Episode Summary
Netfor's founder and CEO Jeff Medley joins his son Ian for the third podcast episode to talk about why AI readiness, not AI itself, is the real strategy, tracing Netfor's 31-year evolution from PC hardware into field services, fulfillment, and customer support before explaining how the company backed into AI development out of fear of being disrupted, only to realize its real advantage was three decades of disciplined knowledge lifecycle management. Jeff walks through why hallucinations happen, why AI has to be governed like any other part of the knowledge lifecycle rather than treated as a magic fix, and why companies should start with simple, low-risk workflows (answering basic questions, gathering intake information) before moving into complex ones that require system integrations, like onboarding a new employee. He closes with a clear piece of advice for any company chasing AI: fix your process and prove your workflow works with humans first, because bad processes fed into AI just become bad AI faster.
Episode Summary
Ian Medley: Welcome to the third episode of the Netfor podcast, I think we're calling them episodes. We have Jeff Medley, who is our CEO, founder, and personally my father, so we can dive right into that little tidbit.
Jeff Medley: Hello, Ian.
Ian Medley: Thoughts on nepotism in the workplace, that's your first prompt.
Jeff Medley: Well, it's necessary sometimes. Very happy to have you in the company, and I learn a lot from you every day, as I'm sure you learn a lot from me every day.
Ian Medley: I sure do.
Jeff Medley: It's been a great adventure so far.
Ian Medley: Yep, sweet. So tell me about your role in the company, where Netfor started, how you've seen technology grow over the past 31 years, I think. Tell me about where you've come from and the journey so far.
Jeff Medley: Yeah, 31 years, started back in 1995, and I've always been a problem solver, that's kind of, as a mechanical engineer by trade, I've always tried to fix things. I used to take apart toys when I was young and figure out how they worked rather than play with them, and that turned into motorcycles and some engines. So I looked at that as probably the passion I had in starting Netfor. About 1996, '97 was the real thrust of personal computers hitting the market from Gateway and Dell, so we got on that bandwagon early on, building systems that quickly disappeared as those companies ate up all the margin. And we got into services, which is really where we love to solve problems. On the service side, we started doing on-site visits.
Jeff Medley: That started our on-site field services. We quickly went into fulfillment, doing logistics for our clients and lifecycle management of hardware, and then eventually service desk and customer service, doing the business process of IT help desk and customer service call centers. And that brought us to today.
Ian Medley: All right, so those five or six services you just mentioned have always been the foundation of what Netfor does. Tell me about the new adventure you're getting into with AI, and where Netfor is heading with that new service offering. I know that's going to be quite a large umbrella of what we do in the coming years, so tell me about where you envision that going.
Jeff Medley: Yeah, that's a great question. To be honest, the industry of customer service and IT help desk has been very mature and has been a lot the same for decades, if not multiple decades. There have been slight improvements over time, but ultimately, someone has a problem, they call a phone number or send an email, someone answers, references either their tacit memory or a piece of paper or a computer-generated knowledge article, troubleshoots, resolves the issue, and hangs up. It's been that way for many years. Scaling that with regard to knowledge has been quite expensive, because all the knowledge we use has to be created, managed, archived, and improved, and that entire lifecycle has to be managed by humans. We've never been able to cure that, it's always been an expensive part of what we do.
Jeff Medley: At Netfor, for example, we have a whole team of knowledge managers managing that pool of information for our clients. What AI has done is allow us to scale that and manage that knowledge lifecycle more quickly, more efficiently, and deliver it to those who need it in a more productive way.
Ian Medley: Yeah, and that segues right into the theme of today's episode, being AI ready. How did you know Netfor was ready to get into the artificial intelligence space? Was there a checklist in your head, or were we already at a point where our processes were so good that we could just layer it on top to enable our knowledge managers and service delivery agents to perform their jobs?
Jeff Medley: I wish I'd had that crystal ball two, two and a half years ago, but I didn't. Honestly, we got into AI development out of fear more than anything, early on.
Ian Medley: Get left behind.
Jeff Medley: Yeah, because early on in AI, as the talking heads were discussing the impact it was going to have on the economy, customer service call centers and IT help desks were the first on the chopping block. People could see, hey, this tool is really good, you ask it a question and it gives you an answer really quickly, so that must be very similar to a call center, and therefore call centers are going to be at risk of being replaced by AI. So that was the initial entrance and consideration in our strategic plan, three or four years ago. As we looked at AI and what to do about it, it really started out of fear.
Ian Medley: Yeah, and do you think that's a big reason why so many companies trying to implement AI today are having trouble, because of fear for their business?
Jeff Medley: We have Netfor delivering our services for us now, look at that over there, we can implement an AI LLM and all of a sudden we don't...
Ian Medley: [inaudible]
Jeff Medley: ...need the services of the humans over here. So that initial fear was of the unknown, and how easily, if at all, it would be to implement and replace customer service call centers like ours.
Ian Medley: Okay, cool. What do you hear from companies that have tried AI that don't feel like it landed? Are you hearing anything from other companies, people in your network saying they tried this and it didn't work?
Jeff Medley: Yeah, operationalizing AI and turning it into return on investment, or a reduction in cost, or any part of a strategy, we've learned over the last two or three years that there's more involved than just turning on a switch. When you look at, you know, there are well-documented large e-commerce sites that replaced their human-driven customer service call centers with AI and had to retract that and bring the humans back in. So we know that now, and what we're focused on is operationalizing what we've been doing for 30 years, which isn't creating AIs or software, or even necessarily the process of answering phone calls, it's managing that knowledge.
Jeff Medley: Our belief and our strategy going forward is to work at building, operationalizing, and automating more of that knowledge lifecycle and knowledge management lifecycle, to essentially bring more insight to the business than what they can get out of today's customer service call centers and IT help desks.
Ian Medley: That makes a lot of sense, and that goes with the same wave as what David was saying last time me and him spoke, that knowledge is probably the most important thing to sort out before implementing AI. David's our CTO, he's our expert in this space, and having him on, I always learn some crazy new analogy or fact about what it's capable of. So it makes sense you're saying the same thing, knowledge being probably the most important process. When you slap AI on top of a broken knowledge base or broken knowledge management and expect it to work, that never works.
Jeff Medley: It doesn't. Back to two years ago, early news reports started reporting what were called hallucinations, still called hallucinations. This is where the LLM will derive an answer or response that isn't based on fact, either cross-correlating information it's finding in different places, or just wanting to please you. The LLM, in simple terms, doesn't have the answer, so it looks for some way to please you with a response to your question. Think, ChatGPT is commonly known as a tool that likes to please you with answers.
Ian Medley: It's always giving me a "good job, that's a really good question, that's really smart of you to ask." It's a little too friendly sometimes.
Jeff Medley: Yeah, the LLMs, they each have their own personality, Claude has its own personality.
Ian Medley: I think Claude's way more straightforward.
Jeff Medley: Yeah, ChatGPT, well, Claude's more thoughtful about its response, more technical.
Ian Medley: Yeah.
Jeff Medley: And it drills in further, whereas ChatGPT is more strategic in its responses and able to correlate more things to provide answers or recommendations.
Ian Medley: Do you use any other ones besides those two?
Jeff Medley: Gemini.
Ian Medley: Okay, I use all three of those for different reasons. I think GPT is a better writer if I'm trying to send an email, I think Claude's better for, every day, "this is what I'm doing, help me think through a strategy," and I think Gemini is best for deep research.
Jeff Medley: Yeah, I'm one of those old dogs, I hooked onto OpenAI's ChatGPT first, so I'm really deep in that one.
Ian Medley: I'm really deep in mine too, and I have trouble letting it go.
Jeff Medley: Yeah, we joke about that, "little Ian" or "little Jeff." There's this term, "second brain," used in some articles today, where that may be where things are going five to ten years from now, where each of us has our own human personality, our own human brain, but sitting next to us or in our computer is our second brain, essentially the AI replication or representation of us as individuals. I believe Zoom is working on this, an easy way to look at it is, think about joining a Zoom meeting, and there's your face and everyone's face as you're all talking. You were busy that day.
Jeff Medley: You couldn't make it to that meeting, so you sent your AI second brain. Your AI second brain pops into the screen, it knows most of the way you think, how you would answer questions, and it attends the meeting, taking notes, even responding.
Ian Medley: Why wouldn't it do that for all meetings then?
Jeff Medley: Because, human in the loop, good question. You don't want a hallucination. If your AI is attending the meeting and approving two-week vacations for the entire company, you don't want that. So having guardrails is important, and there's decision-making. A lot of meetings require decision-making, so putting a guardrail there, so that if the AI Jeff was asked a question knowing that AI Jeff isn't allowed to go into that area, it responds with, "that's a great question, let me get with human Jeff."
Ian Medley: Yeah, easy guardrails.
Jeff Medley: Get this information, and I'll deliver it back to you guys later today when I speak to him. So think of it as a second brain, or the AI version of yourself. I don't fortune tell, but that could be where we're headed, where each of us individually has both our personal lives and our work lives contained in that second brain, and others, our friends, family, coworkers, can interface with it. And if it goes beyond the knowledge of our second brain, it can escalate to us or bring us into the conversation to answer it. That way I'm able to focus on other meaningful things, as opposed to attending a meeting where I'm answering the same question over and over again.
Ian Medley: You've seen the show Severance?
Jeff Medley: Yes.
Ian Medley: Yeah, you're pretty much describing what that is in its own way. I don't want Severance personally, it's a great show to watch, it's probably not something we need to adopt into society.
Jeff Medley: I love the premise of Severance, I really do, I've gotten into that show. The difference is the problem Severance solves is not wanting to remember or have the experience of work, because it's so debilitating, so mind-numbing that no one wants to do it. So those who sign up for the severance procedure, they go up the elevator and don't remember anything after they get to work, and don't remember anything after they leave. Which creates, if you've seen the show, lots of problems, but it's not quite the second brain. It's definitely segregating into the personal life I want to remember and have fun in, and the work life I might find, you know, evil or difficult. That's not me, by the way.
Jeff Medley: I get up every Monday morning loving what I do.
Ian Medley: Okay, well, good, that is a great show.
Jeff Medley: It is.
Ian Medley: So back to a little bit of knowledge, here's something I had a bullet on, I want your take on the pressure to deploy fast versus the cost of deploying wrong. It's not a question, it's a statement, what do you think about that?
Jeff Medley: I think as business owners, this is a very dicey time, 2026, it's stressful. There's a lot of information business leaders are having to absorb at a very quick pace, and it can change from week to week or day to day. The money being spent on AI infrastructure, as a signal, tells us as business leaders that this is here to stay, trillions and trillions of dollars. So we have to look at these different signals out there. I mentioned earlier, some companies are hiring their humans back.
Ian Medley: Yeah.
Jeff Medley: And so there's just a lot of signaling out there, and we all, as business owners and leaders, have precious resources, money and people and our capabilities, and we have to deploy those responsibly. So it's really making the decision of, I call it, how are you going to tiptoe into this?
Ian Medley: Yeah.
Jeff Medley: Without scaring clients away, or spending money where it's not necessary, or doing an implementation that fails and then having to scrap it and go back to the way it was.
Ian Medley: Yeah, so eliminating friction on every single level that tool could touch.
Jeff Medley: Yes, that's where we're seeing success, when we look at organizations like ours or some of our clients, and look for areas that are redundant, workflows that can be automated, and low-risk areas that can be automated and operationalized by AI, keeping a human in the loop, then deployed and managed. That's where we're seeing value, in these easier workflows. The AI we're deploying today is what I'd consider the simplest implementations, those that answer the phone with a nice voice, that gather intent, why are you calling, that can gather some information, can I have your name, can I have your birth date, and then moving into slightly more complex workflows like scheduling an appointment, then following up on that appointment or rescheduling it.
Jeff Medley: These are some areas where we're seeing a lot of success enabling companies to take that redundant work off their frontline workers. And again, we've been in this business for 30 years, we've been answering the phones for 30 years, taking very complex calls, but for the AI, think of it just like a new employee, they've got to start small. Let's get you started answering the phone and answering a few easy questions, and when more difficult or moderately difficult questions come, we'll escalate that to a human. That's what we see as those initial workflows get implemented.
Jeff Medley: And then we target more workflows in sequence with the client, and they can judge and measure them as they get implemented, and decide if they want to keep tiptoeing a little further into the ocean.
Ian Medley: Good, good, that's a really great explanation of what we're doing, thank you for that. The process on paper is different than the process in application, is there a way to differentiate the two before it becomes an issue, or do you just have to test in the field, document, and then come back to write those governing documents?
Jeff Medley: Yeah, because of our history we're very knowledge-centric, and as we mentioned earlier, that knowledge centricity keeps us centered on having information that's very binary. Our clients have never accepted, and no one would ever expect them to accept, our agents or customer service representatives giving bad information. When someone calls in with a question, the expectation is that the answer is accurate and delivered in a kind, empathetic way, and provides a great customer experience. It's strange to me that with AI, that same expectation is assumed, governance is assumed, context is assumed, and it's just not so. So we're focused not on the answering of questions, if you will.
Jeff Medley: We're more focused on how to govern that knowledge lifecycle, and how to make it work for organizations like insurance companies, healthcare providers, government agencies, organizations that just cannot tolerate hallucinations, bad answers, going rogue, going off the rails, no guardrails. We're focused on developing systems that ensure those organizations will have answers for their customers that are accurate and don't go off the rails. But we're also building in the expectation that it will happen every once in a while, and when it does, we capture it and improve on it, making that part of the lifecycle rather than something folks didn't expect to happen. We do expect it to happen, and when it does, we have a way to govern that.
Jeff Medley: That's the way we see knowledge being managed in the future.
Ian Medley: Yeah, every SLA can't be 100% all of the time, and I think you wrote something about that a few years ago, titled that exact phrase. It was a blog about how we know nothing's 100%, we know something's going to mess up, this was before AI, and we acknowledge that, but we still do the best we can and give realistic, transparent answers.
Jeff Medley: Law of diminishing returns. Our clients feel like we're doing good when we're resolving 70% of the calls that come in without any escalation or further conversation.
Ian Medley: Are you saying 70?
Jeff Medley: 70, 75, 80, the higher the happier the customer is. Anytime a customer calls in and can get an answer without being sent somewhere else, they're going to be happy. So we're always trying to improve that. How do you improve that first-call resolution rate, which AI has to account for and humans have to account for? Well, you do that by having more answers, more accurate answers, answers to more complex problems.
Ian Medley: More answers is knowledge.
Jeff Medley: More answers builds knowledge, and having context to it, being able to troubleshoot it and find the right answer, is also part of the knowledge equation. But you always have to deal with the law of diminishing returns. If you spend $100 to increase your first-call resolution from 70 to 71%, it's going to cost you $200 to go from 71 to 73%, it's going to cost you a thousand dollars to get to 80, and it's going to cost you $100,000 to go from 80 to 90. To go from 90 to 100% is going to cost you millions. That is the law of diminishing returns. You can't expect perfection out of humans, and you certainly can't expect it out of AI.
Ian Medley: Okay, that's the law of diminishing returns, I haven't heard that one, and if I have, I don't remember it.
Jeff Medley: I'm a poor father.
Ian Medley: No, no.
Jeff Medley: I should have taught you this when...
Ian Medley: You were three, and it went in one ear and out the other. All right, so we covered that.
Jeff Medley: So what?
Ian Medley: Our first-call resolution rate for some clients is upwards of 92 to 95%.
Jeff Medley: Yes, on the customer service side.
Ian Medley: On the customer service side.
Jeff Medley: Customer service is where you're not troubleshooting, you're answering. Think about a government agency, you're answering a phone call, they're asking about a tax form or a legal document, and there's a rote binary answer to that. IT help desk, there's troubleshooting involved, and when you're dealing with IT systems connected as part of a network, there are a lot of different causes, and that causation has to be determined through troubleshooting, which increases the call time and increases the number of times you have to escalate.
Ian Medley: Do you think AI is going to increase that even higher, the percentages on...
Jeff Medley: On the customer service side? No doubt, it's already doing that on the IT help desk side, it will improve it. But how far it can go depends on, I think, where we're focused, which is having context, context to the knowledge.
Ian Medley: Yes.
Jeff Medley: Is the LLM able to see context in the intent the caller is calling in about, and then use that context to actually find the proper knowledge to deliver? And if it can't deliver it, can it stop halfway through, realize it's not working, and pivot over to another knowledge article, which is what a human would do, "hey, this isn't working, I need to go over here now." That's a question that hasn't been answered yet.
Ian Medley: And this LLM that's answering these questions, either voice or email, it's searching for it, it can't go outside of its knowledge base, right? It's in its own area, it can't go search the internet for why is this computer doing this and making this sound, it can't go search, depending on how you configure it.
Jeff Medley: I think you asked earlier about how we're finding success in AI, or in the different LLMs out there. If you allow the LLM to go anywhere it wants, some of them were trained on Reddit, some were trained on the internet, some continue to be trained on that information.
Ian Medley: David talked about this.
Jeff Medley: Yes, it's very broad. When I talk about healthcare organizations and government agencies, I'm talking about a locked-down system of knowledge whereby it cannot go outside the realm of what's in front of it, a very concise box, and then that knowledge is used to drive the answers, and all the things we've talked about today with those guardrails. But at the end of the day, the LLM is able to see patterns and find the right knowledge articles quicker than humans can, and deliver them in a better fashion.
Ian Medley: It can pick up on things humans would never be able to pick up on, like if there's something coming in from one store having an issue every Wednesday at 2 o'clock, maybe it calls in around 1:59 on a Wednesday, and it's probably assuming that same issue over and over again. Do you think it's doing that?
Jeff Medley: This is where predictability, this is where you're hitting on again, back to our focus, what we're focused on today. For decades it's been extremely difficult for humans to recognize these types of patterns. Whenever a human answers a phone call and resolves it in a certain way that may not be exactly part of the documented knowledge, collecting and understanding the nuance of that additional information they used to resolve it is next to impossible for humans to do. You'd have to literally go through, we take tens of thousands of phone calls a week, what human is going to go listen to those calls and figure out, oh, instead of going to step seven they went to step ten and it resolved it 90% more effectively? With AI, we can do that.
Ian Medley: Yeah.
Jeff Medley: So in this knowledge lifecycle and insight work we're talking about and working to perfect, the LLMs and AI allow us to process recorded phone calls and ticket information more quickly, to raise those insights to the department or division on the service side. How do we make that knowledge better and better on an ongoing basis, so the next phone call that comes in is always a better experience? And secondly, how do we raise those insights to the business?
Jeff Medley: How can we look at the number of phone calls coming in regarding a specific call type, recognize that it's causing productivity issues, maybe even sales issues, at our end users, and gather that data and raise it to the business, so they can make real business decisions about investing in and removing that problem from their environment? A lot of IT departments, a lot of business departments, don't know where to spend their money, and it's buried in these insights we're talking about. If they can connect the end-user experience, the folks calling in, texting in, asking questions, and process all of those interactions, and look at it from the perspective of what impact is this having on our business, those problems can then be picked off.
Jeff Medley: They can be looked at in terms of which one is costing us the most, and they can spend money on the things that will make the most impact on the business. And we say that's a core part of all of this knowledge lifecycle we're talking about.
Ian Medley: Yeah, and getting there isn't just slapping a tool on top of pre-existing processes. Everything you explained, revenue-blocking issues for clients with a thousand stores nationwide, one outage could cost them hundreds of thousands of dollars an hour of downtime if people can't process their payments. So coming back to the overarching theme of what you and I were talking about, being AI ready, AI isn't the strategy, it's the tool you use once your processes are already really good. So being AI ready is the most effective thing a company could do, what's a good word for "preemptively"?
Jeff Medley: Proactively.
Ian Medley: Proactively.
Jeff Medley: It's not that the processes are good, it's that they can be automated.
Ian Medley: But what does a company need to do to be AI ready? What questions should they be asking to get to the point of, all right, we're comfortable with the process, let's now, instead of humans handling it alone, have AI and humans handle it for a better result, those things you were talking about?
Jeff Medley: Yeah, David can drill more into the diagnostic part of operationalizing AI, but I can tell you it starts with readiness. Is your data infrastructure ready? Is it available? Is there a unified data structure within your company? Because at the end of the day, it requires access to your systems, and without that it literally cannot be successful. And beyond the data side, it's identifying workflows that can be automated. If there's a complex workflow, great example, if I'm an agent at Netfor and someone calls in asking a question, and I follow a knowledge article, and that article says step seven, go out to this system and change their password, go out to this system and add this information, then do this and do that, that's called an integration.
Jeff Medley: So instead of me just delivering information to the caller to solve their problem, I now have to go out and touch systems to resolve their issue. Think about new employee onboarding, I need to add them to the payroll system, add them to IT systems, get them a badge, all of those require integration. So the more complex workflows are those that require integration, the simple ones are those that just require passing along information, what are your store hours, they're eight to five, oh, thanks, that's a very simple workflow. Onboarding a new employee, not a simple workflow. So back to your question of AI readiness, you start with those simple workflows.
Ian Medley: Okay, and we're probably going to close out soon, so any parting messages on AI readiness? What should companies, local, national, be doing before implementing tools, just building on what you just said, anything else?
Jeff Medley: Talk to those that have been doing it for decades. From a human standpoint, bad processes fed into AI just become bad AI. One of the advantages we have with decades of experience is we're focused in the area of knowledge lifecycle management, which we've been doing for decades, and having that information and knowledge allows us to take those processes and operationalize them. So whatever you're doing out there as a business, if you're looking at slapping AI onto it, as you said before, make sure the process is sound, make sure the humans can do the process before you expect the AI to, and start on the less complex workflows before you add the more complex ones.
Ian Medley: And how about the mindset? Should people go into it open-minded, fast, or quality driven? I think quality driven is the right answer instead of fast, but I know a lot of leaders, especially IT leaders and customer experience leaders, are under a lot of pressure to get AI integrated, so they, like we started off saying, don't want to get left behind, or the easy answer from execs who maybe don't know what they're talking about is just slapping AI on a workflow without knowing it's broken, which is going to lead to disaster. I know I went in two totally different directions when I just said that.
Jeff Medley: But you're really saying the same thing, the workflows and the knowledge are tantamount to the same thing. We discover workflows through knowledge, and knowledge through workflows, so they start to conflate with one another. But having those conversations early on, I think, is the best way to look at it from a business standpoint. And even the training, one of the things we're doing at Netfor that you're well aware of, as we've been training our managers on AI, we're not looking at success as "you've now been trained on how to use an LLM," we're looking at success as "you have now delivered your first outcome or deliverable assisted with the AI." That really is what it's all about today.
Jeff Medley: Today it's about how we can use these tools to do the more redundant parts of our job, so we can elevate ourselves with our capacity and our education and knowledge to do more valuable work. Back to that second brain, I want the second brain working on the redundant, mind-numbing work, so I can focus on the creative, decision-making, collaborative work that only humans can do.
Ian Medley: Yeah, what's the first step, this can be your parting note, for anybody in training, or just in general, somebody who wants to implement an AI tool, what's the first step?
Jeff Medley: Talk to someone that's done it.
Ian Medley: Okay.
Jeff Medley: And you can look around, I think the companies that failed fast, that have failed at it, are also good companies to talk to, because they failed, but they tried. Some have given up, "I'm just going to hire the humans, I'm going to stay away from that tooling, and let the market figure this out before I dip my toe back in." But others, like us, we buy into failing fast, we try things, we A/B test, we move on to the next iteration, and it's been that way for two years. It's two steps forward, one step back, learn something more, look at the market, present it a little differently, look for improvements, and slowly step your way into a better model, if you will.
Jeff Medley: And the AI will come. It's there to assist us, it's not there to take our job over.
Ian Medley: Yep, I think that's a great phrase to leave this on. So thanks for coming out, thanks for talking to me, you always school me on something, so I'm always learning.
Jeff Medley: Thanks for having me, my first podcast.
Ian Medley: So I've now, certainly not your last...
Jeff Medley: The first one out of the way, I appreciate you having me on.
Full Episodes Available on YouTube and our Podcasts page