SPEAKER_01 0:08

Welcome to the Technology Translator. I'm Vic. Every episode we break down technology, cybersecurity, and AI into plain English for Australian business leaders. Welcome to the podcast.

SPEAKER_00 0:28

And I'm back for another episode of the Technology Translator. I'm Vic, and if you've been listening to the podcast over the last few weeks or the last few episodes, you'll know we've just spent an entire episode talking about AI for business. We spoke about what AI actually is, where it's genuinely useful, where businesses can get themselves into trouble, and why AI is actually an incredible tool, but it's just not magic. If you haven't listened to that episode yet, I'd recommend going back and giving it a listen first, because today's episode builds directly on top of it. Last episode was all about what AI can do. And of course, that keeps changing. Today's episode, though, is about what your business needs to do before AI can do it safely. And honestly, I think this is the conversation that people are only just starting to have, if at all. Because everywhere you look at the moment, AI is being marketed as the answer to every business problem. Every software company has AI. Every salesperson has an AI feature. Every conference has AI keynote speakers. Every board meeting seems to include AI somewhere on the agenda. And you know what? I even went to the doctor the other week and my GP said, hey, do you mind if I record this with my AI assistant so that I take all the notes? Wow. And if you're a business owner, the chances are you're sitting there wondering whether or not you've actually fallen behind. But here's what I'm telling you right from the beginning. I don't think most businesses have an AI adoption program. But what businesses do have is an information problem. And I'll explain what I mean right there. So a number of years ago, I was involved in discussions with a government department wanting to introduce AI into their environment. Now, this was well before Chat GPT exploded into the mainstream. Businesses were talking about AI, but it wasn't sitting on everyone's phone or computer just yet. This organization had a vision. They wanted an AI assistant that could search through all their internal information and something where particular staff members could ask very, very specific plain questions and receive answers from their own documents. It's essentially what we all imagine when we think about Chat GPT for business today. The conversation started with what seemed like a very simple question. So what would it cost? That's fair, right? But as we started unpacking what would actually be required, the conversation changed completely. Because before AI could actually answer a single question, we needed to know where all this information lived. Who owned it? Who could access it? And whether some of that information was confidential, whether some of it was public, whether it was commercially sensitive. Did it contain personal details of people? And then how was it organized? Was it classified? Was it duplicated across things? And then was it even accurate? And suddenly what everyone thought was going to be just a simple AI project of we'll give you these documents, you read them, tell us what we need to know, turned into this massive data governance project. And I'll never forget the look around the room. People genuinely had not realized the amount of work that sits underneath a successful AI project. And not because anyone had done anything wrong, it's just that most people think AI starts with buying the software, signing up to the subscription. But it doesn't. It actually starts with understanding your information. And that right there stuck with me. Because years later, ChatGPT has become mainstream, and I've started having this same conversation with businesses all over again. Except this time, instead of asking about some futuristic AI project, people are asking about Chat GPT, Copilot, Claude, Gemini, every other AI tool in the market. And the question's all the same. We want AI. And my answer is also the same. Fantastic. But let's work out whether or not your business is actually ready for it. Because here's the reality: the AI side is not the hard part. Preparing your business and your business information is. And it's created another language. Yep, seriously. Acronyms. They are IT's bane of their existence. Every second conversation now includes acronyms and terms such as LLMs, agentic AI, AI agents, anthropic, rag, MCP, reasoning models, copilots. And I'm sure that you've probably heard a few of them and thought, I've got no idea what on earth they're talking about. And don't worry, you're not alone. That's actually one of the reasons why I started this podcast. Technology has always had a hover of making simple ideas sound super complicated. So let's actually translate them. So let's jump into translation time. Anthropic. This is one that's been in the news a heap. We're talking July 2026. Anthropic is a big part of the news. And all it is, it's just another AI company. And most people know OpenAI because they created Chat GPT. Anthropic created Claude. It's like Apple and Samsung. Different companies, different products, doing their own thing. But OpenAI and Anthropic are two of the biggest organizations building AI models today, particularly for businesses. So neither one is simply right or wrong for your business, but just different tools. Here's one that's a little bit more interesting. AI agents. So most people think AI is just something you ask questions of. You type something into Chat GPT. It's like Google but on steroids. It gives you an answer, and that's where the story usually ends. But AI agents take things one step further. Instead of just simply answering the questions, it actually does the work for you. So imagine AI to follow up a customer, update your CRM, book a meeting, create a proposal, generate an invoice. Instead of giving you instructions on how to do those things, it actually just goes ahead and does it. And that's an AI agent. Now, imagine you don't just have one AI agent. You've got five. You've got one managing sales, one doing marketing, one's doing finance reports, another one handling customer inquiries, one scheduling meetings, and they're all just working together. And that is a gentic AI. Instead of having one digital assistance, you have a digital workforce. Now, are we really there yet? Not really, but we're seeing a lot of progress in this space, and it's very early days for many organizations. However, I do genuinely believe that this is where the market is heading, and one of the reasons that businesses need to start preparing now. Another acronym you'll hear, especially if you're talking to AI consultants, is RAG, RAG. It stands for Retrieval Augmented Generation. Sounds terrifying and really complex. So what does it actually mean? Let's say you ask ChatGPT about your company's leave policy. It doesn't know, it doesn't work for your business, it can't see your internal documents, or I hope it can't see your internal documents. RAG changes that. So instead of answering purely from what AI already knows, it first searches through your organization's trust information. So when most people are setting up, say, a co-pilot example in their organization, Copilot will typically use RA to search through the company's documents first. This is where you can get into some awesome things like hey, based on my company's brand resources, create me a PowerPoint slide that has the look and feel of my company, but with this information in it. It's a lot of fun. And from someone who's from a sales background, I'm sure I'll drive marketing nuts with it in the future. The last term I want to bring up is MCP, model context protocol. Again, that sounds pretty complex. It's a pretty simple idea, however. Imagine you've got AI sitting in the middle of your business. Now imagine it talks to your CRM and then your accounting software, your HR platform, your project management tool, and your document management system. Historically, every one of those systems needed a custom integration. MCP is aiming to become a common language that allows AI to connect to different business applications in a standard way. So think of it like a universal adapter, a type C USB that everyone's being told that they have to start using. So now you've got one connection that's working across many systems. Will you hear this term more over the next few years? Absolutely. Do you need to know all the detail about it? No. You simply need to understand what it's actually trying to achieve. And now there's a good chance you're sitting there thinking, look, it's interesting, but why does that actually matter? Because every single one of the technologies we've discussed has one thing in common. They all rely on your business information, your documents, customer records, procedures, contracts, financials, intellectual property. And if that information isn't organized, if it's not actually protected, if no one knows where it lives, then it doesn't matter how good the AI becomes because the results will not be what you're hoping for. So now that we all speak the same language, let's answer the question that really matters. What does it actually mean to be AI ready? Because this is where the real work actually begins. So over the years I've found there's really five questions that every organization needs to answer before they're ready to introduce AI properly. Now before we jump into these, I want to make one thing really clear. You do not need to have every single one of these solved before you start experimenting with AI. If you're using ChatGPT to help write emails or summarize meeting notes, that's very different to connecting AI directly into your organization's systems and information. So what we're talking about here is organizational AI. The type of AI systems that have access to your business information, your customer knowledge, your procedures, your IP. And that's where preparation matters. So let's look at the first question. Number one, do we actually know what information we have? And this is probably the biggest piece of work that organizations underestimate. You know how I was speaking about that government department before. This was a conversation that completely changed the project. Because AI doesn't understand your business. It doesn't know that payroll information is more sensitive than the office lunch menu. It doesn't know that executive contracts should be protected more carefully than a company newsletter. It doesn't know the difference between customer information and public marketing brochures. Someone actually has to sit there and teach it. And that's where data classification comes in. And all data classification is, is simply the process of labeling the information. It's like putting stickers on folders. This is public, this is internal, this is confidential, this is commercially sensitive. This one contains customer information, and that contains our financials. Different organizations do use different labels. Some have three, some have five, some have ten. And the important thing isn't the names, it's that your organization agrees on what those labels mean. Because once information is classified, technology can start making intelligent decisions. For example, maybe confidential information can't be emailed outside the org. Maybe highly sensitive information can't be uploaded at all to AI tools. Maybe customer information can only be accessed by certain departments. Without those labels, everything else just looks like another document. And occasionally you hear, well, can't AI just work all that stuff out? Unfortunately, no. AI is incredibly intelligent, but it's not psychic. It still needs humans to provide context. And that's why one of the biggest jobs organizations have ahead of them is this information data classification project. So question number two. Where are your procedures stored? Who owns them? I've worked with organizations where exactly the same document existed in six different locations. Copy and Outlook, someone's download folder, one in OneDrive, another in SharePoint, another in the shared drive, another attached to an old email. Which is the latest. No one knew. And now imagine AI asking that same question. What version should I use? How would it know? It can't. Because it only knows what you've made available to it. One of the biggest opportunities AI gives businesses isn't actually AI. It's forcing organizations to finally clean up years of messy information management. And honestly, that's probably overdue in many businesses anyway. Which brings us to question three. Can we trust the information we are giving AI? And this question often gets overlooked because even if your documents are perfectly organized, it doesn't automatically mean they are correct. Think about policies. Have they been updated recently? Or are staff still referencing procedures from four years ago? Think about customer information. Is it accurate? Have old records been removed? Thinking about product information, pricing, internal processes. If AI is searching through outdated information, it will confidently give people outdated answers. One phrase you'll often hear in technology is garbage in, garbage out. It's been around for decades and it applies more to AI than I think it does anything else. Good information produces good outcomes. Poor information simply produces poor outcomes just faster. Question four. Who should actually have access? This is where cybersecurity and AI start becoming the same conversation. Because AI can only work with the information it has access to, which immediately raises another question. Should it have access? And more importantly, should the person asking the question have access? Imagine someone from marketing asking AI about payroll, or someone from sales asking AI about executive employment contracts, or a junior employee asking AI for next year's acquisition strategy. The AI shouldn't answer those questions. And not because the AI is doing anything wrong, but because the person asking wasn't authorized in the first place. And this is why identity management remains important in this process as well. Things like multi-factor authentication, role-based permissions, conditional access, zero trust, they're also incredibly important because AI doesn't replace your existing security. It actually relies on it. And if your access controls are poor today, AI will simply expose that problem a lot faster. So, question five: are our people actually ready? And this is probably one of my favorite questions because technology projects almost never fail because of technology. They fail because of people. People don't understand it, people don't trust it, people don't know how to use it, or people start using it without anyone knowing. Now, I've scoped out a few projects recently where someone's actually said, look, to be honest, I need to look at a model where I am paying for the one subscription and anyone within the organization should have access to it. And my reasoning for that is because there are some people who might use it 100 times a day, others who might use it once or twice a month, and there's other people in my business who will not touch it regardless. And AI subscriptions can quite often be per person based, and it can add up very quickly when you're talking about, say, 50 bucks a user or a hundred users every single month. That gets pretty pricey. But in saying this, it does provide an important distinction. If a business does not invest in AI and figure out who is using it, why they're using it, and making it appropriate and applicable to their jobs, people will start using it without anyone else knowing. Which sort of brings us to another buzzword that you're going to hear a lot over the next few years. Shadow AI. And before I explain Shadow AI, I need to explain Shadow IT. Shadow IT has existed for years. It's when an employee starts using technology without the knowledge or approval of the business. So this might be you use Dropbox. It's not a business version of Dropbox, but you sign up to Dropbox because it's easier than the company file server. It might be using Trello because you don't really like the approved project management software. Might use your own password manager. Or it might just be the fact that you're really sick and tired of not having an Adobe subscription, so you use your company card and sign up to the full Adobe suite, just clicking the boxes and saying, yep, cool, approve everything. So usually it's not malicious. It's just you trying to do your job and trying to get your job done. Technology has always moved faster than organizations, and AI is definitely no different. And now replace Dropbox or Trello with ChatGPT, Claude, Gemini, or a different platform that's out there. And that's Shadow AI. It's employees uploading customer lists to get that better spreadsheet or putting a contract into Chat GPT to summarize the legal risks. It's salespeople asking AI to rewrite the proposal. It's someone uploading the meeting notes or financial information or HR documents. And now no one's trying to do the wrong thing. We're all just trying to work faster. Because I believe the biggest risk, um, AR risk that's out there isn't malicious employees. It's that we're actually trying to be helpful. We're trying to save time, trying to do a better job and be more productive. But without realizing, in doing those activities, we've actually exposed the business to a level of risk that we've never known before. And this is why education matters and policies definitely matter. Governance around AI really matters. And yes, technology matters too. But there are many security platforms out there today that can identify when sensitive information is being uploaded to AI websites. Some organizations use data loss prevention, often to call DLP, to stop confidential information from leaving a business. Others use cloud access security brokers or CASBs to monitor and control how cloud applications, including AI tools, are being used. Microsoft Purview can classify and protect information across Microsoft 365. Identity platforms can ensure AI only has access to the information each employee is already authorized to see. And the important thing to understand isn't the names of these products, but it's the principle behind them. The technology does exist, and the challenge is deciding on how your organization wants to use it. So if you've listened to today's episode and you're thinking, honestly, I'm not too sure. Is my business even ready? That's completely okay. I'd expect that answer. Because becoming AI ready isn't a project that you actually finish by a certain day. It's actually a journey. And every journey starts with asking the right questions. So here's the challenge I do leave you with. Sit down with your leadership team and ask these five questions. Do we actually know what information we have? Do we know where it lives? Can we trust it? Do the right people have access to it? And are our people ready to use AI responsibly? And notice something interesting. It's not one of them being what AI tool should we buy? Because that's the wrong place to start. The software will continue to change. And 12 months ago, many of us had never heard of Claude, and now it's one of the leading AI assistants of the world. A year from now, there will be new tools. There'll be new models, new acronyms, new buzzwords. The technology will always continue to evolve, as it always does. But good information management, good governance, good cybersecurity, and good business practices, they're things that don't go out of date and they simply become more valuable. So I do want to finish up where we started. Back in that meeting with the government department. At the beginning of the conversation, everyone thought they were actually budgeting for AI. By the end of the conversation, they realized they were budgeting to prepare their organization. Years later, I don't think that lesson has actually changed at all. If anything, it's become even more important because AI is moving faster than any technology we've ever seen. The temptation is definitely that rush to buy the latest and greatest to try and keep up. The worry is that everyone else is getting ahead, and if you're not doing it right now, you'll be left behind. But you need to think about it differently. Don't focus on chasing every new AI announcement. Focus on building an organization that's ready for whatever comes next. Because if your information is organized, if it's secure, if your people understand how to use AI responsibly, if you have strong governance, then it doesn't matter what AI platform becomes the market leader or the most used one in all the back technologies, you will be ready. And that's exactly where every business should want to be. AI isn't something you buy, it's something you become ready for. I really hope these episodes are starting to make technology just that bit more easier to understand. And thank you so much for listening to another episode of the Technology Translator. Until next time.au And I guess I wouldn't be doing this properly if I didn't say if you like what you hear, hit the follow button. There will be more of these episodes coming up.