You don’t need to be a “techie” to use AI for your business

AI isn’t just for large corporations, but for small businesses too, offering the power to transform operations, improve customer experiences, and boost efficiency.

With her deep passion and enthusiasm, Heather Murray is here to demystify AI for all you non-techies and empower you to harness its potential in your businesses and daily life.

Heather cuts out all the jargon to help you overcome any reluctance or fear when it comes to AI, making this tech accessible for everyone.

In this new digital age, businesses must adapt and keep up with the ever-evolving market changes to thrive, and this episode can help you do that.

Here’s what we cover:

• What is generative AI and can a small business owner use this?
• Removing the stigma around using AI and where to start
• How to work out which tasks to use AI for
• Where to avoid using AI
• Crafting better prompts to get better outcomes
• Using AI to increase your revenue
• Grow your business by becoming an early adopter of AI
• Be wary of AI bias
• Become more transparent about using AI
• How to use AI safely and securely
• Who owns the copyright for AI-generated content?
• What does the future have in store for AI?
• A fun place to start your AI journey

What is generative AI and can a small business owner use this?

Bex Burn-Callander:

Well, I was doing some research into the polarising nature of AI and that a lot of business owners feel like it doesn’t even apply to them, that it’s not something that they can use in their day-to-day lives.

Can we talk a bit about what we mean by AI? Because we’re not talking about these futuristic, sentient systems that have feelings and approximate human beings. We’re talking about a different kind of AI. Do you want to break it down for me, please?

Heather Murray:

Yes, of course. AI itself has been around for a really long time, so it’s pretty much since 1950 actually. Every time you pick up your phone or use your computer or go to the cash machine or walk past the CCTV camera, we are using AI throughout everything.

There’s lots and lots of different types of it, but the bit we’re all talking about now and the reason that everybody, every conference you go to, the reason that most people want to switch off when they hear AI, the reason that your LinkedIn feed is completely saturated, is generative AI.

That is something that’s completely new as of the end of 2022, and that is the ability to create something new.

So generative AI, generate something new, whether it’s a piece of text, an image, a video, audio, music, or voice. There’s so much you can create, and then you can analyse and summarise that information as well. It’s something we’ve never been able to do before.

Whenever I do a session I say, “Who’s new to AI?” And I always get a chunk of the room saying, “Yeah, I’ve never used it before.”

And then when I start to explain that your predictive texts, your email spam filters, everything, that’s all AI already. It’s generative AI that you’re new to.

Bex Burn-Callander:

And can generative AI, and for our listeners, that’s what we’ll probably be focusing on for most of this chat, can generative AI be used by any kind of business owner? Because obviously big corporations you can see the use case, but what about, as I said in my intro, man in van, driving around, drainage engineer, how can generative AI help him in his business?

Heather Murray:

Oh, absolutely. I actually think the smaller businesses and one-man bands, or one-woman bands, have the greatest advantage here. And I am a one-woman band, I run this business by myself.

You can integrate it into pretty much every part of your life. There are so many use cases both inside and outside of work, but even if we focus on inside work and your personal productivity.

So how you check your emails, how you manage your diary, how you have your meetings, if you have your meetings online, but if you don’t, there’s also AI for that too, if they’re outside of that. How you prepare your proposals. Pretty much every business prepares proposals in some way.

So whether it’s off the back of a phone call, off the back of a meeting that you’ve had, scribbled notes, there’s just something for everything across the board.

And I always recommend people do start with that personal productivity piece because that’s where you start to see the biggest results and you go, “Oh, I’ve got an hour back then,” or, “Oh, I’ve got 2 hours back.” Or, “That task used to take me 4 hours and now it’s taken me 30 minutes.”

That’s the magic of it.

My husband is actually just ending his 26-year career working in hospitality and starting a new business as a dog trainer. So very much not a digital business at all, but I’ve taken it on as a project. I’m working alongside him to make it as AI first as possible.

I suppose it’s not bricks and mortar because he’s out in the grass and the parks, grass and parks business shall we call it, but you can apply AI to so many parts of that. From coming up with his business name and slogan to just digging into who’s his ideal customer.

We’ll be doing leaflets, who should we be giving leaflets to? What benefits should we be pointing out to them? How is he going to handle his payments and transactions?

There’s so many things that I can help him with. I can’t think of an actual business type that wouldn’t benefit from AI right now.

Removing the stigma around using AI and where to start

Bex Burn-Callander:

That’s so interesting. Because I’m a journalist in my day job and there was lots of doom-mongering about how my profession would be dead in 10 years, but I thought what I’m going to do is I’m just going to get comfortable with AI.

So I signed up for ChatGPT account. I don’t use it to do any writing, but I do things like when I finish a piece, I’ll ask ChatGPT to write the same piece because it’s a useful way of seeing what I’ve missed.  Or if I’ve missed any glaring errors.

I would never use it for facts or sources or data at this point.

But then I also had to do a PR and marketing plan recently, and I just made ChatGPT do it because I’m so bad at writing formal proposals.

I had to edit it quite a lot afterwards, but that was a real game-changer.

Heather Murray:

There are so many different things you can use it for. And actually one thing I want to mention, just as you said that, you said that with almost a bit of shame, “Oh, I cheated with AI,” and actually it’s not.

Just like if you were to do a Google search, it’s a tool like anything else. A tool like the internet is for us, a tool like mobile phones are for us.

If you calling someone, you’re not cheating because you’re not going round to see them face-to-face. I think that’s one of the main stigmas we need to remove is that there is a shame because you think, “Well, that was too easy. Amazing”.

Picking up the phone is too easy compared to going round. Putting something in the dishwasher is too easy compared to washing it yourself. It’s just a progression of technology and I think that’s something we need to get around.

But most people start with Chat GPT. I call them dabblers. And that is honestly the vast majority of people that I work with.

I often work with rooms of 100 people and probably 95% of them will be dabblers, which means that they’ve tried a couple of things, usually ChatGPT.

They might have written an email, and they ask it to be rephrased in some way. So it’s stuff like, “Make this a little bit more friendly or more formal”, or “Give me some ideas for this.”

That’s where most people start because you open it and there are no instructions, so you have to feel your way and work out how to use it.

But there’s a whole world of AI tools out there. So Chat GPT is called an LLM, I hate technical language, hence my business name AI for Non-Techies, but it’s a Large Language Model.

Basically, it’s an AI chatbot, and there are a few of them. There’s Gemini, which is by Google. There’s Claude, which is by Anthropic. There’s Copilot by Microsoft, and there’s ChatGPT by OpenAI. They’re your big four.

They all have different purposes and different strengths and weaknesses, and they’re the best place to start because they’re multipurpose tools.

So start with those chatbots because they can do so much. Of course you can create content, but you can simulate difficult conversations, you can analyse information, you can build plans, you can get expert advice, you can summarise information.

It’s just endless, and it’s usually £20 a month, there’s always free versions when you first start, but that £20 a month is the best £20 I spend all month.

Actually, a little note when you’re saying about using ChatGPT to write or not to write in your case, ChatGPT’s strength isn’t in writing, it’s more analysis and research.

Claude is the one for writers. It’s much more human sounding.

But you start off with the LLMs and then you’d move forward into more specific tools that are designed for a specific purpose. And there’s a whole other group as well.

So you’ve got your LLMs, you’ve got your tools for a specific purpose. You have sales ones, marketing ones, finance ones, all sorts. There are probably hundreds of thousands of them out there at the moment.

How to work out which tasks to use AI for

Bex Burn-Callander:

For our listeners, they shouldn’t just be seeking out AI tools, like looking up so-and-so with AI because that might not add that much value, but they should be thinking about …

Well, maybe tell me the process. Do you look at your work day and think about the things that are manual tasks that maybe aren’t generating a lot of value for the business, but are taking a lot of time, and then look for a tool or figure out a way to use AI to cut that down?

What’s the process to follow?

Heather Murray:

Yes, that’s definitely the right way to do it. I always like to advise people to start with your business goals or your own goals, what are you hoping to achieve.

And then where are your blockers, exactly like you were saying. You don’t need to necessarily think at that stage what’s time-consuming, but AI is brilliant at time-sapping, repetitive admin-type tasks, but it’s also good at others so I wouldn’t factor that in at that early stage.

So what are your goals? What’s stopping you from achieving those goals? And you’ll find that lots comes out.

And then I tend to use the impact effort matrix, it’s a really simple matrix, just Google it, or Perplexity if you’re into your AI, that’s the new Google but we won’t get into that. It hasn’t got quite the same ring to it.

But, the effort impact matrix, or impact effort matrix, helps you set out which of your blockers are quick wins, and which are strategic projects.

And the other 2 sections, it’s in 4 quadrants, the other 2 sections aren’t really to bother with because it’s if they’re high effort and low impact, there’s not really much point in doing that. And if they’re low effort, low impact, again, not much point doing that either.

So then you’ve got your list of quick wins and strategic plans and then you need to look at and see if AI is the answer, because AI is not a silver bullet.

I think a lot of people are trying to plaster and patch problems with AI where it doesn’t fit. It is incredible at some types of tasks, but it’s absolutely horrendous at others. And that doesn’t mean that there aren’t tools in those horrendous areas as well.

So I think people need to start to educate themselves on what it’s actually good at. And like I say, a great place to start is those repetitive admin tasks. If you’re doing something every morning and you’re doing something on repeat, it’s a great case for automation, it’s also a great case for AI.

And I think people often get those two things confused. They’re two separate things, AI and automation. They work beautifully together, but that requires quite advanced knowledge of both.

Where to avoid using AI

Bex Burn-Callander:

And it’d be good to drill down a little more into when not to use AI, because sometimes when customers are expecting a very human response, for example, it’s become quite easy I think to know when it’s a ChatGPT or a generative AI response to something.

Because there’s just a certain cadence, a way of speaking, that is kind of fine and correct, but it lacks warmth and it lacks a certain dimension of humanity, I guess.

And I know that, for example, certain editors in publishing won’t accept generative AI pitches, because they get so many that have just been journalists going on and saying, “Give me 50 article ideas in this topic,” and then sending them off to editors.

So there are places where you really shouldn’t apply AI. Can you tell me a bit about that?

Heather Murray:

Yeah, sure. I think anywhere where you’re trying to outsource your thinking. AI itself doesn’t have experience. It doesn’t have opinions. The way it’s trained is on a snapshot of the internet.

So imagine the internet has stopped and it’s been downloaded and everything’s been mashed together.

And what it does, it finds patterns, it mashes all that information together, whether it’s image, video, everything, and then it finds patterns within that information, not distinguishing between good quality and bad.

And then it gives you its best guess as to what you would like to hear. And it’s passable, isn’t it? You think it’s okay, but it’s really hard to put your finger on exactly why.

I’ve done a lot of work into why, as an ex-copywriter and somebody who ran a copywriting agency as well, I was thinking, “Oh, what is?”

Bex Burn-Callander:

Something jars, doesn’t it?

Heather Murray:

It does. It’s a recurrence of certain words, like “delve,” “embrace,” “ever-evolving,” those sorts of things. There are some really interesting graphs on the internet that show the evolution of the word “delve,” for example.

And this was in scientific papers, so these are people using ChatGPT to write their papers. So it sort of naturally kept on a line and then in 2022 and 20203, it just shoots up. So I think everybody started using it.

But certain words, certain terms, if you’re good enough, there is a way to get around that and produce human-sounding copy, but in order to do that, you need to know what you’re doing.

So you need to be a person who’s got a lot of examples of good quality writing, and the ability to identify good quality writing and describe it as well is very useful.

A copywriter, with the right instruction, can create genuinely good copywriting using AI, but it needs their skill and their guidance. You need to learn how to instruct it properly, how to prompt.

When people are talking about the term “prompt” or “prompt engineering,” it’s just the instructions, or more realistically the conversation that you have with AI.

And that is much more complex than people think. Because it looks a bit like Google, we tend to type in just a short sentence and we’re just like, “Give me 5 ideas for this,” and that’s nowhere near enough context.

My prompts are usually half a page to a page long, they’ve taken me hours to write, and I’m attaching 5, 6 different documents for background information, but the results I get are incredible.

And they really are, they’re indistinguishable from human writing, but it sounds like me because I’ve trained it to sound like me.

So outsourcing your thinking is definitely not good. Anything with specificity, particularly when it comes to image generation as well.

You can create these amazing images with just with one word, you put “apple”, and you’ll get all these different photographs of an apple. Whatever you want. Your imagination is the only limit with that.

But you can’t move something slightly in the image. You can’t slightly tweak a colour. It’s really quite random.

You could put that same prompt in again, and you’re going to get something completely different. Anything where it needs to be very bespoke and specific is difficult too. So not outsourcing your thinking and specificity.

I’m trying to think what else. I think it’s just that humanity. Anything that requires sensitivity. I’m saying that, but I think AI is evolving very rapidly in its ability to imitate sensitivity. It’s called artificial empathy that they’re programming it with.

For my newsletter, I actually joined something called Sonia.ai, which is a CPT therapist, an AI CPT therapist, with the full intention of absolutely slagging it off afterwards, I thought, “This is going to be so irresponsible. This session is going to be awful.”

And it’s done by voice. It says, “Lie down.” And I thought, “Well, I’ll give it the best chance.”

I put the phone on the pillow and interacted with it. And it was genuinely an excellent session.

I felt that this AI therapist was listening to me and understood, and the questions were really considered and the exercises I was given to do were really useful. And I felt I made progress towards a better me in that session.

And that really changed my opinion of AI and sensitivity and empathy as well. So it is possible.

But it will have limitations, obviously for very sensitive situations, and they need to be qualifying their customers a lot better because I was able to just go in and do this therapy session. They didn’t know what my background was, or my proclivities were or anything.

So it’s evolving very quickly, I think lots is possible, but it really needs to be in the hands of the human expert.

Bex Burn-Callander:

I can’t believe that about AI therapists. That’s blown my mind.

But then also, there’s a massive shortage of therapists and there are people waiting years to access any sort of help and counselling. So that’s amazing and maybe a bit wonderful, even if it’s simultaneously quite terrifying.

Crafting better prompts to get better outcomes

Bex Burn-Callander:

But, Heather, I wanted to go back to your point about prompts. So how can we learn? You mentioned that your prompts are half a page long and you attach lots of documents. That’s really fascinating.

How did you learn how to do that? Was that just trial and error over time, or can I go online and take a course in crafting these sorts of prompts? How do I create better outcomes in my generative AI?

Heather Murray:

Yeah, of course. I’ve got a 3 Cs framework that I built, and it was off the back of a ton of trial and error, and I mean a ton. I was spending 2 or 3 hours every day just learning, learning, learning.

And as a writer, that gave me a real leg up in how to construct and how to understand, “So I put this in, I get this out, okay.”

I like experimenting with things as well, so I’m very well suited to doing the job that I’m doing now.

But I soon found that there were 3 things that were really important, and I call them the 3 Cs. Character, context and clarity.

So character, it’s like a role play. Every conversation you have with something like ChatGPT or Claude or Copilot or whatever you’re using is a conversation.

It doesn’t know who you are, and it doesn’t know who it is, but it’s got access to this huge data set. And as we say, this mish-mashed information, and it needs to be narrowed down in order to give you the result that you want.

So if you ask a broad question, you’re going to get a broad and generic answer back because it’s got so much to draw from. It’s like going into the world’s biggest library and just asking for a book. It’s too broad. You need to be really specific as to what you want.

So character is really useful. If you’re writing some LinkedIn posts, you might say, “Act as,” or, “You are a LinkedIn ghostwriter with 10 years’ experience in generative AI,” even though it’s only been around for 2 years, this is AI, you can make it up.

But in doing that, it will say, “Right, I’m going to take from my LinkedIn knowledge. I’m going to take from my ghost-writing knowledge. I’m going to take from my generative AI knowledge.” And you’re starting to narrow down that huge data that it’s able to draw from and get more relevant responses, just in the first sentence.

You could give that character personality as well, if you wanted to and say, “You are really impatient,” or, “You are really helpful and warm,” or, “You are very inquisitive and you ask questions,” it depends what outcome you’re looking for. So setting that character.

Then tell it who you are, because people forget that when we sign up to these chatbots, we don’t answer any forms. You’re just in straight away. You give your email, you’re in. It doesn’t know who you are, or what you do.

A lot of people think that, “Oh, actually, it remembered this about me.” And there are things on ChatGPT that do remember little bits and pieces, but they’re pretty flawed at the moment.

So just think with every new chat, it doesn’t know anything about you. I would say, “Act as a LinkedIn ghostwriter with 10 years of generative AI experience. I am an AI trainer and I am looking to do X,” and that is a great start to a prompt. So that’s character, set in a character.

The next is context, so giving enough background. So the analogy I like to use here is, imagine that you’ve recruited an assistant and their job is purely to do this task that you’re trying to do when you’re doing this prompt.

What information would they need to have to do a good job? You wouldn’t just say, “Write me some LinkedIn posts.” You would say, “Okay, so write me some LinkedIn posts. Here are some examples of LinkedIn posts I’ve done before that I’ve done well. Here’s some information about my audience and what their pains and problems are. Here’s my tone of voice guide. Here’s some information about my product or service.”

They can’t possibly know that information without being given it.

And as soon as that clicks, you go, “Of course. Well, how would ChatGPT know that?”

I think people think ChatGPT or all of these LLMs know everything. And they kind of have bits of everything, but they don’t know which information to select.

So sometimes you’ve got to wave things in front of their nose. So using that pay-per-click function or the plus function on some of them, and adding some really good background information will get you much better responses.

And then the third thing is clarity. So being really, really specific as to what you want. Don’t just ask for a report, say something like, “I’d like a report that answers these 3 questions,” and give the questions. That’s always a really good one.

Or that has these different elements in like, “These are the section titles.”

Or actually, and you could go back to context and say, “Here’s an example of 2 reports I’ve written before,” and it’s going to read those and find patterns in those as well.

And just being really clear, breaking things down. So using white space, don’t just do a big dense block of text.

Break things down, “Here is your task. Here is my background. Here are the limitations. Here are some things you need to know.”

And then also just being really specific, “I would like a table with 5 columns and these are the headings.” Just explain exactly what you want because it’s got so much information it needs you to be specific.

From running my marketing agency, so any sort of marketers out there, think about a good copywriting brief, for example. It would have all the background information and all those documents. It would have who you are meant to be and what the purpose of this is all for.

So a copywriting brief would be the perfect prompt as well. So the 3 Cs.

Bex Burn-Callander:

That’s an amazing insight. I feel like you want to treat it like a work experience.

Presume no prior knowledge or experience or even the ability to take much initiative, just treat it like briefing a work experience.

Heather Murray:

Yeah, a really intelligent work experience person, who’s enthusiastic and keen to please and they just need pointing in the right direction.

And similarly, I think constructive feedback is really important. I watch people prompt sometimes in my classes and they will type something in and they’ll get something back that they’re not happy with, and then they’ll go, “Oh, it didn’t work. Oh, it’s boring.”

And I go, “Oh, no, no, that’s not the end.”

You then give constructive feedback, and you say, “Oh, I like this bit. I didn’t like this bit.” Or, “That was too broad. I want you to be more clear.” Or, “I want you to break it down more or add more detail.”

Just give that feedback in the same way you would with your work experience. You wouldn’t just say, “Oh, that’s rubbish, that’s it. Writing you off.”

You’d say, “Right, okay, so let’s work on this bit a little bit further.”

So that’s when it becomes much more, you start with the instructions and then it becomes a conversation back and forth. Hence, having this character that is assuming this role and knows who you are, so it’s going to respond accordingly.

Using AI to increase your revenue

Bex Burn-Callander:

And, Heather, you are so passionate about this, and you are so well-informed and you put it all in layman’s terms so it’s really accessible.

But tell me how you fell in love with AI, because you are a non-techie and you mentioned a background in copywriting, for example. So take us through the journey that brought you to this point.

Heather Murray:

Oh yeah, of course. Yeah, definite, non-techie. I never have the latest phone. Some people now look at me when I go and do speaking gigs and if something goes wrong, they go, “Oh, she’s the AI person.” And I look at them and go, “Nope, I do not know how to fix this situation. I’m not a techie person.”

I think that’s one of the most magical things right now about AI. You don’t need to be technical to use it anymore. We can now interact with just normal human words and we can build stuff. We can build our own apps, we can design stuff, we can create, and it’s just so exciting.

But I was running the marketing agency and we had a client that came up to us and said, “I’m trying to sell this big high-ticket thing, this £10 million thing. And we are knocking on all the doors and nobody’s responding.”

So I looked and I said, “Oh, how are we breaking down your audience?” And they were breaking them down by job title, by industry, by pain point, all the usual stuff.

And I had a little fumble around the internet and entirely by accident came across an AI tool. I didn’t even know it was AI at the time.

And this tool analysed people’s digital presence, so they’re called digital breadcrumbs, all over the internet. Wherever you have a profile or a bio on a website or you’ve commented on a YouTube or you’ve got a LinkedIn post, it hoovers them all up and analyses your personality and creates this sort of sales playbook for you specifically.

And at first I was really cynical, thinking, “That’s not going to be great.” There’s going to be 5 different versions and it’s just going to randomly allocate them, but it worked.

So we got $60 million worth of pipeline for this company, and then we did it again and again, $15 million more. So this made me go, “Hang on a second, this AI thing.”

I was a tiny company. My pricing was so low considering the results we got for them. I was just there sitting in my little home office in Birmingham, but that really piqued my interest in AI.

I’d started to notice and started to research and think, and this was probably early 2022, late 2021.

So we would pay a little bit of attention, but like everyone else, we were super busy, and just thinking, “Oh, this is interesting.”

And then ChatGPT came out, and I remember sitting here in this very office and I just thought, “What is this …”

I had 2 thoughts. My first thought was, “This is going to kill my marketing agency.” This can do or will eventually be able to do, and I didn’t realise how fast it would move, what my agency does now.

So much of the kind of persona-building work and all that. People started to do that in-house and I started to panic.

But then my second thought was, “I need to learn everything.” This is new. How often are you at the beginning of a new technology?

And I thought, “I don’t need to know any coding. This is really exciting.” So I also thought this is an entirely new business as well, but I didn’t realise how much it would do.

I decided to focus and spend a good 3 or 4 hours a day, every single day, including Christmas day, even though I was cooking for a massive amount of people, on that thing. When I get excited about something I go all in, and I just learned as much as I could.

So I ended up travelling around the world, met AI tool founders. I prompted and prompted and prompted, trying to work out how these things work.

And I read papers and listened to things and bought books and just devoured information. And then most of all, I experimented.

I believe that people running their own businesses just as one person can now get to much higher revenue levels by using AI. And I think that is by far much more exciting to me than big enterprises using AI.

I think those individual businesses can go, “Oh, I’m used to getting £10K month, now I can get £20K a month. That’s incredible. And I don’t have to hire anybody. I don’t have the hassle of hiring and the responsibility of managing a person, but I can do all these new things and things can run more smoothly, and my services can be deeper and better as well.”

So I started sharing what I’d learnt. I started a newsletter called AI for Non-Techies, and that just went whoosh. And so eventually, I think my eye just completely went off the ball with my agency because there was so much opportunity and people were asking me to speak and share what I’d learnt, and people were so hungry for this information.

So I ended up having to very sadly wind down my agency and my team as well. It was one of the toughest decisions I’ve ever made in my entire life. It was heartbreaking.

But in the end, one was going this way, and the other one was going this way, and you have to make these decisions as a business owner sometimes.

Yeah, that was March this year, and it’s just absolutely shot up since then. Because I was doing both for a while, doing both for a couple of years.

I was out doing the AI speaking and now all I do is the AI stuff. I train people, I speak, I build programmes, I do boot camps and things like that, and I absolutely love it.

So that’s how I got into it. It was by accident, as is the case so many people have.

Bex Burn-Callander:

That’s the best business idea or best business journey is when it’s like, “Oh, I just couldn’t not do it. I stumbled across this thing and it reached out with its arms and it grabbed me and then that was it. That was what I was doing for the next 10 years.”

Grow your business by becoming an early adopter of AI

Bex Burn-Callander:

And this is an area that’s exploding, and there are a lot of people out there that need help and need the education piece, but is that going to be the same in 10 years time? Are you still going to be giving the same sort of training, only, I don’t know, 2 levels up?

Because there’ll always be a gap between what the person, regular person on the street knows about AI, and what the actual potential for it or what the kind of usability has reached.

So is that a bit of job security, I guess, for you going forward?

Heather Murray:

Yeah, I think so. The technology adoption curve, where you’ve got the early adopters, and we’re not even in the early majority yet, I think the vast majority of people haven’t even begun on this journey yet.

So where I’m positioning myself right now is AI for beginners. Non-technical. Because I see this too much and I still see it, I’m often popping into things and seeing what other people are doing, and I find it starts about 3 stages further than a lot of people need it to.

I usually start all my talks by going, “This is what I’m talking about by the way. This is generative AI.” And they go, “Oh, okay, right. I get it now.”, “So that’s what’s different. Okay.”

And then I’ll just explain things in a really basic way. I hate the term “for dummies,” but you remember those books.

Just simplify things, and that’s what I plan to do.

So sticking with that beginner book. My forte is working with other small businesses and other people who want to, just like me, who want to double, or triple the size of their business without hiring. And just to see what that feels like to have this beautiful, almost like other team members with specialisms, but they’re AI.

And then knit that all together really nicely with automation and you’ve got something really quite special.

Be wary of AI bias

Bex Burn-Callander:

And do you have any worries? I mean, there’s a lot of fear out there about the dark side of AI and what it could do to jobs, to the way that we socialise, the way that humanity feels about itself.

Do you share those fears, or are you just really quite bullish on AI and really positive about the impact it will make on the world around us?

Heather Murray:

Oh, absolutely. Every technology in the hands of bad actors for a start is a scary thing. Look at the internet. But there’s a huge flip side, and I think I’m 50/50.

I’m very much a kind of live in the present moment, and I’m thinking about what it can do right now.

But when I think ahead, and a lot of people ask me this question of, “Where do you think we’ll be in 2 years’ time? What will have changed and where do you think we’ll be in 10 years’ time, 20 years’ time?”

It’s really difficult to say because it’s changing so quickly. I know it’s a cliche, but it really is. As somebody who works in it, I’ll do the same session 2 days apart and something will have changed. I can never get my slides right.

But there are lots of risks. There are lots of ethical issues as well around generative AI. I think we’d be talking about them for hours if we got really into that.

But things like the possibility of misinformation has now just got so much worse because of people using AI incorrectly. And that’s the main reason I’m doing what I’m doing, making it accessible and helping people understand how to use it properly.

If you don’t use it properly, if you’re just trusting what comes out the other side and you’re not fact-checking it, because it hallucinates, it makes things up, it’s biased.

Look at the data set that it’s being trained on, it’s trained on human content, which means it’s got human bias.

It’s built by humans, and then it’s coming out with this bias content. So we have bias, we have the potential for deepfakes as well.

I can create a video now with a couple of sentences in 30 seconds. Again, in the wrong hands, that’s a terrifying thing too. I think safety is so important, and transparency is so important too.

Become more transparent about using AI

Heather Murray:

So when we were talking earlier about the shame of using AI, that people feel like they’re cheating, I think we need to learn to be much more open about our AI use and say, “Hey, I used AI for this”, so that we are not starting to cover things up.

And I love LinkedIn, they’ve just started to do this thing where you get a little CR on the images.

Bex Burn-Callander:

Yeah. I was just going to mention that because I love the fact that it was, “Use AI to expand your reach,” or something. And then it puts a little tab on it that says, “This was made with AI,” which I love.

Heather Murray:

And I’ve seen a few startups now who are building products like that all over the internet, so you’d be able to click on it and see its provenance.

So I think that is such an amazing and important step as well. I don’t know how we do that with text, but I think there are a lot of people working on that too.

I mean, there are a ton of risks and ethical issues, but there’s a ton of mitigations being worked on as well.

How to use AI safely and securely

Heather Murray:

So it’s like with any new technology, and we really are in its infancy, which is another thing to know for people who are just starting out using it, it breaks all the time. All the time.

It’s very rare that I have a day where I’m training somebody on, say, ChatGPT and it doesn’t do something weird. So don’t be put off if it does, just give it a minute, give it an hour, it’ll come back.

Or if it gives you a weird response, don’t write it off. Just be very careful.

I think the most important thing to know right now for people starting to use it is that whatever you put in, could come out in somebody else’s response.

So when I’m talking about attaching documents and things like that, be really, really careful with the information that you give it because it’s being used to train and it could, it’s very unlikely as I say, but it could come out in somebody else’s response.

So avoid sensitive data, if you’re using free versions of things. Generally, you pay for data privacy when it comes to AI.

So for example, I have Chat GPT Teams edition, and that’s because I am constantly feeding it sensitive data. I’m doing lots of work on it and I know that I’m secure there.

But in the free versions, your data is not secure at all. So anonymise it or just play around without it.

But if you do want to start feeding it real information, get one of the paid versions instead. And just look into the security side, do your due diligence in that way, I think it’s really, really important to do that.

Who owns the copyright for AI generated content?

Bex Burn-Callander:

And is it true that if you generate any sort of content on these platforms, you don’t actually own the copyright? Is that right? It’s all owned by the AI company itself.

Heather Murray:

It’s a Wild West. Nobody’s able to answer that question right now, and a lot of people ask it to me. When it comes to regulation right now and who owns what in IP, it’s a Wild West in the UK. We’ve got the EU AI Act, which is obviously going to set a precedent for us.

But at the moment in the UK, we’ve got bits and pieces, and the bits and pieces are great, but they’re not joined up. And that is something we need to be working on right now so we’ve got set clear rules for who owns what and how transparent we should be and what needs to be declared and risk levels and all of that sort of stuff.

But yeah, there’s no answer for that yet. That’s how early we are into this. And this is rare for a technology that we all have such easy access to this new thing.

 And it also makes it a dangerous thing as well, that we can all go so wrong, so easily. But nobody knows who owns it.

Bex Burn-Callander:

It’s rare that I feel pity for lawyers, but my goodness, this is a mind-boggling area where a lot of work, a lot of man-hours, are going to go into figuring all this out.

So I do, I have sympathy for the people who are going to try and put this all down in rules and laws and make it all comprehensive. That’s a big job.

What does the future have in store for AI?

Bex Burn-Callander:

And, Heather, we talked a bit about the future of your business. Now while I was talking to you, I was just imagining it, we’ve got holograms already.

 So if you wanted to grow your business, could you just create an AI version of yourself to go out, flick a switch, the hologram comes on, and you are delivering training?

I mean, could this be something that, I don’t know, in 5 years time even you could do if you wanted to grow your training business. Is this the way technology is headed? Is this the kind of thing you are thinking about when you are looking into the future of AI for Non-Techies?

Heather Murray:

Yeah, I mean, I’m really torn because I deliberately deliver my training in a way that’s so human. It’s just me talking. There’s no slickness. If I make a mistake, I deliberately keep it in as well, if I trip over my words. I think it’s really important now.

I’ve actually got an AI version of myself, which is an AI avatar trained on my image, not quite in hologram form yet, but that’s bound to be coming soon.

But this digital Heather, she looks like me, she sounds like me. My own dad, when I played the video, couldn’t tell it wasn’t me. She has my voice. All you do is you type in what you want her to say and she’ll say it.

Bex Burn-Callander:

You deepfaked yourself.

Heather Murray:

I deepfaked myself. Yes. Which seems really stupid when you think about it. A lot of people go, “Are you not worried that you’ve actually given your likeness?”

It’s with a company called HeyGen and they have really good guardrails. And there is a lot of trust you have to put in these tools if you want to experiment, and I chose to trust this tool.

So I’ve had a look at avatars for training purposes. I think the beauty of training, is that being engaging is about responding to what people are saying and doing.

And also, just the thoughts that come into your head as you’re talking. It’s not a rehearsed script. I would never sit and read off slides, I’ve never had a script for training in my life.

I like to go, “Oh, and that reminds me of this,” and “Oh, this came out yesterday and this is actually really relevant.” Or, “That’s actually not as up-to-date as it could be because this happened yesterday.”

I think that type of lively delivery would be something that AI would really struggle to do. It could be possible in the future. It could sit and listen to me delivering training and be able to mimic it and then be tailored into the news somehow.

There’s something called Google NotebookLM that came out that really made me go, “Whoa.”

It’s free. I really recommend people check it out just to see where voice has got to, AI voice. So it’s free. You just load in whatever information you want. I mean, you can just type in a paragraph. I typed in to get a paragraph about my husband leaving his career and starting this business that I’m going to help him out with, and just like a few sentences, but you can upload loads of stuff.

I also uploaded a different one with all of my AI training in it.

And it creates a podcast with 2 humans talking to each other about the topic, and it comes up with insights that you hadn’t even thought of.

But the podcast, you would never guess it was AI. They’re laughing, they make jokes, they express opinions. It’s incredible. And that really made me go, “Oh, AI voice is a lot more advanced than what we’ve got access to right now.”

So there’s so much that the humanity of this podcast was so realistic that I’m not deeming anything impossible right now, but from where I am right now, I think very human delivery is what I’m focused on.

Because I think that’s keeping me apart from competitors, is just that I’m very normal and I think people relate to that like, “Oh, she’s like me.”

But I’ve spent the last 2 years researching AI, and that’s the whole point. I’m not a gatekeeper of knowledge, or that I know things you could never understand. It’s not that. The whole point is that you can understand it, it’s really easy. Let me show you.

Bex Burn-Callander:

And your enthusiasm is infectious. I’m literally on the edge of my seat and I’m leaning in, my head is going to look enormous on YouTube because I’m properly bent forward, hanging off your every word.

So I do think that you definitely have a niche with that. And I feel like we ought to apologise to our listeners because they’re going to lose, what, 40 hours next week at least…

Heather Murray:

Oh, at least, at least.

Bex Burn-Callander:

… playing with all these tools, going down the rabbit hole. I’ve been scribbling down things to play with. Maybe I’ll spend a day where I’ll send all my emails in the style of George Orwell.

Heather Murray:

Yes. Exactly. And that’s another really important part is play. Play safely, but play.

This is the time to be playing around. A lot of the tools are in their infancy, but getting to understand how they work and how they respond will prepare you for when they’re not in their infancy and they’re a bit more advanced.

And they’ve got all these capabilities that will jump you ahead of everybody else who just starts to learn at that point.

But it’s so much fun. It’s loads of fun to play around with these. I’m constantly squishing people and turning people into cakes and writing random things and all sorts of different things.

My husband is sick of me coming home and going, “Look at this.” And he goes, “What is this now?” But it’s fun. It’s a fun thing. So play, but play safely.

A fun place to start your AI journey

Bex Burn-Callander:

And then I guess just finally, so if someone who’s tuned into the show has 1 hour that they’ve set aside to look into this tech, what would you tell them to do with that one hour? Where would you send them? It’s their first play, as you say, with this tech, where should they go?

Heather Murray:

I still think Chat GPT, its free version is really broad, but try to think outside of content creation.

Ask it to act as an interviewer for you. Ask it to be these different characters. I think just playing around with characters for an hour on ChatGPT.

If you are applying for a job, you can set it as your interviewer and you can practice. You could say, “I am X. I’m applying for a job X.” You can even attach the job description or paste the job advert in.

“And you’re going to be the interviewer and you’re going to be quite harsh and ask difficult questions. Start the simulation.” And you can go back and forth, back and forth.

Imagine the people that you would like to book a one-to-one with, but are probably a bit too expensive, set that conversation up on ChatGPT and I bet you’ll be surprised.

You could say, “Act as a sales consultant that usually costs £1,000 an hour. I am a business owner, and I would like some advice on my sales. Ask me the questions you’d normally ask in this situation and come up with a plan at the end.”

Then you’re going to get this amazing expert advice. It’s incredible what you can achieve. So yeah, ChatGPT, the free version, is a really good place to start.

Inspired by this business story?

Wherever you’re listening or watching, subscribe to Sound Advice on Apple iTunes here.

We are also on Spotify and anywhere else you get your podcasts.

Join our community to share your insights and stories on X @SageUK using the hashtag #SoundAdvicePodcast, on Instagram @SageOfficial or in the comments below!

Want to know more about Heather and AI for Non-Techies?

You can find out more about Heather on her LinkedIn.

You can find out more about AI for Non-Techies on their website.