services. He was actually working on this well in advance off the announcement from Nikola Willis, but he put it all together and published it basically the week that mandate came down from the Finance Minister, from call centers and human resources to health education and even how we test new policies before they ever hit Parliament.

Brandon has been looking at the role AI could play in an ideal world. We get into exactly how difficult this is going to be in the podcast. In our conversation, we'll unpack some of the ideas from that list, the ones that could potentially trim headcount the most well. Look at Brandon's proposal for shared AI enabled contacts in a platform across government, a space for around six thousand people currently work, and why he thinks up to eighteen h full time roles in that area alone are addressable by automation.

We'll talk about the push for common cloud based HR, payroll and case management systems where another three thousand staff are currently employed, and ask how much of that work could realistically be consolidated or redesigned with AI and the loop. But it's not just a story about cutting people, Brandon argues, we're framing this moment the wrong way as a blunt job cuts exercise, rather than a once in a generation chance to improve citizen services, grow new export businesses out of public sector expertise, and build a proper New Zealand digital service to set standards and avoid another decade of technical debt.

We'll explore US four bucket framework for categorizing government processes from fully automatable tasks through to those that must always stay in human hands, and what would take to actually map that across the state sector. Later in the show, we'll leave Wellington behind. Brandon has just come back from Microsoft's Quantum computing labs in Redmond, which I visited a couple of months back. It's where engineers are racing to build fault tolerant quantum computers based on a radically different kind of cubit, the type of bit that exists in the quantum world.

We'll talk about what he saw there. Way he thinks quantum needs to be on the risk register and opportunity map of every major New Zealand organization and how it intersects with AI on everything from drug discovery and new materials to the future of cryptography. So stick around. That's all coming up with my guest, hso's head of Quantum, Brandon Hutchison.

Brandon, Welcome to the business of tech. How are you doing? Can't complain. It just got back from the States a few days ago and no jetline this time, and so I'm pretty happy.

Sunny Seattle this time? Was it? It was sunny, perfect temperature inside and out, but it didn't really get to see much of the outside. Yeah, it was the same.

We both went there to visit Microsoft, and we're going to talk about that. Particularly, the highlight off my tour and I'm sure yours was the visit to Microsoft's quantum computing labs, which are under Redmond Campus. So we're going to talk about that and a big announcement from Microsoft as we record this on the fourth of journe just yesterday about big advances in their quantum computing efforts, So really can you get your take on that. We're going to talk about this really interesting document you published online basically a a micro site, one hundred and sixty ideas for transforming the public sector here in New Zealand with artificial intelligence.

Very topical and almost perfectly timed to stimulate discussion after the announcement about the cuts to the public sector. Before we get into that, though, to find out a bit more about your background, I know you from this great company, Aware Group, which really is very much a Microsoft integrator, works in the Microsoft platforms, has done a lot around artificial intelligence. But take us back before Aware Group, how you started out in business and got into the tech. Sec wagon too, the tech sector.

I remember I was fifteen lying on the lilah of my dad's pool wagging from school and my dad came home, caught me drinking one of his bourbons and said, you got two choices, son, one go back to school or go do something with yourself. So I decided to take him up on the second option and go to university. So I masaged to get early entrints into the University of Waikodo, which was quite daunting at the time, being fifteen going on sixteen, And it took me five years to finish a three year degree and it did part of my post grad in cybersecurity under Ryan Coe when he set up the cybersecurity center at the university.

I remember when I had eighteen, the first thing I did was not celebrate my birthday. It was go out instead of a company. So the first company I set up as a company called develop it or develop it, and it was a small web in software development house based in Hamilton. It still was going today.

Actually it's been probably one of the longest running businesses have been semi part of. But ran that for about three to four years, then had a kit on the way and realized I had start earning some proper money, so it left that business, went contracting in the public sector as a project manager. So this really gave me my first I guess run through of how bureaucracy works within a public sector agency. Did that for a few years, and while I was going about doing that, discovered there is a huge demand that's going to happen in the data space.

Jumped on, got a few of those early tertiary clients as our flagship customers. Realized that my now business partner, John Templeton, we couldn't service all these different clients by ourselves. So between that and pear pressure from Microsoft over in Seattle to set up a company, we got together one night and we had was it about three hours in the basement of one of our ex shareholders' houses in Seattle to come up with a company name. So we thought, what is a name that can span all countries, it's not trademarked, and it's not going to upset anyone.

So that's where the name Aware came from, or Aware Group. The original vision was for us to have lots of different AI spinoffs or data spinoffs as we realized how the technology was going to mature, so jumped in started that. Our first actual contract was actually with Microsoft in the States. We got given due to some awesome New Zealand connections that work over at Microsoft.

We've got a contract to build AI demos for nearly every single Microsoft conference in the world. This was when, or back in the day when people actually physically attended them more often, So we got essentially crash coursed into building AI IoT demos that had to work in front of thousands of people right from the start. So yeah, you can imagine the kind of stress that put us under, especially since AI was just essentially wasn't cool yet. It was one of those things that Microsoft is still trying to market to and convince people.

So this was demos from everything from showing how computer vision would work on from manufacturing perspective, you scratch it, but a still computer vision detects the scratches On't there nice practical way to teach the public or the audience how computer vision works. So we end up crossing a huge amount of different areas from agriculture working on Satia and Nadala's pit project at the time, which was farm beats, which was a lot of fun. We managed to bring that to New Zealand and rolled out out across what was it I think it was from Terra darien Z and kind of showcase where the technology would sit there.

So I think we ended up presenting at or building those demos for about one hundred and thirty worldwide conferences and that kind of gave us the stepping stone to bring all this ip that we could use from different countries around the world that we built this full fuller conferences like Hannover Messe Industrial all the way through to like Ignite and bring it back into New Zealand and showcase that to our New Zealand customers over here or future customers. That really made a stand out because when we first started there was us in about four competitors and to date only one of them still exist.

So from the eco system that started, we ended up going outter ten years selling to HSO which is a big Dutch company which is at the time was funded and backed by Carlile, big private equity firm, and at the start of this year they got bought by Bank Capital. So that was a ten year journey. You were there for pretty much the whole tenure of that, and you're still with You've stayed on as head of Quantum at HSO. Yes, so after we got since you purchased by HSO, we well I got the option.

They said most founders end up leaving after one or two years after they sell off, so they were quite aware that this could be the since the potential case. So they go to me and go, hey, what do you want to do if you got to choose anything? And I was like, where's the future going to be? My expertise is more predicting what technology is going to come forth, the novelty of it, and trying to lobby people to actually go through back it and engage with it.

So quantum it was something that we've been dabbling with for years now. We actually built a bunch of quantum demos for Microsoft about three four years ago, and I was like, I'm pretty sure this is going to be the next big thing, especially with the convergence of AI with quantum and the economy that we have in New Zealand, which is heavily based on essentially simulation of materials from everything from milk through to manufacturing. And we are an export first country. So package in that all up together, I was like, this seems like a great direction to go on.

We'll drill a bit more into that potential opportunity and the quantum stuff, but let's start with artificial intelligence. I know, Aware did some really great projects things on university campuses, you know, looking at how people move around campuses. The robotic dog stuff. You were with some of the first to bring that to New Zealand.

What was that called the Yellow Dogs. Spot from Buston Dynamics. So man that demo had was I think that was the most attention. We doubled our market share across public sector.

As soon as we did one post on that. We had things about twelve to thirteen government agencies coming to us going hey, can you come over and show us how this could work? So we end up rolling that out too. We took that through the Christ's Cathedral to do the analysis inside, we took it through si on research facilities, a test, it through forests.

So I think Aware how to gain a reputation. Obviously, it was deeply embedded in the Microsoft ecosystem, leveraging those AI tools and the Azure platform, and that so did very well. And so you're now with HS. So a lot of consolidation has gone on in the New Zealand sort of tech scene, integrators and that sort of thing.

So that's been sort of part of that trend. But look, we now find ourselves grappling in the private sector and in governments with how to implement artificial intelligen since it's moving very quickly, and then sort of wham out of sort of nowhere. Really a couple of weeks ago, this mandate comes down. Eighty seven hundred jobs must be trimmed from the government headcount.

Two point four billion dollars in savings are required over a very short period of time, and a bit clunky the way they announced it, basically mashing that together with artificial intelligence. And I'm sure it won't all be AI responsible for filling that gap, but the government is definitely leaning heavily on technology to make this happen. You've done a lot of work in government, so you know what the reality is. What was your take when you first heard that from Nikola.

Willis that sort of mandate that has been handed down to government leaders agency chief executives knowing what you do about how our government sector works. First thing I thought was on no, you framed it as job cuts. I was like, no, this is the worst way. It's like, okay, we're gonna frame it as job cuts, then try to convince the people we're going to cut their jobs to then be in charge of replacing and automating their own functions, and like that was the worse mistake that we could have made.

At the High Tech Awards, which was a few days after that, I actually told that to Nicola. I'm not sure how she took it, but we do need to take a different approach, like we should be framing this with hope and looking at how we can do this to support future exports. How can we benefit the smeeths or the subject matter experts within those agencies to hopefully we could spiral out smaller subcompanies that could actually build those agents of change. Instead of going, hey, let's when you go automate your own job and do a good quality job at doing that.

It's impossible, not no one's going to do that. It's a conflict of interest. That was definitely the first thing that I thought, and then I guess the second one was there is a lot of AI ideas that inside the public scept that people have wanted to do and the private sector. There is a lot of experts in the country that just aren't being tapped to come and help solve the problem.

And there's a lot of highly intelligent senior tech leaders across government sector, but it's very very rare that any of them have a foundational knowledge of AI, and we're expecting them to go in and essentially architect a solution that's going to fast track things that faster than any other country has done it, to a better standard, make limited mistakes because we're in a no forgiveness society, and somehow succeed this in a timeframe that none of them have been consulted on.

I think that's the big argument that most people in New Zealand are looking at it going job cuts instead of going, hey, how can we provide a better citizen service to everyone and make this more efficient? Like, there's different ways of framing things. Most of the commentary has been pretty negative about how this has been framed. A lot of people pointing out that just the lack of capability to do this and the potential to really blow trust and confidence in public services and an artificial intelligence underpinning them as well.

However, I suspect you a little bit like me, where you're also a little bit excited at the prospect of an AI driven revolution and what are often quite clunky and inefficient public services. And I sort of saw that in this website that you put together, where you've laid out these one hundred and sixty ideas, and we'll put a link in the show notes so you can beautiful website, very easy to navigate. Chapter zero opens with what you call thirty practical ideas that are actually tied to these twenty twenty six reforms.

And it's the sort of stuff that I hear people in the private sector talking about all the time. A shared contact center knowledge assistance, which is pretty standard stuff now, but we have, according to your site, we have nearly six thousand people involved in contact centers across government departments. I was just staggered. I guess it makes sense when you bundle in iid MSD police all of them.

That's a lot of people sitting in contact centers. That alone, if you streamline that, put them all on one platform, and gave them the tools to be more efficient at their job, that alone would probably trim headcounts significantly and save money. Yeah, with the share contact center. This is kind of a universal agreed that this should happen across most agencies of at least the first step of triage.

And if we assume like MSD runs things about one hundred and sixty contact centers alone, you start multiplying that up by every other agency looking at initial triage that has to happen when half of it's just source and information that's required, the step or the process that is required for that citizen to go about doing it. And then yes, there is specific human related and more complex tasks. But if we essentially bulk any of this, any of the AI solutions that we're looking at deploying across public sector into four categories.

One, we identify the process that can be fully automated, nice and easy. Second would be the automate but with a transition plan, so we deploy it with human involvement as the AI model or the agent actually matures to the point that we're happy for it to transition from that stage two fully automated. Third is essentially partially automated with permanent human involvement. This is where we deal with things like high stakes cases ones that do deal with exceptions to the normal rules as well.

There's lots of processes that won't be able to be fully automated, but that's an exception. And if we assume twenty percent of exceptions for every process was a give and take, that's still a large reduction that could be made. And then the fourth is human only related processes. And the biggest issue that we've faced New Zealand is that none of these processes have actually been mapped or categorized against which of the categories from an AI perspective could be done.

The cour center one is nice and easy because it's been proven trying and to test like air, New Zealand's done a great job and I'm automating lots of their course c in functions. We were involved in that a few years ago and the numbers that we were seeing was upwards of twenty percent increases just by improving the services through AI, not replacing anyone, because nobody wants to be stuck on the phone waiting for information. So it's a win for both the government and it's a win for the citizen itself.

Yeah, so I wonder because is what I'm hearing from the likes of one end Z and two degrees and others that have big contact centers and have made very good use of AI. What they're finding is that customer service improves, their their net rating scores go up, their agents are more effective on the phone, but they're not necessarily leaving the contacts, you know. So I think it's probably a bit of a false economy to think that. I mean, if you put in the in your document here, there's maybe up to eighteen hundred ft addressable jobs in those contacts centers.

Is that suggesting that we could see a headcount reduction of that much? Yes, yep. This one's pretty straightforward, nice and easy. We have, like most contact centers are some of the most well document to find processes of what people's requests are.

Like, we do recorded majority of the cares that come through what they're for, build up our knowledge base on that information. This is the easiest one to go. Hey, which ones of these most common practices or processes can we actually automate and don't do it all at once. Start with the triage of still allowing an automation on some processes, human on some and then as the process to get more and more mature, more stuff gets automated as it goes through tillets.

So that's an obvious low hanging fruit. The other one is a common cloud human resources, payroll and case management system across agency everyone on the same platform. Is about three thousand people across government working in HR payroll, PASEE management. That seems a center as well.

And I know there have been mandates from DIA to basically say we want you to more centrally procure these systems and not be constantly going off and tendering basically for the same thing separately. That would be I guess part of that drive. But three thousand people there potentially that a lot of those jobs could be consolidating. Huge part of it could be consolidated to a degree, so not every single part of their functions could be But if we're looking at trousing where an invoice goes and how this process, and we've got ten different people doing that across ten different agencies and going okay, this is a remittance.

This is an invoice, this is a Paceluve, et cetera, and been able to actually automate the flow from that. That one's nice and easy for us to lock in, but there will always be a human component that we haven't thought of yet as well. A big problem is that we haven't defined a lot of these processes for what is going to take place, So it's trying to solution things without understanding what we're getting ourselves into. And that's what strikes me when I read your list, and it goes from these sort of bog standard things that are sort of the ADMIN to back off as functions of government all the way through to some really exciting innovative things.

Synthetic population for policy testing. A lot of synthetic data is being used now, so that's really cool. Real time legislation, impacts simulator, so all of this stuff. You know, you could do some really innovative things that make a big difference in government.

But as I look down this list and they cover the whole of government, from health and preventative care to environment and climate to start up economy and export opportunities, it's like, boy, we've got a lot of work to do to understand how to do this and do it in a way that you know, we're a bit prickly as a nation around combining sources of data together and things like that, you know, things like a digital driver's license, kiwis recoil a bit at the thought of it, and agencies sharing information and two big brother we've got sort of culturally and capacity wise in government, we've got a bit of a way to go.

The philosophical side of making these decisions is just as much if it as the automation. Like I'll bring up an example that I've been talking about probably for six seven years now, and that's if we look at our education sector and we look at one specific process, which is role taking in primary school, every child, it takes about five minutes a day for role taking an intermediate twice a day, five minutes in high school five minutes per class is five classes a day on average.

You add up every one of those five minute intervals and then calculate that how much time is taken by role taking over the child's entire education. It's nearly half a year of school days. We haven't even started the discussion of not just automating processes and roles, but how do you actually provide a better citizen experience the impact of our economy by even solving that one little problem, we come up with different avenues. Do we use that time for give that back to teachers to actually improve education and with that extra half a year of teaching over a child's entire life, make a big difference to our educational standards.

Would we start looking at shortening our educational time by making some of these cuts to twelve years not thirteen like the philosophical thing, into a lot of these bigger AI pitchres that aren't just an automation of a process is the exciting part. But as a country, this stuff has been available for six seven years and no one's jumped on this. I remember sending Chris Hopkins an email back then and then it stuff did a bit of a expos on the facial recognition side, and nothing happened.

There's no central procurement across education within the schools. The amount of effort that would take to actually implement something like that, we're not set up to. Since you adopt the AI gains that we could be, we need to start at the right start. I reckon it would take six months to go across most of the agencies do proper business process mapping, categorize every one of the processes under those four categories, fully automatable all the way to human only use those there actually start prioritizing entriaging, which since she functions that each of the agencies should go about considering the use of AI for or just normal digital automation.

Then start the actual process. And this is this should all start with one central agency that it doesn't do the AI, but is in charge of defining the standards. So like if you look at in the UK, it's what's UK digital Service, we should essentially be sending up the similar thing these on digital service. It's essentially in charge of only monetary and standards across every one of the agencies that do submit.

Kind of like how people submit apps into the Apple Store. You've got to go through a big process that checks the standards, the quality, make sure it's good to go, goes in and there is continuous analysis over those apps. That's way you can get kicked off the store. It should be the same thing regarding the functions itself.

Each of the agent essentially, each of the agencies should be responsible for building their own agents, but based on the same essentially standardized architecture that we can not some repeat, but we can and ensure that there is consistency from a maintenance perspective, because one of the biggest issues that we are going to face in the way that we are going about trying to implement AI at the moment is we're setting ourselves up for another generation of technical debt. Every single agency has going about looking at AI automation in their own way.

It doesn't take a lot for us to define one standardized way of going about doing it. It doesn't change the benefit, doesn't change the essential of the technology stack that we do, but it does allow us to if one function is essentially archived or could be replaced it, we can replace it very very quickly. You know that is as you say, technical debt is a long standing issue in our government, and so we're paying the price for that. I guess the flip side of that is the fears of a sort of a mind molithic tech provider across government.

You know, Microsoft actually is big across government, particularly on copilot things like that Microsoft three sixty five. It's a bit more diverse on the cloud side, AWS and Google and others data com local players are in there. But how do you avoid that sort of vendor lock in when you go down that path of standardizing things. How do you make it sort of modular so that other providers all of that data can be quite easily shifted out to another provider.

This is a sincely where New Zealand Digital Service as a concept would define those standards agnostic of platform like I'm while i'm since she's single cloud integration partner over the last ten years, I do believe that having a single point of failure is the worst thing that we could do in the country as well. So define the standards does not mean we have a lockdown specific partner and everyone like we already know that we're not mature enough as a country to implement half these AI initiatives considering that everyone goes copilot can do all of this, that's not true.

Copilot can do components of it, and changing the maturity of the people involved probably is one of the first steps. But we should be encouraging all the cloud providers and essentially open source tech providers as well, to go about the challenge of actually automating or building agents for those functions. In a particular way that is compatible to the New Zealand Digital Services standards. That would be the way we'd go about doing this as well, and it essentially it could be gamified as well if there is different providers that can provide Say we've got a model or an agent that's been submitted to New Zealand Digital Service.

It does a process eighty percent, Well, maturity hasn't adjusted in six months, so no one's actually no one from the agency is maintaining it because the provider has a mating changes. We should be allowing other people to go through and compete for Hey, I've got a better agent that can do this, and allow this constant series of competition because we do not want any of the agents going forward or the automation to be stuck at a single point of maturity. We want the maturity to constantly grow and this is where something like New Zealand Digital Service would actually go about ensuring that from a maturing perspective, there is pressure onto the agencies to constantly and continually improve because this is what we're going to go about doing.

We're going to build this AI to solve one law problem and then no one's going to maintain it. No one's going to look at the maturity of the improvements. Then we're going to end up six months after releasing that having some issue about bias or just something that hasn't been done that degrades public trust or erodes public trust and what the public sector is actually doing, and it's going to create large larger capital costs going forward in the future for us to go through repair it and do one giant, big AI project to try to fix all these issues versus having a continuous improvement mentality.

So at the moment, I guess the partner of an internal affairs is driving a lot of the tech related changes in government. So you suggest you basically have a unit there or something separate that it is responsible for setting the strategy, the agenda, the standards, maintaining those standards, the reusability of things, and ensuring that happens across government. Absolutely one agency that does that but does none of the AI development. What the nzes should be looking to actually do is essentially three main things.

One is it helping, is it improving? And is it trusted? Those are the three metrics that they should be recording against. Cost down, completion up, it's the same number.

Both sides are happy with the citizen and the agency. Essentially, the maturity is autonomy going up while mistakes stay down. So we want to see the continuous improvement and maturity go up. And the third is the sally ensuring that it's trusted.

Public sector isn't going to jump in and use any of the citizen services that are provided by AI if the trust is not improving. And like I know that lots of people in the past, I think someone on a insid herald this morning was talking around from a trust perspective. All it takes is one instance for a public sectorm to not be trusted to provide that service, and well that's going to happen regardless. So we do need to shift our marketing approach for these agencies to be tied to usage but also improving trust.

So this is a week ends DS would actually jump in and measure every one of the agencies based on those three metrics. Is it helping, is improving and is it trusted? And I guess the other thing is making sure that there's opportunities for New Zealand companies to participate in this. And do you have any reflections on the capabilities within IECH sector to be part of this.

Sure, you know we're not going to develop models necessarily that are used by government departments, but there are lots of things that wrap around those. Our startups are established companies could be tendering for as well. I think having a series of like it's just my thoughts, I think thrown out and ensuring that we've got a few strategic partners that are aligned, so ten to fifteen different New Zealand based companies, those partnering with the subject matter experts within the government agencies to whatever form of automation that we do around those government processes.

There is a good export opportunity for us to look at. If it's successful in New Zealand, why don't we export assuially government improvement as an offering to different markets that are doing it. Like no one has gone about doing like Central AI managed platform across all of government in the world so far. Like there is places that have done some components, but in Estonia is probably the closest that we're looking at.

And if we start like there's so many different things about considering the future versus ten years ago. Like ten years ago, we could build this waterfall project of implementing AI, and it could could potentially work. Nowadays it can't. Like if you look at different forms of technology that are coming as frontier technology is going forward.

Even if you look at digital identity for example, currently we don't even have what NHI and one other number which we use is our single point of a number that we can track every citizen to. But we're not even allowed to, like agencies aren't allowed to querium like query those numbers to create a single identity like that would be a great start for us to change those rules and allow that to actually happen. But then, okay, we go about doing that in three to four years. If we start looking at how sinially new biometrics will change the digital identity landscape with things like this thing, there's about six different new enabled quantum technologies that will be able to deal with biometrics, and we look at two or three of those markers, we stop being able to sini she spoo for emulate other people.

What happens to all of the work that we've done in the last three years. We need to have this foresight of going what is going to happen in the future and architect what we're doing now to be future. You're thinking not just dealing with the problem that we're doing right now. Consider it the same with roads.

We don't want to build for what's now, we want to build for what's going to happen in the future. I think you're dead on with that. You know, the digital identity and trust and it's a framework that the government is developing on that more for sort of accessing both public services and private sector services like banking without handing over your data all the time to third parties. It's basically verified in a private vault on your phone or something like that.

So I think that's part of it. But there's a more fundamental issue. Off we don't have a social security number that ties one citizen to every government government service, and Singapore, for instance, has seen how valuable and how efficient that can make government service. If we look at from an AI perspective of actual citizen improvement.

Simple example, I go into say we've got like a my New Zealant, which is one single platform where everyone engages with it could be voice, that could be like a chat interface everything else, and say you're moving house, I am moving to christ Church for example, having that go through every different government service update it right from the start. I know real Me has tried to do this in some level, but it's not just it's not the ease of going about doing that. And then you've got one step above that.

What happens if you're updating your information through the same interface, for I know, ID use some metaphor which has got your new address, even having that small little agent that can identify that your address is different than your master record and proactively go through and update your records, even just cuerit to you, Hey, is this your new address? Like that this here will improve our data quality through in New Zealand as well. But also from a citizen experience, it's one let like we're reducing the contact points, not just solving the exact same problem that we have, which is if I have to do something, I've got a phone, multiple agencies do it over and over where she's stopping the problem from happening.

I think there's a bit of naivete as well and among ministers, and you know, Nichola willis about how much AI costs to implement. You know, it's sort of seen as a efficiency driver. And sure there'll be a return on investment from this technology if it's implemented properly. But you know the token industry, the tokenomics, you know that it's expensive stuff, and you know it's going to need to be serious investment to get that return down the line.

I couldn't name one minister that has essentially a foundational knowledge of AI and how the costs work. So predicting what's going to happen from those costs is one thing. Like it's not just a model or a token cost, it's also the cost that no one's talking about, which is actually the data connectivity. A lot of people are talking about, Okay, we've got a data quality problem across government, we've got a essentially the cost of actually implementing the AI, but no one is talking about the big one, which is actually the data connectivity problem.

We do not have a great, essentially foundational layer for how data is connected between different agencies, and the way that we're going to develop a going for in the future is going to be essentially very siloed into different areas, which is going to build like it's going to increase the costs of what you're talking about, which is the token costs by having such a disparate like series of agents and models. It's a real problem. So just to wrap up on your work on AI and the government, what do you hope to achieve from this?

Have you had any bites from people in high places and government saying hey, come and talk to us, we need some help here. Not from a minister up speak to except for my brief conversation with Necklar at the High Tech Awards. But last week I did attend as probably the only commercial partner with about ten different agencies in Seattle, which was actually quite good a lot and had actually read what I published, which was quite surprising because it was quite on a quite unpromptu the way that I went about publishing.

It was going to be an article I wrote and published later on this year, but just due to the media, I was like, now I'm going to bring it forward a little bit more. The conversation is that is huge appetite within the agencies and there is a very clear mandate we have to do this. Like there's not one of those agencies that attended that i'd assume does not get the message that they have to implement some form of AI. But understanding the maturity of their senior leaders in adopt in it.

They have their own learning curve to go over understanding the technology, and I believe that we're going to make a series of very bad architectural and future decisions on how we implement AI. We've got a once in a generation opportunity to be like Estonia and get this right, or it could set us up for decades of pain. So hopefully we do the former. Anyway, Let's say you talked about quantum there.

Let's get on to that. You're probably the only person in the commercial world in New Zealand that has quantum in your title. There are researchers, some of whom I've had on the podcast, who are at Dodd Wall Center and others who are really interested in working on quantum technologies, photonics mainly other things quantum sensors. But you obviously see this as the real future or part of the future off HSO and the tech industry in general.

You just came back from Seattle from Microsoft's quantum Labs. I was there a couple of months ago. So maybe just visualize this for us so that people understand it's relatively small facility. Two stories.

You've got chip manufacturing going on their fabrication, and you've got these quantum computers upstairs. Oh, you've got those two, and then you've got all the satellite buildings with That's one of the craziest things that you go there. You expect, Okay, you're going to be some centralized research functions, maybe a few scientists here and there. What surprised me was just the quantity of different experts across every different like facet, Microsoft is investing so much into what could happen, not what is going to happen, but what could happen in different areas, So like, that was one of the big key takeaways is that just the level of the level of intelligence across half the experts.

It's just one of those things that you just don't normally get to experience, Like not downplaying New Zealand's like intelligence level, but when you're stuck there surrounded by people, you feel inferior, Like you see that these people are thinking so far forward and what they're planning for no one's thinking about. That's what I took away from actual quantum labs itself. Holding the little chip is quite fun as well. It is a lot heavier than you could imagine most most time, you think Archat's is going to be like a lout of the Intel penting for chips.

It's quite light from back in the day, But no, the actual the new quantum chips are quite heavy in itself too. You're talking there about that. They call it what the Mayorana chip. The first version of that was there.

I had a look at that as well. Microsoft has just announced a second version of that. Maybe we should just talk a little bit about Microsoft's approach to quantum, which is a bit different to the likes of IBM, which has been doing this for a long time, and even Google. This is based around topological cubits.

Can you give us a breakdown brand and of what that actually means. Essentially, there's four four or five big players that are investing in the quantum space. So you've got Microsoft, which has gone down the topological approach, which is one approach. Then you've got IBM, which has got their essentially quantum chip I think it's called Flamingo.

You're at Google which has got Willow, and you've got an Australian company called said Quantum, which has taken a different approach as well. I think Australia has invested a bit a billion Australian dollars into that one, and they until yesterday they had actually promoted that their quantum they were going to be the first to have their the first fault tolerant a quantum chip in the world. But now Microsoft is now competing on their twenty twenty nine surprise that Microsoft didn't say anything when I was there last week, but there was a different approach.

So you've got essentially a topological we got superconducting am I on, and then there's things that's about two to three other different approaches. There's no one right approach on going about essentially that investment into these quantum technologies or as quantum approaches. No one knows what is going to win. But so far, with Marcosoft's big announcement yesterday, they are definitely the front runners at the moment.

Previous to last week, I would have said it would have been IBM. It'll be interesting to see what happens because in the last like three weeks has been huge announcements. IBM has just got signed I think a dollar for a dollar massive contract with the US government around developing sorry the US first essentially sovereign quantum facility as well. Yeah, and they've also the US government has put money into I think I on Q and d wave and others, so which led to a big spike in their year prices.

So quantum is is hot at the moment, getting hotter. I think Microsoft, you know, is taking a bit of a more radical path. It's trying to engineer something that's funded mely more robust. These cubits, which are really hard to imagine, but the quantum equivalent of bits of data, you know.

The maintaining the quality of those and making sure there isn't a lot of errors in them, and they are really difficult to manage. That's the key really to an effective quantum computer. So they're taking a different approach. They've got this new chip Maarana II, which they claim has one thousandfold improvement and reliability over the chip that we held in our hands.

So that is massive if they can scale that up and twenty twenty nine very aggressive. They told me they're working with the company to produce a one hundred cubit computer I think by the end of the year as a sort of a prototype. And you can just sort of see I guess you know, you've worked in the Microsoft environment for a long time on the Azure platform. If you have these quantum computers dotted around the world that are accessible via Azure by any small business, you don't have to have a quantum computer yourself.

You're basically renting time to do this high capacity computing on a quantum computer that's sitting in Redmond or open this somewhere else. Oh absolutely so. Essentially, my guess on how Microsoft is going to go about sincely integrating their quantum into the existing services. How is Microsoft actually going to incorporate quantum?

Is it going to be pushed into Microsoft services standlan services? Whereas Microsoft going to be sitting on essentially their playbook around this, And there is definitely going to be specific dedicated quantum services built in specifically with Azure quantum. Quantum is not great for every single problem. It's actually great for any very small amount of very dense mathematical problems that we currently have very inefficient or no ways of going about solving.

So there is a great flow chart of going like what form computer is good to solve this problem? And it is going to be a hybrid solution on everything that we do Microsoft does fill the box and every one of those other areas for this essentially giant hybrid solution same as IBM. I'm same as Google as well, will be a big push in doing since you're taking that approach. One of the things I had to take away from the visit over a Microsoft though, is majority of the focus on the Quantum site is primarily around research.

It's not around optimization or solving optimization issues. It's actually around as you can see in most of the videos, drug discovery. If it's around new material science, those are the key focus areas. Originally, when I jumped into this, I was like, I think the essentially the quantum optimism like optimization, like geopolitical simulations with the fun path to go about doing this.

But no, it's definitely around the actual drug discovery. So from New Zealand's perspective, we should be focusing our small pharmaceutical sector that we do have, building those partnerships up so we can be competitive in the future from essentially a material science perspective, Like I know, Fonterra is actually already working with Microsoft's quantum team and the discovery team over there, which was really cool just to hear I don't know what they're doing, Olish I told me, but just hearing that New Zealand companies are already jumping on the bandwagon, have their roadmap, but it is the private sector.

What we do need to do in New Zealand, specially across our CRIS is actually open up the access to these services, and I guess nationally negotiated access to both the quantum computers around the world fast track research grants to actually enable that. The second is have the training from scientists to actually know when what help with these quantum technologies. You and I were at a conference in Adelaide recently. I saw you briefly there.

It was Quantum Australia, so and I was blown away a little bit like you were saying when you went to Microsoft and you're like, oh my god, we're so small and underdeveloped in New Zealand. I've sort of had that feeling just being in Australia because there's so much going on there. But the team from the Advanced Technology Institute was there and they're currently considering this, do we chuck a lot of money into quantum as a nation, So they'll make that call I think in July next month, But what was your sense seeing what's going on there in Australia is the opportunity for us and either pockets of capability we have that can be applied to quantum that.

Would take the Australian approach of not one quantum technology is better or superior than others as well, like Australia's done a great job and putting an emphasis across all four different areas of quantum quantum computing, quantum cryptography, quantum communications and quantum sensing and New Zealand the specialists that we do have from ends IAT and others and another it's going to be slightly unpopular, but there is a big focus on our actual quantum sensing from a photonics perspective quantum comms, there's a very minimal foresight into what the cryptography and the computing side is.

It's part of the reason why I'm jumping and trying to advocate for this because like there is nobody else really pushing it. There's a lot of cyber security experts needs on that do care about the crytography side, and the return on investment from a quantum computing perspective is significantly higher. I think it's about three to four times the quantum sensing and it which is higher than quantum comms and everything else. So we do need to cover our investment and make sure that we are not just focusing on what we've done well so far, but also put the investment into the areas of quantum which we do not have a strong I guess the foundation on which is the quantum computing and quantum cryptography side.

Yeah, And I guess they're looking at we've got seventy million dollars, which really probably equates to like seven million dollars a year or something like that. Do we have enough resource to forge into a new area, particularly quantum computing, which can get expensive very quickly. So I guess it's a bit of a dilemma. But I'm sensing that year of viewers is that if we are focusing on advanced technologies in this country, sure, AI is a bit of a no brainer, but increasingly quantum is as well.

AI is now just in flight. There's conversations that we having like this one, which is like, how can we mature the use of technology which is or essentially adopt it and play catch up. I think that's a different question to the way that we should be looking at essentially quantum going forward. We should be looking at quantum going forward based on the timelines the risk if we don't do it, which is the same question we should have been, well, we should have been asking ourselves with AI years and years ago, but actually taking the risk, the risk of actual investing into things that may or may not work, and looking at what could the sential return investment be if it was to succeed.

And you talked about cyber there, and presumably that's a bit of a focus for HSO and the tech industry. You hear cyber security analysts talking about supposedly q DA, which is the day where quantum computers break are public key encryption, you know, which underpends a lot of our sensitive transactions and data storage and that sort of thing. So it seems like the timeline for some functional quantum computers is coming forward. It's something we really need to get a handle on.

And because there'll be infrastructure, there'll be algorithm, there'll be software upgrades that are required to keep your stuff stafe. Yeah, so i'd actually combine the quantum cryptography or the quantum risk for QDA. We can actually combine that a little bit with AI risk from things like methos and all the other topical security risks of AI too. So every organization New Zone should be looking at how quantum is going to change the security standards as well as how AI is going to change it too.

One big project go through modernize it. Like NEST, which is the National Institute of Security Standards in the States, they have already released a series of what they would consider safe approach which for transition. But the thing is in New Zealand, most agencies don't have someone called like a cryptogaphy cryptographic asset inventory. We don't know what we currently have, what form of encryption that is currently in place, and we don't understands what we could be transitioning it too.

We do have to have that upskilling exercise because from a quantum perspective, there's going to be two big areas which will be affected. So one is essentially factorization, which assures algorithm that is going to be the single hand lead the biggest risk. And it's like the modernization will happen a lot of the time by the platform provider itself, so there will be a small subset which will be required from agencies or from businesses in New Zealand, but there will be a lot of it which will just be done by the likes of Microsoft as part of their transition plan.

The second is like Grover's algorithm, which is search based functionality. To those two alone is probably where the quantum risk actually is. There is a separate risk that no one's talking about, and that is what the halves now decrypt later is. There's the first issue, which is dealing with Okay, with modernizer the security stands now to prevent breaches in the future.

Cool, But how do we go about dealing with all the encrypted data sets that have been hacked already but just not decrypted. So this is called harvest now decrypt later. And how many times within a board like a CEO is replaced every five six years or so, that CEO is probably not going to a fundamental understanding of data that was hacked ten fifteen years ago, and what the risk to that business is. It's just reactivating previous hacks.

I'm makeually talking about this it in seut of Directors AI forum in about two weeks as well, trying to get the message out there there is bigger risk than just protecting now, it is also the pr risk of what happens when essentially someone is stolen the safe, they can't get into it. We've now got the tools when QDA hits to break into that safe. What was in that safe that is super important to your organization and how do you go about trying to get it back or not getting it back?

Yeah, it shuddered a thing how much data that has been exfiltrated from government departments, from businesses, Sensitive data that we just don't even realize is sitting there on a service somewhere. Someone's waiting for the day to be able to access that. Some of it will be out of data, no longer relevant financial data mapp is. Some of it will be explosive when it goes public.

Just finally, Brandon, you know a lot of KEI we founders have gone through this journey where you get brought out by a multinational and then you're highly respected, You're retained in the company as an esteemed person for advice. Now what does that feel like? It's no longer your baby, but you still have you're still looked up to someone who has a really important role to play. It must be quite a transition mentally to go through transition.

I'm quite fortunate that quite a few of my friends have sold their companies before me. One of the key biggest bits of advice I was given is do not stay in your same role like That's one of the issues with doing that is that people do look at you going and having the same expectations of what you were able to do, the power you could do, the freedom that you'd enable, but you're unable to do that because you now have a bigger boss. So that was probably the biggest bit of advice was move into a different role so that the natural transition of the company purchasing gets to essentially merge or have that natural fight around the culture without me being right in the middle of it.

I still jump in and back both sides and kind of smooth things out of it when needed. But the benefit was we're actually purchased because they wanted to deploy our way of deploying AI in New Zealand to all of their different markets, which was quite fun. We actually won New Zealand Partner of the Year for Microsoft for that year as well, so perfect timing that definitely helped with evaluation. But having that proven track record of we know what we're doing better than you was definitely helpful in that negotiation because we had multiple offers at the same time that we had to work through and this was the one that gave us the most creative freedom for our New Zealand staff to deploy the way that we do AI.

So we haven't had too much. We didn't lose one. We had no resignations from our team for the first year after we got purchased, which I think that kind of shows that it was quite smooth sailing in the approach that we did around the merging or the algamation. That's great, what a success story and so cool that you get to wear the quantum hat.

And really hopefully we will see more roles like that in the private sector in New Zealand and that that expertise gets drawn on by the government because boy, they need it both for AI, for Quantum and just getting a coherent vision of how we adopt advanced technologies. So good luck with all of that, and thanks so much for your insights on the hundred sixty ideas for the public sector for AI, and thanks so much for coming on the business. Of Tech awesome. Thanks so much Peter.

That's it for this episode of the Business of Tech. My thanks to Brandon Hutchison for taking us inside. Both is one hundred and sixty idea blueprint for an AI driven public sector and the cutting edge of Microsoft's Quantum program. If you want to dive deeper, you'll find a link to his ideas site and to more background on a government's AI guidance in the show notes.

If you enjoyed this conversation, please follow the podcast, leave a rating or review that really helps, and share it with someone in your organization who needs to be thinking about this stuff. I'm Peter Griffin. Thanks for listening. I'll catch you with another episode next Thursday.