How generative AI and automation improve customer experience without losing trust, privacy, or the human touch—practical AI strategy for growing businesses.
Primary Growth Pillar: AI & Automation
AI is not here to replace people — it removes repetitive work so your team can focus on real relationships.
Colin Mackenzie helps businesses use AI and automation to improve customer experience without losing trust, privacy, or the human touch.
In this episode of Knack 4 Business (Season 4, Episode 17), host Bernie Franzgrote and co-host Wayne Pratt sit down with Colin to answer a big question:
How can you use artificial intelligence in business without sounding robotic or losing customer trust?
Colin explains that AI is not magic. It is math, data, and models. Used the right way, generative AI and large language models can automate FAQs, scheduling, and contact center tasks—while humans handle empathy, trust, and complex conversations.
Key topics covered:
Who this podcast is for:
Startup founders, solopreneurs, scaling business owners, operations leaders, and IT decision-makers exploring AI & automation.
What you’ll gain:
Actionable AI implementation ideas, clear explanations you can repeat to your team, and practical steps to improve customer experience while protecting trust.
Watch / read:
Guest link:
Sponsors:
With thanks to:
Read more on Ghost, then comment with your biggest AI question. You can also connect with Colin or message us at info@kreativinsight.com . For more episodes and tools, visit Knack 4 Business and subscribe.
Bernie (00:05)
So question to the audience. How can you use AI to improve your customer experience without losing the human touch? That's exactly what Colin will unpack in this episode. He'll share.
how businesses of any size can turn complex AI tools into simple human-centered solutions that create real impact. From contact center integration and intelligent agents to knowledge retrieval and data sovereignty, will reveal practical ways to harness AI innovation beyond the hype. And our guest today is Colin Mackenize And he's a seasoned AI applications innovator who spent over three decades turning complex tech into tools that feel simple and human.
From building early hospital networks to shaping global multimedia services, he's led major advances in cloud communications and artificial intelligence. His work includes filed patents in multimedia tech, teaching advanced computer science, and guiding organizations through the shifts to cloud automation and now generative AI. Colin focuses now on intelligent agents, contact center integration,
knowledge retrieval and data sovereignty. That's gonna be an interesting one. We're talk about that. And with Elvis and I, I'm making customer experiences smarter and more meaningful. He's known for blending deep technical insight and practical wisdom. He's a trusted voice at conferences, workshops worldwide. Outside of tech, Colin loves to travel and once toured as a professional musician. Colin, welcome sir. How are you? And do you have your favorite quote or saying you'd like to share with us?
Colin (01:40)
Well, I'm a philosophy fan, so I always, ⁓ I like that one quote from Socrates, the only true wisdom is knowing you know nothing. But anyway, I like it because it's, know, sort of intellectual humility, you know, being curious and open and instead of thinking that you have all the answers. And coincidentally enough, speaking of travel, I was actually lucky enough to visit Delphi in Greece.
a few weeks ago, which is where Socrates started off his journey of questioning. So it might have been thousands of years ago, but it still kind of feels just as relevant today as it was then.
Bernie (02:12)
No hemlock for you.
Colin (02:13)
Not today.
Bernie (02:14)
That's right. So it's interesting to hear someone talk about blending the human connection and cutting edge technology. And you know, every once in a while you see something new coming up and going, how does that work? And usually it's the prototype and then when it leaves a few cuts and edges and bruises until you get to the final, you know, end good result. What drew you into that space? What made you go the user experience?
Colin (02:40)
Well, I mean, I've always been.
sort of fascinated by the intersection of people and technology. I my first roles were in hospitals, which are less about machines and more about how to connect people together better, doctors, nurses, patients. So technology is just the enabler in that context. So I think that carried through my career. So whether it's multimedia, cloud, or AI, the question is still always, how can this make life easier, more human, more connected?
Wayne (03:09)
There is a lot of people talking about AI that don't have a clue what they're talking about. What are some of the misconceptions that you've seen that you'd like to clarify for us?
Colin (03:20)
Well, I mean, there's a couple that I always sort of look at a little bit sideways. I mean, there's the idea that AI is going to take over everything like Skynet or the other end that is just hype. I mean, the truth, of course, is always somewhere in the middle. It's powerful, but only when it's applied thoughtfully.
I think another one is that AI is magic. It's not, it's math, it's data, it's models. Without guardrails, context and integration, it could be as wrong as it is right.
Bernie (03:46)
It's interesting. We've had conversations before this call, before this podcast, and you were explaining to me that it's almost like the genie in the bottle. He's in tiny space. Big universe can come out. And the reason I know, it's right out of Aladdin from Disney. But you've described something that you were explaining how AI can fit inside your phone.
cruise around with and yet it has a lot of capacity. for me, prior to that moment, I always thought there had to be this large server farm breathing fire, you know, to make this come together. And you're just telling me, it's raining computer, a little small, you can have a large language model kind of cooking and baking for you. What's behind that?
Colin (04:29)
Well, I mean, like I said, it's not magic. It's just math, data and models, right? And again, you get into the math side and you could have a number with 10 decimal points. You can have a number with three decimal points. You're just kind of changing the precision of it a little bit. But as you can imagine, know, the less absolute you make the numbers, the more of them you can kind of fit into any given space. So, I mean, they've done some phenomenal work in the last couple of years of taking these big, huge model
I
LLM, know, generative AI models, the chat GPTs of the world, and shrinking them using some clever math, you know, I mean, things like precision rounding of numbers and stuff like that, but basically getting them into into a footprint that can fit on your phone. So a lot of the new phones that are, you know, coming out, the new operating system updates, they're literally including, you know, LLM models that are similar to the chat GPTs of the world that you're seeing on
the cloud, but they're working on device and on the phone. Which is not only, I think, great from, you know, obviously very responsive, you know, can get answers to stuff really quickly, you know.
But I think beyond even the technical capabilities it gives you, there's a huge advantage just to privacy. The ability to have something done on device where it doesn't have to go out to the cloud, having all that data travel around the cloud and get absorbed into everybody's data centers and resucked into training the next model. There's something that I think is a good step for the industry to take because there's always
bit of a trust issue when it comes to, know, where is my data really going? You know, so keeping it on device is a valuable thing from that point of view as well.
Wayne (06:11)
So how do you keep things secure when it goes back into an LLM?
Colin (06:16)
Well, mean, so first of all, by having something on device, it never has to actually leave the device. So that's obviously one way. But I mean, that is one of the risks. Like, that's one of the areas that.
you know, that came up about data sovereignty, you know, before. there is a challenge with when people are using these, again, the chat GPTs of the world, you know, these generative AI services, you know, you're throwing stuff up into the cloud, that content, you you might've thought it was private, but maybe it includes some kind of personal information, maybe it includes, you know, some patient care, some ⁓ proprietary company information, that's now all gone out into the cloud and that can then be used to
to sort of reinvest in training the next model down the road. But there are, you know, there are approaches that can mitigate that. So I mean, one of the main ones is, you know, you can pay a little bit more for licensing. So you pay a higher monthly fee instead of a lower one, get an enterprise account, so to speak. And you get a guarantee that the data that you submit is going to be kept private and not used in that way. But what a lot of companies are moving towards is a little bit more of that kind of that really private data
sovereignty approach where they'll basically install their own models. They'll basically get their own servers and their own data centers. They'll install their own kind of AI engine to shovel all of this stuff through. And then it never actually has to leave their data center in the first place. It doesn't even go out to the cloud. Obviously there's more work and effort involved in doing something like that than just taking advantage of the things that are out there all in the cloud. But a lot of companies are starting to find that
because customers don't always trust AI unless they trust how their data is being handled. And it's not just the customer or the company side, but regulators around the world are starting to catch up and demand that as well. So sometimes it's even a compliance requirement, not just a nice to have.
Wayne (08:05)
I find very interesting that I like to talk to professionals about, if you are 60 years old, you've seen a telephone tree where you saw technology being used against you to try to actually talk to a human that could help you. Now that we have billions of dollars worth of hardware and some very, very smart minds, what do you think is actually going to be implementable and makes life better instead of just different?
Colin (08:32)
Well, I I think that comes into the conversation around why are you automating things in the first place if you're trying to make the customer experience better? mean, obviously people are trying to maybe reduce costs. They're trying to push things to a self-service.
maybe an AI agent because it's cheaper than a human person to be on the end of the phone. But at the end of the day, the customer service has to drive that. it shouldn't be the case that people are kind of forced into that model where they have to talk to an AI agent. It should be more about...
Where do you draw the line with the right kinds of things? What are the right low-hanging fruit where it makes sense to do something where maybe a user or a customer would rather just get some piece of information quickly? They just want to get an account balance or something like that really fast. Do that automated, move those things ⁓ to an automated model where AI can help, but then keep the more...
the more relationship building sides or areas where a customer really wants to have a conversation or a connection or when that's important to have that established with your users. Keep those things with, because a human being does clearly much better at that. So I think it's more about where you draw the line as a company to the kinds of things where you want to use automation to improve. Use it for the low hanging fruit where somebody just wants a quick answer. Don't use it to try and replace the sort of authentic
and customer connections that I think are valuable for all companies.
Bernie (10:05)
You mentioned something earlier about AI being in house, The needing servers. But some of the utility of having AI is being able to pick up information that's not in the house, because it's out in the wild, right? Whether you're referencing Wikipedia, which is not necessarily accurate, or you have, there's another data set out there that's more reputable. Not that Wikipedia is reputable, it's more,
verified. How do you take the two sets of information, so like one internal data, which is company secrets and company process versus what's out there so that you can pick up the best of both worlds? Is that something that's doable?
Colin (10:48)
absolutely doable. Yeah.
So I guess there's maybe two ways of answering that. One might be information that's out in the world. And I think one of the nice things is that a lot of the models that even if you're going to, even if you want to deploy something to keep it private, you you want to do something on premise or in a data center and kind of keep everything so walled off from the rest of the world, you're still basically leveraging models that have been trained on data sets using the outside world up until a certain point. So even if you go and install one of those today and you put it on a server and it
data center, it still has a pretty comprehensive set of real world knowledge to draw from. But a lot of what's being done, because I think it's even an inherent thing with large language models to begin with is...
they're not deterministic. we've all seen and heard about generative AI hallucinating, making stuff up. It doesn't very, very well. It's very convincing what it does.
But there's always going to be scenarios like that where it's not just a technical issue. In a lot of cases, you have to be able to trust that some of the information that you're relying on is not just made up or generated out of an LLM, but it's grounded in that real world knowledge source. And what a lot of people have been doing on that is basically combining different sets of information. There's a common acronym called a RAG pipeline, a retrieval augmented generation, which is basically
a way of saying, hey, somebody's asking a question, why don't I take that question and then look in my, know, dip into my real world knowledge source first, and then go and figure out based on the things that I found there, now how do I put these pieces together, maybe synthesize it, you know, put it into a nice answer to send out. So it's kind of a blending between traditional generative AI, LLMs and knowledge sources, which, you know, there's kind of newer technology that's being used.
to that easier to store, easier to kind of ingest stuff into and very, very quick and rapid to be able to get things out in a way that's usable in that kind of pipeline.
Wayne (12:46)
Colin, we've been hearing a lot in the last couple of years about AI and there's everything from it's going to create an underclass where we're all going to be, you know, starving and cold. And the other is we're going to be doing so well and be so productive. We'll have people feeding us grapes as somebody plays the loon on our island. In the short, medium term, three years or less, what do you see actually happening?
versus what people are saying in either of those two ditches.
Colin (13:14)
Well, I mean, it does go back to my comment about, you know, how sometimes you have people that are thinking that it's going to be one extreme or the other and the, you know, it's not going to be Skynet and it's not just hype. The answer is always, you know, somewhere in the middle. But I think, you know, I think the reality is, that AI is a tool. So, I mean, even in the next couple of years, it's really
It's not about replacing everything. It's not about replacing people's jobs. It's really more about augmentation. So I mean, if we think about this, like in a customer service kind of world, how do you help somebody find an answer faster if they're on the phone with somebody or draft a response or how do you make a self-service interaction more conversational? So it's really about kind of augmenting what people are doing in the short term.
And, you know, I mean, in the long term, I mean, there's newer technologies that will that will get more autonomous and be more able to help kind of, you know, automate more sort of aspects of what people are doing on a day to day basis. But it goes back to that human connection. You're never going to replace that. It's more about replacing the low hanging fruit that that people don't necessarily want to do, that it's repetitive and then
pushing businesses and users a little bit more to have the customer connection and the personal connection on the things where it matters, where it's complicated, where you need to build those relationships together. And if somebody just wants to find out your business hours or get an account balance or things like that, sure, leverage technology to do that. I think those are the areas that it will take more and more of. It'll become part of that part of the ecosystem a little bit more thoroughly.
Bernie (14:54)
Right now, most people, when you say, you mention AI, think of one of three things. They go, hey, pick the name. It's say it right now, but I get a prompt back, right? We all work here. Oh no, they're all talking together. That, or you think of, I've got to put a question in and I got to do some prompt engineering, which means there's a of a chatter or a chat bot.
But I think there's a lot more under the hood. Where else is AI making an actual impact, positive impact? And kind of where's it going next? Is my car gonna know, Bernie, you've come back to the car? Yeah, yeah, I've unlocked the door for you. I recognize your footsteps, padawan.
Colin (15:40)
Well, there is a
You know, I mean, it may do that too. But there is, I think there's some really cool areas where it's going. I mean, we have gone from a world where AI was used to do little bits and pieces here or there. It was almost like little spot technology fixes in places. It's become a little bit more prevalent through areas like conversational AI and bots.
the generative AI, the chat GPTs of the world and everything that our devices will answer for us are getting smarter and smarter. But I think the more interesting direction is some of the things that are starting to come out around what's called agentic AI. So to me, that's kind of the cool thing that's coming up in the future. I mean, it's here now, but it's not as maybe as prevalent.
And so I think what's really cool about that is, you know, we're dealing with systems that aren't just responding, you know, you're not just asking a question and getting an answer back. They're not just anticipating something, but they can act autonomously. So they can, they can chain tasks together so they can solve a problem end to end. You know, if they know that they have to do five things in a certain order to, solve a problem for you, they can go off independently and do those five things and then come back after and say, Hey, I've done this whole thing.
for you, I've solved it. And I think that's sort of a cool thing, especially like for smaller businesses as well. mean, because that's
It's just leveling the playing field. You can have an AI assistant that can handle more of the routine work at that level than you would have had otherwise. And so the owners are free to focus on trying to grow their business and relationships and stuff like that. But the agentic AI, I think, is the cool area and the direction that it's going in.
Wayne (17:31)
On the video, we can see a whole bunch of guitars. So I'll ask you the musical question. Now that we have all this new technology and we have things like Auto-Tune, and we're all romantic about what happened in 1966, is this technology improving us, destroying us, letting us pick more carefully? Where are we?
Colin (17:57)
⁓ Yes.
All of the above. So it is, you know, and it's very interesting as well. I mean, there's a bunch of stuff that's in the news about about even artists where the entire, you know, the entire food chain of writing and recording a song is in some cases being done automatically and AI driven. So it's I mean, those are some sort of extreme examples of
⁓ where AI is starting to kind of infiltrate the music business a little bit. And sometimes you can't even necessarily trust exactly what you're hearing. I may be a little old fashioned in that sense because I had a musical background and as a performer and a writer, I...
I know I'm going to sound like I'm walking uphill both ways to school in the snow, I remember it back when it was tough. You had to learn an instrument, had to learn how to sing, you had to write songs, you had to record it. So it's not to say that there isn't clearly a benefit of having technology. ⁓ I'm still pretty involved in that industry and I know folks that are still recording artists and they're able to still be creative.
but they're able to put something together in a matter of days versus weeks or months. And so that's.
phenomenal. mean, as an enabler, you know, technology can really help, but there is a risk that it can go a little bit too far. And there's a risk that there could be a trust sort of issue. You mentioned auto tune. I mean, that's a perfect example. I actually have a good friend. I used to teach guitar lessons ⁓ back in the old days as well. And I had a guitar student who's now ⁓ he's a big Nashville ⁓ artist. He does session guitar playing and he's toured around the world as well.
and has his own writing. And I remember hearing a story where they would go into a Nashville studio and they wouldn't even...
even record people without auto-tune being on in the listening environment because they didn't actually care how the track sounded when it went on tape. They cared how the track sounded going on tape through auto-tune. So I remember hearing about that, that it was a little disheartening that it was so pervasive that they, you know, it was just a given that it was going to be part of the, you know, that it was going to be part of the signal chain. But so we can go a little bit too far that it can be relied on as a crutch.
But I think there is still an aspect to AI and music where there are some commonalities. Yeah, I mean, if you think about music, both of them are deeply rooted in math, harmony, rhythm, timing. There's lots of overlaps between music and AI on a good day anyway. But you have all these areas where, you know,
you can benefit from the technology. There are overlaps for sure. mean, even from a creative point of view, you're dealing with kind of listening and adapting and creating things in real time. I mean, they're not totally dissimilar sort of areas and fields, but if they're relied on as a crutch versus as a tool to help drive creativity, I think maybe that's where some of the more spirited debate comes from.
Bernie (20:58)
know, interesting enough, it's a tool and I get that piece. After watching tradesmen doing a house reno here, I'm going, I can do that. And I already know the answer is yeah, I can do it. But it will be a dog's breakfast when it's done, laying tiles on the floor. How does a business know when they're looking at an AI tool? Yeah, I can do that. Is there like a sandbox environment you kind of test it out in and then you unleash it to the wild?
Is there, when you know it's too much, like, whether it's a interactive, like a chat bot on a website or it's part of the phone tree system going, hello, Mr. Smith, and yes, Mr. Anderson, you know, we have him at dialogue, right? No matrix jokes here at all. But where do you find that you kind of like use enough for the tool to benefit, but at the same time, stay authentic to the task, stay authentic with the client?
Colin (21:52)
Well, I mean, think that goes to what I was mentioning before, you know, that there, the human connection is important in any company, in any business. So, I mean, it's important to retain that. So I think what that comes down to is, you know, any business should be looking at...
What are the areas where it makes sense to do something, where it makes sense to invest in? I mean, the best starting point is to look at things that are going to save time. mean, are there back office tasks that can be automated, invoicing, scheduling, things like that? There's no real value from a business point of view of doing that work manually versus having it more automated. So those are great places of applying AI as tools to be able to help you do that.
because it just acts as a kind of like a force multiplier. But I think, you you mentioned chatbots. I think that's another area where customer facing automation that leverages AI. I really what it's going to come down to is just going to help ⁓ a company be more responsive. mean, if it's trained properly and has the right guardrails in there, it can respond to something instantly instead of maybe having to wait minutes or hours for somebody to get back manually. So even a small chatbot that can answer FAQ
as well is going to make a smaller business look bigger and more professional. so I mean, those are, I think, good areas where it makes sense to kind of look at those things as investments because you're either going to save time or you're going to help ⁓ be more responsive, maybe look a little bit bigger than you are by having a better, more polished sort of customer presence. I mean, I think those are the things that can make a difference.
Bernie (23:29)
You said something just a second ago, guardrails. And, you know, is that something that is the business? Is that something in the company that's setting up the components? what is a guardrail? What is it meant to do? And if it fails, how terrible can it be?
Colin (23:44)
Well, I mean, you can ask Microsoft and some of the companies that have not done a historically a terribly good job putting guardrails up and had AI bot experiments publicly fail just in a colossal fashion. They're important for sure.
I think I made the comment before about trust when we were talking about data sovereignty. I think trust is maybe one of the aspects where that comes in. Like I was saying before, you've got large language models that aren't deterministic. They can hallucinate. They can say stuff that isn't true. So it's important to ground it in real world knowledge like we were talking about before. But
But there's also other areas that can come in as well. I mean, you can have bias in terms of how models are trained. you know, if the data used to train a model has bias, the outputs are going to have a bias as well. So that can affect trust. know, I mean, there's been stories about law firms using, you know, AI to do, you know, sort of legal cases, you know, hiring, you know, banks doing it in terms of lending, you know, how customers are treated, you know, so that can affect fairness.
we talk about guardrails, I think that's part of where that comes in. So it's not just about tying in actual real knowledge and real sources, but it's about trying to put kind of parameters around it so that the kinds of things that come out in terms of outputs are accurate, are appropriate and are fair. And those are typically done, mean, especially with generative AI. mean, those are typically done
by any kind of larger generative AI system. They're done under the hood with the chat GPTs of the world and the systems that we all use on a day-to-day basis as well. Those guardrails exist in there. They're part of the system prompts that are intended to, again, to keep the outputs accurate and fair. And especially when we're talking about a business that might be leveraging things, you wanna make sure that it's appropriate. You wanna make sure that those guardrails include, what is the voice of the company?
I mean, how do I, know, what's the voice that I want to project to a customer when I'm speaking to them and to keep any output or any generative results kind of lined up with, you know, with that barometer.
Wayne (25:58)
when you read business press and you hear about Sam Altman and Microsoft and Google and Metta and what they're talking about, the numbers they're using are all phone numbers. mean, they're just, they're not where SMEs are, small, medium. Where is this going to the electrician, the plumber, the window and door salesman?
When is that going to mutate down to the guy with the cell phone and the laptop in the truck?
Colin (26:28)
Well, I mean...
I think to some degree, maybe that goes into what Bernie was mentioning before about having AI that's resident in our phones. I think some companies, I think Apple is a good example of it. think some companies like OpenAI are using generative AI very publicly as a platform. Come here and ask ChatGPT a question and I will give you an answer. But I think other companies, Apple, Google, mean anybody that has devices and
operating systems, I think even Microsoft is an example of that. They're trying to figure out how to kind of integrate it in to help solve problems. So it's less about having a specific answer. There's been lots of fanfare around how some of the AI will get more intelligent and help people. For example, things like...
being able to ask my phone, hey, I think my mother's flying in. Can you tell me if the plane is gonna be on time and what time I have to leave here to be able to get there on time to pick her up? So I mean, that's a good example where it goes.
certainly beyond a straightforward question. It's now asking questions that matter to kind of normal people. It's integrating personal information and personal context and it's helping connect a lot of dots on their device to do something a little bit more normal, you a little bit more human. Hey, I just, I need to get in the car and go and drive and pick up my mother at the airport.
help me do that and help me not be late. So I think there's lots of examples like that where some companies are approaching it as a platform that stands alone by itself. But I think a lot of other companies are working really hard to try and get it, I think, out to more of the layman. I just need a thing on how do I use this stuff on a day-to-day basis to make my life a little bit easier. And hopefully that's pervasive across the...
board to anybody, know, it's not exclusive to the technocrat at you.
Bernie (28:25)
So if you're out hunting information and you ask a particular AI platform for an answer and you're looking at it going, how do I know the veracity? unless of course you're asking what's two plus two, right? And you're getting four and you're looking down and you're getting 4.9, then something's going on. How do you, but you're asking a complex question, say it's a medical question and not that you're gonna, you know,
based on the response back. But if you wanted to cross reference it, we do ask another AI platform the same question and see if the answers pair up. In other words, there's two separate LLMs hard at work and you're going, well, look, they match.
Colin (29:09)
Well, certainly a lot of people do that. They'll have LLMs fighting each other and trying to come up to the right answer in tandem. And if they end up at the same place knocking on the same door, then that's probably a good answer. I think certainly from a business point of view, I think the way more people tend to approach that problem is more with that, like I said, that combination of guardrails and
kind of augmented knowledge retrieval. I mean, so there's technical ways that people tend to do that. They'll use vector stores, vector databases, and take all of their company stuff and sort of throw it into a basically a little database, but one that's kind of geared around semantic searches and not more literal searches. And they can tie that into a pipeline that generative AI can feed off of. And so they use that to kind of balance the real world stuff.
Typically today, if you're talking, I if you're just using chat GPT, you might have to fight it out with another chat query. But the companies that are doing this that are incorporating, again, whether it's health information or other things, I mean, in most cases, they're taking that query and then cross-referencing it typically with context about the individual that they're referring to. So they have real world context that actually has information about them specifically. And then they typically have
knowledge that's coming out of effectively like a knowledge store. can think of it like a big knowledge database where they're using that to kind of cross reference and keep sort of the guardrails on that conversation to keep it kind of within factual bounds to make sure that it's accurate. So that's sort of a common way that people are guaranteeing it today. But if you just ask Chachi P.T.
Although the guardrails are starting to get better and the accuracy is getting better, that they clearly make a point of highlighting lower accuracy or higher accuracy, lower hallucinations every time they do an update. to be honest, it's always still a bit of a coin flip at a Wild West in terms of whether you can 100 % guarantee. You don't want to always bet the farm on the answer.
Bernie (31:15)
if the person is capitalizing on the AI, will they feel better off or will they feel like, is it a crutch or is it an aid?
Colin (31:26)
I think...
I think in a lot of cases, as long as it's used properly and used kind of consistent with what I said before, mean, try and try and automate the sort of the grunt work, the low hanging fruit, the repetitive tasks. You know, if I was somebody that was sitting at a contact center, if I was an agent that was on the end of a phone answering it, it would get a little repetitive to keep saying, our business hours are nine to five, our business hours are nine to five, you know, that that can get a little bit repetitive. So I think
I think being able to leverage it to do those kinds of things, I think being able to leverage it to do things like help them do sort of quick lookups of information when somebody calls, because there's a business benefit to that too, if you can kind of shorten the time that somebody is on a call and that's better for the customer and it's better for the agent, because people cost money. So there's benefits for the business, but I think the benefit for
the sort of people that are working in the companies is that they don't get dragged into all these kind of low value repetitive tasks, you know, and they can be a little bit more focused on the higher value stuff, a little bit more focused on things that have an actual, you know, sort of customer connection. know, mean, they're part of the company helping to sort of build its connection with people, you know, as opposed to, you know, feeling like they're just, you know, repetitively answering the same question 50 times.
in a row. I think in a lot of cases they will find that it's a value but it does depend on a company trying to implement it the right way. know, implement the stuff not only that a customer may want to get quickly but implement the stuff that agents or people that are doing the work don't necessarily want to have to do themselves because it's kind of low value stuff.
Wayne (33:04)
There are people selling shiny objects. And there are people who think they can implement the shiny object. And they find out after spending a fair amount of money and time, they didn't even know the right questions to ask. What does the human consultant, salesperson value added bring back to get them to a yes?
Colin (33:28)
Yeah, I I guess there's sort of two sides of it. I you know, I think one side is around, you know, that there's sometimes a lack of knowledge about even knowing what questions to ask about AI, you know.
because it can be a little bit of a black box, you know, can be a little bit opaque, you know, it's hard to sort of know what's in there. And as you said, you know, there people selling shiny objects, and, you know, it seems like magic sometimes. So they come in and they seem like magicians. But I think for the people that are buying technology, they can
you know, they can ask more practical questions. They don't have to understand the math. You know, they just need to kind of look at the outcomes, you know, look, where's the training data coming from this? I mean, what are you doing to prevent kind of bias or hallucination and keep the results accurate? You know, can you show me some real world results with other customers, you know, that are like us, you know, for how your thing performs, you know? So I think there's better questions that they can ask around how that works. But...
I think there's also, I think for the people that are selling the technology, think you mentioned an important kind of aspect, which is a lot of times it's positioned as a silver bullet. And that's, I think a little problematic sometimes because it's not, you know, it's, I think the best way that...
both people selling technology and people that are trying to implement technology, I think the best way for them to look at it is, you know, experiment, know, start small experiment and measure the results, you know, treat AI like it's a business experiment, not like it's a silver bullet and get comfortable with a loop and a cadence, know, it's AI is far more about.
testing, measuring it and adjusting. It's far less like, you know, 20 years ago, if you had a big system, you'd spend a year or two in some big long implementation, it will be done. You'd close it and you'd run out that thing for 10, 15 years and that was the end of it. AI systems are typically not like that in terms of how they're implemented. I mean, there's an ongoing...
cadence and tuning that goes into it, at least certainly in the successful implementations. And so I think when it's sold as a silver bullet or when it's even viewed as a silver bullet by the people that are buying it, I think there's a much more productive way of approaching it than that. Think of it like a journey, not like a destination.
Bernie (35:48)
As long as we don't have a werewolf, we're good because then the silver bullet will have to come out. you have a preferred AI platform?
Colin (35:56)
All of them. ⁓
Bernie (35:57)
Or is it
as or is it a task specific?
Colin (36:01)
I think it's yeah, I think the I mean, the not joking answer is it's always task specific, you know, there are.
You know, there are some like the open AIs of the world that are really good for cloud related. know, I mean, they're going to be sort of on the forefront of what's going on for sort of paid models. You know, they have really good kind of API endpoints that can allow you to automate things, you know, not just use it like ChatGPT. You know, they have aspects that can allow you to do some agentic kind of workflows. have, you know, the...
Wayne (36:35)
Jen, how's
that word?
Bernie (36:38)
It's a color,
Wayne, it's a color. Sorry, that's bad.
Colin (36:41)
D'accord.
So yeah, mean, you so you have the ability to kind of leverage it to do workflows where it's able to work more autonomously, you know, so it can do, it can work in a loop, you know, it can look at a problem, it can break the problem down into chunks, and then it can kind of go through that process, you know, iteratively.
autonomously and chain tasks together until it can solve the problem. companies like OpenAI and others have, they have APIs that can do those kinds of things as well. They also have models that can do things like text to speech. You want to generate some really good quality ⁓ human speech for a transcription for anything else. I it'll do those kinds of things as well. So those kinds of companies I think are great in the open cloud space.
they're not necessarily always the companies that you would go to if you're doing something premise based or you want to leverage something that's open source, if it needs to be an offline kind of air gapped model, common open source ones, Facebook is actually fairly prominent. They've funded the open source LAMA LLM models. Mistral is another one that's very common. depending on what area that you're looking at, ⁓ if it's cloud, if it's premise,
you might have different tools for the job and different vendors have different strengths. And I think one of the really cool things is, as we're sometimes seeing in the news, there's always a news report every month or two about somebody like DeepSeek and some of these companies that are coming out with crazy innovative models that turn the entire industry on its head. And I think that's one of the really...
interesting areas of this area is it's kind of democratized. One of the benefits of people leveraging open source tools and algorithms and technology is that they're all building on each other. And then every once in while you get these really incredible kind of leapfrogs that the whole industry goes, I think we're going to turn right instead of left. This is really cool.
Bernie (38:34)
So if somebody's coming along and they're going, oh my gosh, it's like shopping for a car. As we all know that sometimes, you know, if you have a knowledge base and you understand how the mechanics of the motor, the car works versus, oh, that looks shiny and pretty. Who do you seek for sound advice? In other words, I'm a startup business. Now, maybe my funding is a little tight.
I need to automate some process. I get that. I'm not afraid. I'm trying to make sure that I'm keeping my data sovereign to my side of the fence and protecting my client base. But at the same time, do you just come off the shelf and you plug and play or who should be the person they should be reaching out to to kind of coordinate? Yes. This is Mr. Smith and you're Mr. Anderson. Let me guide you down the rabbit hole.
As for way too matrixy, I apologize. But that said, what would you identify as a resource to guide you into a path? You might have some tech savviness to yourself, but you know, want to make sure you get it right at least the first somewhat close to the first time.
Colin (39:42)
Yeah, I think there's always a benefit of leveraging people that have experience in the field just to help guide you as to what's real. again, we're back to the hype. that word has come up a couple of times. sometimes things are overhyped, sometimes they're underhyped. so it's always, I mean, it's never a bad thing to have somebody with some kind of knowledge and experience to help at least kind of guide you through some of that process. But I do think it's still
I think I would still go back to what I was saying, that the tools are democratized enough that I think there are always some really easy areas in low-hanging fruit where...
where people can start small and experiment a little bit and then just sort of see, hey, I think this might be an area, sure, maybe I'm gonna invest a little bit and maybe have somebody give me a hand on that. But we all have phones, we can all go to chat GPT and type something. mean, there is the technology I think is more pervasive than previous generations of technology. And I think that's a wonderful thing, because it means it's more accessible to people. I think before anybody invests
in an expensive consultant, I think it's always worth playing around a little bit. Start small experimenting, try and find out some area that might be of value, especially as a small, medium business owner that maybe doesn't have quite the annual budget for consulting ⁓ as a large enterprise. And just get comfortable with it and then at least you're gonna be able to ask better questions.
Wayne (41:10)
If this made sense, if you're a listener and you wanted to get a better take, you want to understand a bit better without being talked down to, how do we get a hold of you?
Colin (41:22)
But I think the link that you've got, going through LinkedIn is probably an easy place as any to reach out. I I know that's how a lot of us stay connected in the corporate world. So that's probably the simplest way.
Bernie (41:37)
Colin Mackenize, I want to say thank you so much for being a guest today. You, Wayne Pratt, for being the co-host, and especially you, the Knack 4 Business listeners, because you know what? This makes it worthwhile sharing the knowledge. So Colin has been doing touring as a musician to pioneering hospital networks. He seeks harmony between creativity and innovation, and that's kind of the key.
and he's working to ensure that AI driven solutions remain meaningful and impactful, ethical and deeply connected to your needs as a human being. You got a question for him, reach out to him. He's extremely solid person to talk to and reliable.