Talking Channel #5: AvePoint and Purple Frog Systems

This month's Talking Channel is an AI special, with Will Garside joined by Alex Whittles (Purple Frog Systems) and Christopher Shaw (AvePoint) to discuss the channel's maturity and where partners are seeing commercial success. 

Transcript

SPEAKER_02 0:03

Hello, and we'll go outside editor here at IT Europa and welcome to Talking Channel. We've tried to avoid it, but the inevitables happen. This week we are talking about AI with two expert guests from the community. Alex Chris, thank you for joining us.

SPEAKER_00 0:20

Pleasure. Thank you very much. I mean the experts are stretched.

SPEAKER_02 0:23

With two guests from the general community.

SPEAKER_00 0:26

Here's awesome opinions. My name is Chris Shaw. Lovely to meet you both. And I am the channel director at Av Points, looking after MSPs and resellers within the partner world, predominantly with cloud-based solutions and ecosystems.

SPEAKER_02 0:40

Wonderful.

SPEAKER_01 0:41

And I'm Alex Whittles, founder and CEO, principal consultant of Purple Frog Data Analytics. We help customers with their data needs, which these days, obviously, AI is a huge aspect of that. Implementation, consultancy, advice, and yeah, how to leverage data and AI to achieve wonderful things. I've been a Microsoft platform MVP for 11 years, so uh mainly in the Microsoft space, but uh yeah, data.

SPEAKER_02 1:02

We're never going to solve the AI conundrum in 23 and a half minutes. But let's do it going.

SPEAKER_00 1:08

Do our best.

SPEAKER_02 1:09

Let's let's let's talk from different perspectives. Because AI means different things depending on where you are within the community. You guys, early investor in that types of technology, talking to a part communities around the data side of it. Where are we now, from a vendor hat on, in the AI journey?

SPEAKER_00 1:26

Yeah, I mean it's the term AI itself is very broad, right? We know there's a huge, huge, huge range of technologies, conversations, products, whatever it might be, that do something in this space. And what we see is end users go, I need AI. And partners go, we need to sell AI. And we as vendors go, we've got to get on the back of AI and help with it and move it. But I think for me, what I hear and what I see predominantly in the market is a need to understand how that relates to real-world outcomes and how you can you actually translate AI from this really useful concept of oh AI can do everything for me, it can automate, it can answer, it can you know do a thousand or things and make my toast in the morning to actually driving it. Well, actually, I need to develop some ROI against it, or I need to actually have a real use case against it. And I think the people and the companies, whether they're partners, users, vendors that do well in that space, are the ones that have a clear message around it. And from my perspective, I have point that's specifically around data, cleanliness of data, who has access to it, who can control it, and ultimately what do you want the AO to interact with that data in what format. And that generally tends to be our entry point, um, how we help people understand what AI can do for them for their outcomes.

SPEAKER_02 2:39

Yeah.

SPEAKER_00 2:40

Right, and I think that's probably the the the shift in perspective and the shift in um not just selling but engaging, understanding, you know, building a use case for it.

SPEAKER_02 2:50

Yeah, I couldn't agree more. I think you deal with enterprise customers, you probably I saw you nodding your head when Chris said, Oh yeah. We had a conversation about outcomes rather than inputs. Where are we now in the AI journey?

SPEAKER_01 3:02

Yeah, we see uh probably the biggest or most common question that we get asked is from board members saying, I'm I'm being pushed to have an AI strategy. Help me implement AI. Well, what problems do you want it to solve? Oh, we don't know, we just need AI to do something. I've got to tick a box. And that's why if you look at Forbes, 85% of AI projects fail. Yeah. Why is that? Well, because you mentioned it, lack of data, lack of clean, consistent, governed, secured data. Um, but also AI for AI's sake. And that is a dangerous path to go down. We take the opposite approach of don't try and transform your entire organization, don't just put it in place as a tick box. Identify business problems, identify small areas of bottlenecks and challenges and frustrations and tackle those. Implement it bit by bit in small, manageable, low-hanging fruit projects that build trust, build confidence, build an understanding of how it can be used in a business to effectively improve and make it more efficient. Don't try and rewrite your entire business.

SPEAKER_02 4:00

Yeah.

SPEAKER_01 4:01

Take them all step at a time.

SPEAKER_02 4:02

And it's not like the I'm I'm long and ugly enough to remember cloud as being the first second coming, if that makes sense. But that was a very simple business case and cost analysis. We have a load of servers in our data centre which cost us X. Let's virtualize and move those servers into whoever's cloud. That costs us Y. This is the return, these are the advantages, these are the security advantages, these are the scale advantages. Really, really simple. When we're having the AI conversation, this is not, as you said, not a simple conversation, but also that's there's no real at this moment in time, true cost base for AI. We all know that every time you do an AI search, someone's losing money. I mean, is this sustainable, the current AI mindset?

SPEAKER_00 4:53

I think it's a really good analogy, actually, to link virtualization and the move to cloud. People are still physical, right? There are still people that sit in physical because it doesn't work for them.

SPEAKER_02 5:01

Yeah.

SPEAKER_00 5:01

AI is the same principle, right? It's not going to work for everyone. And as you said a minute ago, you've got to understand the use case, you've got to understand the business drivers behind it. And I mean, if you're if you're being ultra-analytical about it, you can potentially build an RMI case, agenti KI. You can look at that for an example. I need to automate a business process. I need to move something from manual to automation. It costs me X amount of money per month, yeah, whatever measure you want to have. Does this AI, agent KI save me money? And you can look at it through that lens and actually build a relatively robust use case against it. But there are steps you've got to get to before they show the identification of that need and the understanding of the people you're going to give this AI to and the repurposing of that resource. Is it actually cheaper to have AI versus an FTU resource, right? These sort of questions I think are still being addressed, still being looked at, certainly. We have a lot of conversations along those lines. Um but again, there's not one size fits all, right? You can't have this huge square peg going into what sometimes is a rather smaller triangle hole, right?

SPEAKER_02 6:01

And it was easier with things like robotics. If you look at the car industry, this is what a person on a production line cost. This was the capex of a robot, this is what the robot did, and this is how efficiently it did it. It never went on strike, and you could upgrade it every six months, and the first five years of large car manufacturers moving to robotics was actually very expensive. The next two decades paid for themselves with more efficient lower cost cars. Great example, automation solved the problem. AI at the moment is if you look at the hundred something plus billion that's been invested in AI, I'm paying £20 a month to use an AI service, which does not cost £20 a month. How is this sustainable?

SPEAKER_01 6:48

Well I think it's stick with the same analogy of robotics. If you look at Tesla when they first started, they tried to automate too much. And that actually caused massive problems with their production capability and throughput. They had to rein in their expectations and bring people back in to achieve their goals. It's the same with AI. Don't use it to replace people, use it to augment and enhance. There's two aspects of the of the pricing model you mentioned. One is, is it sustainable from the AI hosting providers, from OpenAI, from Anthropic, from Microsoft AWS? No, it's not. It costs them a lot more to run these tools than than they're charging for it. But it's a loss leader to get you hooked, get you uh get you in, and to uh to to wait for other competitors to leave the market. Um a customer perspective, we can leverage that to use this cheap AI capability now to radically reduce the cost of proof of concepts and prototypes to actually explore how it can make a change to the business. If we try that in two, three years' time, it's gonna be a lot more expensive to do. So I think there's a huge opportunity to leverage that cost at the moment. Go back to the cloud uh analogy you were talking about. Most of the customers we talked to back in the in the cloud migration stages were we want to save money, let's move to the cloud.

SPEAKER_02 8:00

Yeah.

SPEAKER_01 8:00

How many projects did we get were involved with that actually ended up saving money? It didn't. It was more expensive to go to the cloud. You move CapEx to OpEx, but it's ultimately more expensive. The real benefit was in the additional capability, the additional uh HADR, the additional robustness, the scalability, the flexibility to scale up and down. And I think that's where we need to focus the AI conversation is it's not necessarily on am I going to save money directly, it's can I improve the efficiency and productivity of my organization, both in resource and decision making, to uh improve our productivity while spending more money at the same time. And so it's not about cost saving, it's about spending it wisely.

SPEAKER_00 8:42

Just if I could add, it's about changing our behaviours and how we interact in the working world as well, right? We're not going to be able to use these tools and do the same things we did without them. And as I say, I mean Microsoft is a great example of copilot, right? It's not there to do it on your behalf, it's there to give you a steer and be a bit of muse and help you take these big concepts and make them quite digestible and understandable, right? I think if we just look at copilot individuality, um people use it to stimulate ideas. You've got an idea, you don't quite know how to articulate it, you don't quite know how to get it from A to B. Can you give me some help? Can you give me some ideas? Can you give me some structure? Can you give me some, you know, ways of framing this? And that behaviour change from the user, you know, how would they have done it a year ago, two years ago? They would have just sort of probably gone and asked a colleague and drained their time and stopped their productivity, right? But still the the key thing here is how we're interacting with it and how we're going to adapt, and how we as users and organizations, whether that's an end user or a partner, how are we going to make sure that we are getting the best use cases out of it? And are we doing the right things to maximize our opportunities with it? You're absolutely right.

SPEAKER_01 9:47

And I think one of the one of the key challenges we see is are you using AI? Oh yes, we've rolled our co-pilot. Well, that's like buying a smart TV and watching BBC One. You've got a huge opportunity there to do a lot more with AI, and copilots or ChatGPT or uh or Claude, they are one aspect of a very simple entry point into helping do something. Most customers I speak to don't understand the concept of creating your own AI skills to actually, if you think of AI as a well-intentioned intern, if you ask them to do a job, they're gonna do it their way, not your way. How do you guide that? Well, you write them a specification, you write them a guide on how you want this task done. Take that, turn it into an AI skill, give it to Claude or Copilot, and it's gonna be able to do the job better, aligns to your business.

SPEAKER_02 10:34

So, what's your your views on? I I had a conversation with it was Anthony Dobson at Arrow, and he made this absolutely fantastic point that the vendor community is building AI into their products. They're just doing it. They're doing it to make their products more intelligent, stickier, and to solve problems that they had that are easier to solve in AI than they are in hard engineering or building code and modules. And his point was, which I thought was really interesting, it doesn't actually require the enterprise customer to have an AI strategy to solve Challenge X. If Software X does it anyway. And his his view, which was again quite nuanced, was the AI revolution around let's build it ourselves is the opposite of where the software industry has been heading, where if you look at enterprise customers now, they do not build their own proprietary code bases. Because there's no point. A bank does not want to be a software company, they want to be a bank. So my argument is this why are we expecting enterprises to deploy and build these AI applications when for the last two decades we've been saying to them by SAP, by PeopleSoft, by Oracle, by Microsoft, don't build applications because there's no sense in it. I'm curious, you're an MVP, too. Maybe I should ask you.

SPEAKER_01 12:03

So it comes back to what is the problem you're trying to solve. Um, if you've got a business problem, you need to solve it, you've got a choice, buy or build. Same as you always have done. The difference was a few years ago, building your own software solution was very expensive. Um, yes, you could outsource it, you could offshore it, but it was still expensive to manage and maintain. These days, the capability of tools like Claude Code uh is quite phenomenal. So you can now build things much more cheaply, but there's a danger and a risk. You've still got to manage that and govern it yourself. But also, where you've got a very tight integration with AI capability with an existing solution you've got in place, it makes sense to add on the AI capability. AI capability in HubSpot, for example, it works pretty well. You pay for it, but it works. Whereas if you're looking more agentic AI, um, we've just rolled out a new agent in our business, where when someone logs a support ticket, the first thing it does, the agent picks up that that ticket, looks back through the history of uh tickets for that particular customer. Is this a repeat question? Is there anything else we've already answered that can help uh solve this question? Right. So our support team can look at that and instantly get a summary of what state that customer's in. They can respond much more quickly and accurately. Now that's not available off the shelf. We can build that in a day.

SPEAKER_02 13:21

But that feature will be available off the shelf. I I've I've spoken to the greatest example is Adobe. My son's doing a creative course at a university and he showed me the latest versions of Adobe, which are insanely intelligent and they have AI within them. He has no interest in writing code. He just wants to make this picture look like that picture. That capability is built into the core platform. You guys have been building AI into some of your core platforms for the benefit of your partners and customers. The partners don't have to then do that. I mean, vendor hat on, do you see any vendor that's not building AI capability into their product?

SPEAKER_00 14:04

I think the good vendors understand how their technologies link into that model.

unknown 14:09

Right?

SPEAKER_00 14:09

You don't try and shoe on it in. The ones that go, oh, we're we're AI ready, we're here, we're gonna do it. And you sort of scratch beneath the surface a little bit and go, well, how do you use that? And they go, Oh, we just asked Copilot. Like similar sort of things, right? They don't quite know how to link this huge, huge, huge trend to their stack and their technologies and their outcomes. And the ones who do it well are the ones who say, well, actually you can automate this or optimise this or you know, change a process here, right? And um it's it's that it's that understanding what us as a vendor can do in this space, right? So for us, just as an example, like you're controlling, you're governing, and you're understanding the risk, right? Because these are risk technologies as well, right? You can't get away from that fact. You can't let everyone have access to every bit of data internally, and you also can't let people start throwing data externally into the free versions that are out there. So, how do you actually control that risk? How do you manage it? How do you ensure that you're not exposed as a business? And that's where we as a vendor start, particularly just as a slight example. Um, how do you ensure access control? How do you ensure history tracking? How do you ensure the right data is returned from the right sources in relation to the person that's asking as well? Can HR actually have you know access to sales? Can sales have access to everything, right? Do you need that link, right? Do you have the gate sort of do you have the control? And so I think understanding what you do as a vendor and how that actually fixes some of these underlying problems that maybe the end users haven't seen yet. Those as you the example earlier, they just don't need an AI strategy. Well, what does that mean? Have you thought about the risks behind? Because it's the same as any technology platform, it's the same as any software platform. You need control, you need to understand what you're using it for and what the outcomes will be, positive or negative to that.

SPEAKER_02 15:52

It's a bit Wild West at the moment, though.

SPEAKER_00 15:53

Somewhat, yeah.

SPEAKER_02 15:56

And it's missing. But it's it's missing because it's not required. Let's be very honest. PCI DSS is a requirement if you want to transact credit cards. Um NIST 2 is a requirement if you want to I don't actually know what it's called critical infrastructure or something like that. So there's this regular regulatory requirements around things. There is no regulatory requirement around AI Act.

SPEAKER_01 16:25

It's in force for the critical, very impactful areas.

SPEAKER_02 16:30

It's 0.1% of the of the number of organizations in the EU.

SPEAKER_01 16:36

But it's still mandated that it has to be in place. So the the the big area is the high-risk area of AI, uh, which won't come into force until I think this August. And that is that controls uh covers things like where an AI can affect someone's socioeconomic status. Hiring, firing, health, things like that. Um we're seeing a huge rise in uh recruitment using AI. People turning up for job interviews, and they're being interviewed by an AI boss. That's going to change because of this EU AI.

SPEAKER_02 17:06

I I I have to be honest, I think the EU AI Act is probably unenforceable. Simply because of how things happen. The good example around their regulatory framework suggests that phrase, economically impacted. It's like saying economically impacted by what? Electricity or steam power or it's it's well-intentioned but unlikely to provide any type of framework for safe operation of AI.

SPEAKER_01 17:38

See, the challenge with AI is it's changing every month. Every month there is new capability, new tools, and new scare stories as well. New ways to break it. 20% of our entire company's workforce spend their time researching, learning, and training on new skills. Because they're moving that fast. So how can regulatory framework ever keep up? But it's a starting point.

SPEAKER_02 17:59

It is a starting point.

SPEAKER_01 18:00

And when you've got fines up to 35 million euros, it actually makes people stop and think.

SPEAKER_02 18:04

Yeah.

SPEAKER_01 18:04

And it'll evolve over time as to how that's going to be uh be managed and monitored.

SPEAKER_02 18:10

And it would matter more if any of the AI pioneers were based in Europe, which would which unfortunately so let's talk a bit about the endgame. Um if you look at cloud, ended up being three large US companies and somebody in China, which is always a caveat. Is AI going to go the same way? Is AI gonna end up in the hands of three, maybe four monolithic companies? It's a difficult question.

SPEAKER_01 18:33

It has to. The cost and size of data centers, GPUs, RAM, we all know the problems at the moment with the supply chain of hardware. Um, the sheer scale of that means small players, unless you have massive investment from venture capitalists, how do you scale to the same and how do you compete? You mentioned earlier about the the price pricing problems. Everything's uh at a loss leader at the moment, that's not sustainable. And so all that's happening is the big players are going to wait until the small players leave the market and can't be sustained or get acquired by the big players.

SPEAKER_02 19:03

Is that good for the channel community having a duopoly?

SPEAKER_00 19:06

I think duopoly, not duopoly, a quadropoly, the phrase is I don't think there's a clear answer to that actually, and it's a bit of a cop-out answer, I appreciate, but I don't think it's gonna necessarily be good or bad, because again you come back to the same thing that people still don't know how to use it, right? People still don't know the outcomes they're gonna get from it. So, how can how can any organisation know if it's bundled into Google or Microsoft? You know. Is that is that a good or a bad thing? Is it a good or a bad thing? Really, most people only really have one of the big three hyperscalers now, right? Is is that a good or a bad thing? And they're not much of a mushness, right? They are uniquely different and they have skills and special areas that they excel in in each of these particular areas, but it's it is an element of picky colour sometimes, right? And it's the same as software at the moment.

SPEAKER_01 19:53

When you pick a software vendor, a SaaS supplier, you don't know if it's hosted on AWS or GCP or Azure. You know it's one of those three.

SPEAKER_02 20:01

Do you care?

SPEAKER_01 20:01

Do you care? Exactly. And AI is going to go the same way where the actual hardware infrastructure, I think, is going to standardize uh among a small number of players, but the value add on top of that is then going to be a very big market. Yeah. But they're not hosting their own hardware, they're using the capability of the big players.

SPEAKER_02 20:20

Absolutely. Yeah, I mean we had we had a luncheon recently and we talked to a group of MSPs about what they were doing for AI. And there was a lot of a lot of opportunity and optimism around AI, but very little practical steps forward for an MSP or service provider to deliver something tangible in that space. There were a couple around the table that were. If you were going to give a piece of advice, because we're nearly out of time, to an MSP that could do something this year tangibly, so that when they get the question from the board, what's your AI position, that they have a competent answer. What would that be?

SPEAKER_01 20:58

I'd say there's two things. First of all, the easy routing, uh, we actually wrote a starting guide for the uh for GTIA uh on how to think about getting started with an AI strategy. Uh GTIA have just launched a um a data and AI interest group um available to all members with regular workshops and uh webinars about this exact topic. Uh but the single biggest thing I would say is identify a business problem that's frustrating and slowing you down, and make sure you've got the data infrastructure to support.

SPEAKER_02 21:28

Something internally or something for a customer. So start with your own house first.

SPEAKER_01 21:34

Yeah, you've got to have that data. You imagine earlier, Chris, you've got to have that data foundation solid, trustworthy, and government.

SPEAKER_02 21:40

So an internal project would be a good way to start that process off before you go externally.

SPEAKER_00 21:45

I I actually agree. From us as a sales organization as well as a vendor, because we all are sales forks, right? We are supposed to help you identify your pain points. You have this problem to fix, it might not be the technology to fix it you think you start from, but it we're gonna get you to. This endpoint, and it might not be because of a financial outcome, it might be to actually genuinely fix that pain point. And can you articulate that in your own mind, Mr. Customer? Can you understand how we're getting you from what you're doing today to where you're going tomorrow and where you want to be tomorrow? These are the big questions the vendor should be asking. And that sales process, that engagement, that articulation of the thought and the ability for us as a vendor to deliver that solution, these are the things that we need to get to grips with independent community.

SPEAKER_02 22:27

I'm gonna give my toothpaper as the final call. I about six years ago, five, six years ago, I did a day's training with an AI expert who at the time we always joked that he was a wizard because he was doing this training and showing stuff on a very early build of ChatGPT that blew my mind. And it was a whole day where we could ask him questions and he'd give us little examples, we did little uh tests of our own, and we spent this just one day learning about the AI fundamentals, we learned about like these how it works. And the best thing I ever did was provided to me by a company I used to work for at the time, and at the time I didn't really have an application in my head that I would use it for, but it gave me this grounding, I think every single MSP to take the senior leadership team and maybe a couple of people, the departmental heads or keen enthusiasts, and get some paid-for structured training. Because at the moment, just playing around with Chat GPT or something co-pilot is just not the same. Get someone to teach you the fun same way we got sales training or technical training or Microsoft training or Cisco training, get some training, and from there you can start to have sensible decisions. Until you do that, you're quite you're not actually committing, you're just playing around the edge, and it's gonna cost money. Let's be very honest. Good training is not free, but it's probably worth every single penny. Right, so that's it. We've just that's we've done AI. We've unpacked AI in 25 minutes. So, Alex Chris, thank you for joining us and thank you for joining Talking Channel on our AI special. We'll see you soon.