Why AI Fails Without Clean Data

Mark Smith: 0:14
Welcome to the Power Platform Show. Thanks for joining me today. I hope today's guest inspires and educates you on the possibilities of the Microsoft Power Platform. Now let's get on with the show. Today's guest is from London in the United Kingdom. He's a founder and Power BI lead at Amagi. Is Amagi correct? Did I pronounce it right?

Harley Webster: 0:46
I pronounced it Amagi and we can maybe talk about what it means later on there we go.

Mark Smith: 0:50
We'll unpack that in a second. He is an experienced data professional, driving business value through data visualization, analytics and process implementation across multiple industries and global markets for the past 12 years. You can find links to his bio and social media in the show notes for this episode. Welcome to the show, Harley.

Harley Webster: 1:09
Thank you very much, mark. It's fantastic to be here and, as I was just telling you earlier, this is my first podcast, so I'm very excited to do this.

Mark Smith: 1:19
It's so cool to have you on, and your area of expertise is kind of one of my favorite topics that has developed in the last 12 to 24 months, which is data and the importance of data if you want AI to work correctly, and so it's cool to have you come on. Tell me about the name of your company.

Harley Webster: 1:40
Yeah, so Amarji is, as I said, how I pronounce it. It's actually a word of Sumerian heritage, so Sumerian being the first human civilization on the earth which is in modern-day Iraq. Nice, yep, yep. If people do look through my work history, you can see I've spent 10 years in the Middle East. I love the region. I married an Iraqi woman Wow, very good. So we have this link to the business name through that family association. It means freedom.

Mark Smith: 2:13
Is that right? It means freedom.

Harley Webster: 2:15
It means freedom, and it can also mean freedom from taxation, which is a bit of a UK tax system based joke as well.

Mark Smith: 2:22
It's interesting that you say that because I've literally walked in from the other room speaking with my wife about. I did a piece of work on myself many years ago which was like what do I want in life? Like what is life all about for me? And you know, the natural thing is you start with things like money and happiness and things like that, and I just kept refining and refining, and refining and I distilled it down to one word, which was freedom. So my defining word for the last 20 odd years has been freedom. Because when you look at freedom, that means freedom of choice, freedom of what you do, freedom of health care, freedom of if you've got true freedom. It's, um, it's incredibly powerful thing. So I love that name very cool so do I?

Harley Webster: 3:15
yep, I love it and, um, I think for me what it means for me. Um, I've worked in many big organizations where there is bureaucracy. As we all know it happens in big organizations there's lots of layers of middle managers. I'm someone who just loves to get things done quickly. I am quite impatient.

Harley Webster: 3:31
I love just to go and fix something, make it done, make it complete, deliver something. You know I'm one of. I absolutely love scrum because it enables that uh kind of I love just to go and fix something, make it done, make it complete, deliver something. You know I absolutely love Scrum because it enables that kind of methodology in our work. So for me, amarji means freedom in terms of this is my business. I can go and just do stuff quickly and get it done.

Mark Smith: 3:51
That's superb. I love it. Tell me about food, family and fun. What do they mean to you?

Harley Webster: 3:56
Yeah, they're all very, very important things to me, but we've got to start with family first. Um, I've already mentioned my wife of 10 years. I couldn't have done anything without her in terms of her pushing me to to start the business, uh, pushing me to go out of my comfort zone. You know, things like appearing on podcasts or making linkedin posts or things like this, which I'm so grateful for. I have three children Keep me extremely busy, so you can imagine what that's like having three kids in a business.

Mark Smith: 4:26
Yep.

Harley Webster: 4:27
My kids. I've got one in the teenage years, I've got one and two girls close together in age seven and five and yeah, it's fantastic. And food I am an absolute foodie, like I. I think you are as well, mark, because otherwise you wouldn't ask all your guests the same question as I've been picking up on um. But for me, I'm an absolute foodie.

Harley Webster: 4:48
I've been lucky enough to live and visit many, many places. Like I said, I had a big uh stint in the middle. I've also worked in Copenhagen, in Denmark, aberdeen, scotland all the glorious places oil and gas takes you. But I always, always, always think, okay, where am I going today and what I'm going to eat there? That is the thing that I always think of first, however, one of my daughters has got several very severe food allergies, so she gets anaphylactic shock from a variety of foods. Um, so it's. I'm always the chef in the house as well, so it has. You know, when we cook a family meal, I'm being very creative now to ensure we're creating food that we love still, but ensuring that we've got a safe meal for my daughter. So I'm still cooking curries and chili con carne and fajitas.

Mark Smith: 5:43
How old's your daughter? She's six. Interesting because my boy had it, as in food allergies, that we always had to have an EpiPen with him. I only once had to use it when he was about would have been about nine. He licked the cream, the milk off a coffee cup and that was enough to send him. So he had dairy allergies, he had nut allergies, fish allergies and he's now 19, and he has grown out of all of them. 19 and he has grown out of all of them. But he didn't grow out of a lot of them till he was like 13, 14 years old before he did and he no longer has to have an epi pen. He's been phased onto the various things over time and tested with the doctors and stuff. But I know what that's like, as in you gotta read every food package label to what it has in it because you don't know what's going to set it. You know set it off.

Harley Webster: 6:41
So, particularly when eating out, it was a thing that was looked at very closely and from a young age you got good at reading food labels themselves yeah, well, you heard about the natasha's foundation uh, the girl who unfortunately consumed sesame seeds in heathrow airport that was not on the label of a sandwich and unfortunately passed away. There is a charity in her name in the UK which conducts a lot of research into food allergies, and it's one of the charities that we support.

Mark Smith: 7:11
It's crazy how the effect something like that small as a sesame seed, the impact it can have on on different people's bodies. Um, what was your career? Uh, journey into the area that you work in now. How did you, you know, find yourself in this position?

Harley Webster: 7:32
yeah, I think most people probably don't choose it, it just happens and that was absolutely the same for me. But anyone you know young listening who thinks I really want to get into into data you know, don't listen to what I'm saying because I just fell into it. Uh, so I'm a geologist by background, so I studied geology at university of portsmouth and um managed to get myself in the oil and gas industry as a data manager. So data manager in oil and gas terms means you are responsible for quality checking the data that comes from drilling oil wells or seismic surveys. So a huge amount of money is spent drilling these oil holes and you collect a lot terabytes of data as you drill. It gets actually stored on SD cards and hard drives on the device itself. So you're measuring all the properties about the rocks, so the radioactive properties, porosity, permeability, all of these things and that data comes in, gets packaged up, delivered by the service companies, the big companies Baker Hughes, halliburton, schlumberger and somebody's got to check all of that data, make sure it's right, make sure all the metadata is correct, load it into the right databases or applications for the analysts to then use.

Harley Webster: 8:56
So I had many years as a data manager in various businesses in oil and gas, and that's probably given me my kind of little bit old school approach to data quality and data governance as a foundation for everything, because without those two things you're not going to have any good analytics and AI. So from being a data manager, I was actually part of the biggest oil and gas transition in history. So what do I mean by that? In Qatar, I was working for Maersk Oil Qatar and they actually lost the rights to one of the oil fields, the Al Shaheen oil field in Qatar, and they lost it to Total or Total won the bid, should we say.

Harley Webster: 9:41
So you've found yourself in a position where you've got a 300,000 barrel of oil a day, field with thousands of employees and petabytes of data, which has got Musk's intellectual property, intertw, intertwined throughout it, and you have to hand that over to another company. Wow, absolutely mammoth project, and I'm just so proud to be a part of it. So I actually just got a phone call one day from um it was actually from the head of pmo from musk and he said I searched the directory for data and you're the only person that came up that was online. So I picked up my desk phone. You're the only person that came up that was online, so I picked up my desk phone. You know the old Cisco phones we used to have on our desk.

Harley Webster: 10:23
And he said can you come and have a chat? So I had a chat and he said we need somebody to design a mechanism for tracking and sending our data across to the data room for this transition. Can you do it? And I've always been a yes man in my career. I've always just done things like move to Qatar when I was 21 or just said yes to stuff and yeah, I said yeah, okay, yep, absolutely cool, let's do it. And I was suddenly found myself in a PMO team at the top of Musk Oil, making these data tracking templates in SharePoint and Spotfire.

Mark Smith: 11:01
Wow, wow. I expected something much more elaborate. Oh, no way. That's incredible.

Harley Webster: 11:08
Whatever you think about any industry in the world, you know that they all run on Excel and that is our endeavor to. That's one of our endeavors to change. But yeah, so we were tracking all of this data we were sending. We would get requests. You know we'd get hundreds of requests for information every day. The request would come in. We'd have to send it to the right department, collect the data, strip the IP, send it across and make sure all of that was tracked.

Harley Webster: 11:34
So I started building these dashboards and I started saying, right, this is how much, how many requests for information we've had. This is how many we've satisfied. This is the SLA we've put on it. We're not overdue on anything. And I would make these reports and we'd be suited up going to the top floor of these skyscrapers in Doha presenting to Total and Qatar Petroleum saying we're on track. And it was all because we had this whole data visualization system working just super efficiently. Wow. So yeah, absolutely amazing. It's a year-long project that was. And then I did transfer over to Total, where I ended up being the head of performance management for the subsurface division.

Mark Smith: 12:15
So let me get this right. You went from the company that didn't win the Ford contract and you then moved to the company that won the Ford contract.

Harley Webster: 12:24
Yeah, as part of the deal, they had to offer everyone jobs. Wow yeah, fantastic project.

Mark Smith: 12:31
So a couple of things spring to mind. You talked about you had to strip off the IP. What did that mean practically, give us an example of a piece of IP that would need to be removed from a data set to transfer over.

Harley Webster: 12:47
Absolutely so. When you have some raw data from an oil field, the intellectual property would be the analysis of that data, the end result.

Mark Smith: 12:58
Yes, yes.

Harley Webster: 13:00
So what you can have in oil and gas is you have log data or seismic data. So log data is a y-axis line chart. Essentially. It might be incomplete, it might need some correction, it might need some data processing and it needs analysis. So anything that Maersk employees had done to that data was intellectual property, so we just gave them the raw data now from musk's perspective.

Mark Smith: 13:26
I assume that the interpretation was of no value to them after that fact, because they no longer had the contract for that particular well as an example, it could be, or if we had similar oil fields with similar properties in other parts of the world.

Harley Webster: 13:39
You would also want to know and have that experience yeah because, as you know, the earth crust has moved over time, so you get oil reservoirs which are exactly the same thousands of miles apart.

Mark Smith: 13:50
Wow, this is so, so interesting. And so how did that transition for you to what tools do you use today? If I said what are your data building blocks and why I said before data has become such an important area, is because in the area of AI, the results have hallucinations, in the generative AI, for example and I go, hmm, and I'm talking about if this data is coming from your organization, and so from a microsoft perspective, that might be the graph that they are running like co-pilot runs against the graph and that's might be all the organization knowledge. Every human being, in whatever you do each day, whether you're doing a PowerPoint presentation, updating an Excel spreadsheet, working on a Word document people introduce error. Right, you didn't round out that calculation correctly. You didn't remove certain sentences from a RFX document, whatever it is, and then you create copies of this and then you say, hey, ai is trained on all my organization data, with probably millions of micro errors built in all the way through. And then you go, oh, I got it wrong and you're like, hang on a second, you trained it on all this error prone data that you have inside your organization.

Mark Smith: 15:15
That's why I believe that the true ai story is a data story, because until you get your data clean, filtered right and probably narrowed down, rather than trained on the ocean, it should be just trained on whatever the task you're trying to achieve, the data that's specific to that, you're going to have a problem. And so, if we look at a modern you know the type of projects that we're going into these days around AI and around the Microsoft stack of building products, whether that be fabric, whether that be purview, whether that be the power platform, dynamics, etc. Dynamics, et cetera we often are finding that we have to back up and go back to the data and go hang on a second. Is it the right data? Is it clean data? Is it enriched data? If you look at the tool sets that you use now in the projects that you're involved in, what are they? What are your tools of trade?

Harley Webster: 16:17
Yeah, so I think the revolution of data engineering has happened with fabric. It's, it's absolutely fantastic what we've got our fingertips, uh, and it's the main tools I use and I promote in my day-to-day projects. So if we're talking about ai and we're talking about getting that foundation right the data quality, the data governance, the data provenance you know where does it come from so important. We have to look at what we've got our fingertips now. And what do I? What do we use? So imagi we specialize in data engineering, data visualization and also the business consulting side of it, which is how to get value from what we've done.

Harley Webster: 17:00
Microsoft Fabric has this beautiful way of just combining all of the old kind of different named tools that Microsoft had under one blanket and the data pipelines. To pull data from different data sources, clean it on the fly, store it in the data lake house is absolutely brilliant. There's also Dataflow Gen 2, which looks a lot like Power Query in your browser. We can use that to get data from other data sources combine into the lake house. And then we talk about medallion architecture inside data lake houses, where we have our bronze, silver and gold. I think everyone has a different definition of what they would put in each bucket. But you know, you can just really quickly say bronze is raw data, silver might be you've cleansed the data and gold might be you've built a view or data model that you can just use straight away without having to do anything to the data. So the tools I use on a daily basis are all around Microsoft Fabric and actually the easy bit is the data visualization, because I think anyone can pick up Excel and make a nice table and open Power BI and go there we go. I made a graph, there we go. I made a table, an open Power BI and go there we go. I made a graph, there we go, I made a table.

Harley Webster: 18:22
But the value comes in when you're working in organizations who have got data in completely different systems, different geographies. You might get mergers and acquisitions. There's a recent client that we worked on who they've acquired three very similar businesses, but they all use different tech as their day-to-day tool. So can we build a dashboard that just needs to find out one KPI across those three businesses? Well, they all use different tools and they call data different things in those different tools. So we need to combine this data, we need to clean it, we need to agree to the governance structure, we need to identify data owners. We need to ensure the quality is as expected. So when you do present that KPI, you don't get one of the MDs saying Power BI is wrong. Yeah.

Mark Smith: 19:05
Yeah, I was involved in a project of doing a whole transition of software for a rail company and they just recently bought three or four other rail companies in different geographies, or four other rail companies in different geographies, and I couldn't believe that, even though these organizations are in the same industry, they had different terms and phrases and things for the same thing, and if one person in that role talked to somebody in another company and they used these words, for whatever reason, they found it very difficult to go no, that's what we call, or that's what we call blah, blah, blah. And so one of the first things was was building a common data model, right in a common language, that all acquired companies moved to and had an agreement on, so that they could both communicate effectively about the same thing and what their data was telling them. How do you do that for the customers that you work for? What's that journey look like for them, that you take them on?

Harley Webster: 20:09
I love calling it Babelfish. Absolutely love it. It's an absolute Babelfish. This is where the business consulting side comes in, mark, because you have to get people to agree to something and that's always very difficult, especially now post-covid over teams. So I do recommend going to hold sessions in person where you have to come to a common ground on these types of events, um. So we have to go to the stakeholders who can make a decision. So it might be you might have to say you know an opportunity where you find there's two mds in the same place. I'm going to grab you guys for 30 minutes. We need to agree on our businesses um regions so we can good example yeah, so we can create kpis based on our regions.

Harley Webster: 20:58
You're you're calling, you're calling this the regions. You're calling this the Northwest and you're calling this the North and you're calling this the West. So can we agree that it's the Northwest? Yeah yeah, these sessions have to happen. So you're all speaking the same language, you're all singing from the same hymn sheet.

Mark Smith: 21:15
Yeah, yeah. In the work you've done in the last couple of years, what are the biggest pain points you're seeing that organizations have around their data?

Harley Webster: 21:25
I think, apart from the, there is the low hanging fruit of. People are quite good in Excel and Power BI at a base level, I'm finding. So they're quite happy making manual reports in excel. They're quite happy maybe taking those data from various sources, downloading it as csvs, combining it all together in one excel and then putting power bi on it. I find that that definitely where I'm working on what I'm seeing in in the region I'm in at the moment, it's quite people are pretty good at that.

Harley Webster: 21:59
I think everyone is aware that automation and data engineering and data warehouses and lake houses exist, but they don't really know what it is and what it does and how much it costs. So today I was talking to a client who said look, we've got this oracle data warehouse, we've got microsoft fabric, we've got power bi, but we're downloading our data into excel and it takes us ages to um do this every week. How can we automate this? So that is what the challenges I'm seeing now. It's matured, I think, a lot in the last two to three years. And then you ask people okay, cool, yeah, we can do that. What's your data quality like? Oh, it's great, yeah, it's fine, yeah, it's really good. We're really careful when we type. And then you ask people okay, cool, yeah, we can do that. What's your data quality? Like, oh it's great, yeah, it's fine, yeah, it's really good, we're really careful when we type stuff in.

Harley Webster: 22:51
So have you got any data quality dashboards showing you what your data quality is like? One example would be I worked with a business who do waste collection and they have to start their shift, pick all their bins and end their shift, and you want to get some kpis on your durations on each day, on each round, and you've got durations that go on for days and days and days because no one's ever clicked end shift. It's things like that. So we create, you know, exception based reports, power BI, which are really valuable as well.

Harley Webster: 23:22
Where, for data quality, we're looking at the data you want to report on and then we're putting in these statements like only show me data that is wrong in this instance. So you're straight away finding the needle in the haystack, rather than having to go through tons and tons of data to find something to fix it and then you have to implement that into the business. It's all well and good having a dashboard telling you you've got this problem here. You've got this problem here. Your data needs correcting here. Someone's got to look at that and then actually take an action. So the implementation is extremely, extremely important implementation is extremely, extremely important.

Mark Smith: 23:59
That's yeah, you're so right. How do you balance technical and business acumen right? Because data to a lot of people and anything technical to a lot of business-minded people is kind of it's a foreign language and my entire career has been around. If I was to distill it down is that I can take tech concepts and turn them into business outcomes and I can take business needs and turn them into tech outcomes. That's my, my skill. How do you? How do you balance that as someone, um, that is highly technical, but know that sometimes you need to explain things in very non-technical terms and sometimes and you need to maybe do that two ways. You know, when working with teams of technologists, how do you balance it?

Harley Webster: 24:53
Yeah, absolutely. It's a really, really good question. I think you have to if you're anything like me, me, mark, and I'm sure lots of people listening. If someone told us to go and analyze a hundred things, we'd go and analyze a hundred things because we just love doing it and it's all useless because it's not actually providing any value to the business. You have to have your high level milestone and keep going back to that milestone.

Harley Webster: 25:20
What are we trying to do here? What is the roi? How is what we are going to do have an impact on the return on investment? So you know, you hire a project and you pay x amount. I guarantee that you'll get an roi on that, because I'll keep making sure we keep going back to that sentence every meeting, every morning, every single time.

Harley Webster: 25:42
This is what we're focused on. You can do a lot of cool things and get lost by doing. You know, oh, you might see an opportunity to do some machine learning on this thing, but it doesn't actually have any impact on the ROI. So you're right and I think that is to be honest. I think that is kind of something that I'm proud to say is one of my strengths, that I can always give it back to the business to say, look, this is how you're going to get value out of what we're doing. But yeah, you're right, struggling to explain things like the importance of Data Lakehouse to stakeholders has been interesting. Holders has been interesting.

Harley Webster: 26:22
What has really helped me and that what might really help some people listening, is if you go back 10 years, let's say what are we looking at to get a data warehouse in an organization in terms of investment and in terms of support staff on site? Because you're going to have an on-site server. You're going to have people on site that have to maintain that. You're going to be looking at, you know, if you're having data integration layers. You're looking at Denodo or you're looking at Oracle. You're looking at really expensive high commitment on-prem solutions. So you say to them we need a data warehouse. In those days that's going to cost $100,000 a year, let's say for a medium-sized business. Now we're talking about I can get you a Microsoft Fabric for you know, I think we need this many SKUs. I think we're going to need this much compute power. It's going to cost us $1,000 a month on the SaaS Wow, that usually makes people listen.

Mark Smith: 27:17
Yeah, you know what surprised me and what you said there was is that one, that fabric cost point very interesting, because I've heard people go fabric seems expensive, um, and I'm like, uh, do you know what is actually running in the background, like how much heavy lifting is done for you? And you just did a great comparison of what is done for you by fabric and therefore the price point is crazy good. The other thing is you said ROI from a data person and I'm like how do you quantify return on investment for an organization? Because I'm like that's like how many grains of sand are on the seashore? How do you quantify an investment, and you know, even at $100,000, I thought that'd be, you know, for a large organization in the multi-millions, you know their data strategy cost. And how do you then, you know, quantify to a CXO-type role the justification for this outlay which is going to might be a multi-year project before they actually start seeing some of the returns. How have those conversations gone down for you?

Harley Webster: 28:41
Yeah, fantastic. The obvious one is the efficiency gains. That's the obvious one that everyone goes for. Automation get data quality means you'll make the right decisions faster. You'll have faster access to data. You won't need to create these daily reports. You won't need to create these weekly reports. You'll have better KPIs that look forward, that predict, so there's an efficiency gain there.

Harley Webster: 29:09
We've done some maths before. I think it was around with one of our clients looking at a before and after shot. So this was their daily reporting load was around 30 emails, so 30 excel sheets they were making sending out emails. It's a relatively small client which has grown um and we looked at how long it took them to prepare those reports and then what we saved in terms of automation. Uh, it was something like 40, 40 faster, yeah. So there's that one.

Harley Webster: 29:44
And then when you're looking at much bigger organizations who need more than just an efficiency game, you can start looking at other areas of um roi, and one really interesting one is market share gain, or protecting your business from other businesses taking your market share Interesting. So if you are not a leader inside your business areas of data and your competitors are, they are going to be better decision makers, faster decision makers and they're going to take some of your market away. So if I asked you let's say, mark, you are CEO of a big company, and I'm going to asked you, let's say, mark, you are CEO of a big company and I'm going to tell you is there? If your data is absolutely bang on, and we are using AI and we've done the whole shebang, we've got everything implemented do you think there is a 1% chance that will protect 1% of your market?

Mark Smith: 30:49
That's an interesting question.

Harley Webster: 30:53
Let's go down. Do you think there's a 0.5% chance it will protect 0.5% of your market and if I say yes, then you can say cool, that's 50 million.

Mark Smith: 31:03
Yeah, yeah, crazy, right, crazy, very right, crazy, very, very interesting, very interesting space to be in. My final question for you is what happens in the scenario you go into an organization. They say here's our data and you do or don't believe them. And what I'm talking about there is that how do you really find out all the data in an organization, the piece of software that is not storing that data in a format in a location that's accessible? It's you know. You haven't bought it into a lake yet.

Mark Smith: 31:40
How do you kind of do a, a stock take of the data of an organization? Do you have like a methodology, a process that you go through to make sure that you have the data inventory down? There's not something hidden away that's kind of going to be the missing link into a massive project because, oh, we didn't consider that was important, we didn't, oh, didn't. We know you needed that. You know and people will tell you, you know. The other common one I come across is our data's clean. You're like, right, you know. Do you have frameworks that you use or patterns that you use to ascertain? One have we got all the data the organization has before we start getting the outputs that we need? And then two how do you ascertain the robustness or the data integrity that you're asked to work with?

Harley Webster: 32:41
Yeah, I mean your first point. I don't have a magic bullet there at all. And if we go back to the beginning of our conversation, I said I'm someone that likes to do things quickly and get results, so I look at the outcomes first. Tell me the top 20 KPIs in your business, and where does that data come from and where is it stored? Okay, cool, let's get those 20 done. There we go Results. Fantastic, We've got them done.

Mark Smith: 33:07
I love it. So you work from the end and mine backwards, absolutely.

Harley Webster: 33:11
And for your next question about the integrity and data quality and our data is clean. I mean coming from a consultancy. If you go into a business and you tell them, you know, I think we can do this in two weeks and on day one you realize that data is an absolute mess, then you've shot yourself in the foot. So we do like to do a little bit of a recce before we do a proposal. Yeah, and I did mention data quality dashboards and that is something that we will do straight away and we'll show the client.

Mark Smith: 33:41
Yeah, so have you got your own kind of tool set that you would run in and quickly, that you would get a sense check of what you're dealing with?

Harley Webster: 33:52
Yeah, absolutely. It depends what industry. We do some repeat work in waste industry in the UK we partnered with a brilliant company called VWS Software who make the main tool and database for a lot of the waste management companies in the in the uk and I believe they've just expanded to australia. Um need to get to your islands next um, we're a very small market yeah, um, so they're repeat um.

Harley Webster: 34:21
We have repeat clients and repeat industries so we can go in with a template straight away. We've already made the views, we've already made the dashboards. We can go in and go boom, your day is 60 clean. So we're going to need a week to clean your data before you get actual results. Awesome, um, other industries it's obviously more difficult because you need to learn the business. You need to learn what data you're looking at, what limits to set in the data, um spikes, to look for this type of thing. But with the user interface you get with Fabric now I wouldn't need to do anything like that in Power BI. I would do it using SQL or Python, just in Fabric, yeah.

Mark Smith: 35:04
Harley, this has been so interesting talking to you. I feel like you'd be a good person to sit down and have a drink with to the fat over war stories. Thank you so much for coming on the show.

Harley Webster: 35:16
It's been a pleasure, mark. Yeah, absolutely fantastic, really enjoyed it, thank you.

Mark Smith: 35:21
Hey, thanks for listening. I'm your host business application MVP Mark Smith, otherwise known as the NZ365 guy. If there's a guest you'd like to see on the show, please message me on LinkedIn. If you want to be a supporter of the show, please check out buymeacoffeecom. Forward slash NZ365 guy. Stay safe out there and shoot for the stars.