AI Expert Explains: 97.2% of People Are Using AI WRONG

0:00Amelia Miller:70 % of the skills we use in our jobs at the moment will be obsolete by 2030. My name is Amelia Miller. I'm the co-founder and CEO of a company called Ivy. Recently named Forbes 30 Under 30. Famously, 84 % of the world have never used AI. But a lot of people are still utilising AI to maybe 5-10 % of its true capabilities. And I had AI running for me in my sleep. The worst thing you can do for yourself or your brand or your company is to be seen sounding like you're using AI. And a grad hiring market has taken a massive nosedive because those are the roles that AI is automating. Large companies, they are using it to reduce headcount.

0:40Amelia Miller:It might spend thousands of pounds on your card or your company's card. Every CV that had the word women in it, for example, Women's Chess Club or Women's University, it immediately rejected those candidates. Any bias that exists in the world is amplified by AI. Now is the time and there's no women in the room. Every company is now an AI company or it will die. Can you be on a train in two hours to film an episode of Dragon's Den at 9am tomorrow morning? And I turned to Lydia and I said, we need to create a fake version of the platform. Amelia, we have a meeting. Welcome to the podcast. Thank you for having me.

1:21It's so good to have you here. big open question to kick off. Who are you and what problem do you solve?

1:27Amelia Miller:My name's Amelia Miller. I'm the co-founder and CEO of a company called Ivy. And we are solving the problem that is AI is here to stay. And nobody is training people on how to use it, how to leverage it, how not to get left behind. And doing it in a fun and interesting way that isn't really boring. Nice. I like that. It's going to be a good conversation. so if anybody's listening and they've just tuned in why would you be the best person to give them advice and stick around i'm not i'm kidding i'm kidding i can go through my roster um i'm i would say above everything cv wise i'm i'm just really passionate about this space i'm really passionate about the fact that i don't want anyone business or individual to get left behind as ai just completely transforms the world it's incredible what's happening but it's moving really quickly and it can be very scary to a lot of people and I think with that you need to have a huge amount of passion for actually helping people learn rather than just making a bunch of money or adding to the overwhelm by making an AI tool yourself to just throw on the pile of the thousand that have come out this week so I think my passion makes me well equipped in terms of my yeah CV or roster or whatever you want to call it I'm I'm academic by trade so published researcher from the University of Cambridge I'm currently a fellow there at the moment looking at skills prediction in the labor market so things like you know in previous tech revolutions what has happened to the workforce what skills have been needed what skills have died out I then spent some time in banking at Goldman Sachs in London here in foreign exchange sales and trading half loved it half hated it um and realized quite quickly that it wasn't scratching the entrepreneurial itch so then I jumped into the world of startups haven't looked back really um I'm a member of the women in tech UK government task force so do a lot of work with the government around how they're thinking about AI skills particularly with a gendered lens in in my work there and then recently named Forbes 30 under 30s not that that has anything to do with what we're building at ivy but hopefully it's a good you know uh string in the bow to show that potentially we're we're doing good work but i'll let you guys be the judge of that congrats i feel quite intimidated with that uh with that cv um i'd love to know from your perspective what you think people are getting wrong about ai i mean what's the big frustration you think i wish people would understand that a little bit better it's a great question i think what the main thing a lot of people are getting wrong at the moment is they're using ai like a glorified google search uh so they have their chat bot maybe they use chat gpt maybe they use gemini maybe they use claude and they have great conversations with it but it's a back and forth conversation and that is really level one of 10 in terms of where we're at with ai right now and most people are still in that phase if in that phase at all because famously 84 of the world have never used ai wow and everyone listening to this podcast i'm sure uses ai we even we live in an echo chamber we live in a bubble um and but there are levels to it where ai is extremely powerful is where you break out of that chatbot stage where you need to be there you need to be present you need to be babysitting your llm or your model um and you can really get into the automations the agentic workflows that's where AI is incredibly powerful nobody really understands how powerful it is yet but the technology is here it's live it's been built but a lot of people are still utilizing AI to maybe five ten percent of its true capabilities could you maybe think of an example of the maybe the first time or the most impactful time that you were like oh this can do something totally different for me in business great question I think for me it was the first time I used a tool called n8n which is an agentic you know automation tool and that for me was the first time i personally exited the glorified google chatbot stage and i had ai running for me in my sleep i remember one particular where we'd had an inbound at 3 a.m from a u.s client obviously i was asleep and when i woke up the next morning an email had been sent that sounded deeply human a meeting they had replied a meeting had been scheduled it was sat in my calendar for midday or you know 2 p.m the next day with that client because they were really keen to jump on the phone and for me that level of pace and that level of taking technology applying something human to it in the copy that that you have written or you want it to write um and the like marginal gain that that did for us as a business was just obviously one tiny example of the power of this tool where it is working for me and making me money in my sleep wow so let's take it back to i guess the the core because there'll be people listening to this that are maybe like using it as chatbot or they're dabbling and some of the tools that you spoke about we're going to get into further into the episode from my standpoint there's probably like three core ai tools that are your your classic chat bots which are chat gpt gemini and claude when you think about those three tools i'm sure there are different ones as well but what are you using and for what or are there any tools that you think actually get rid of that you're wasting your time absolutely well yeah the ones you mentioned And a lot of people might not be familiar with the fact that there really are only three main foundational models.

7:21Amelia Miller:What do I mean by that? That's the brain power. That's like the furnace of AI. And those are Anthropic, who have built Claude, which is their large language model. Google DeepMind, who have built Gemini. And then OpenAI, who have built ChatGPT. and they really are the ones putting can't even tell you how much money into building and training these models and I can obviously go into a bit more detail around what it actually means to train a model everything else is not a foundational model it's what's known as a wrapper so they actually pull on those models so Copilot is a great example Copilot have not you know Microsoft have not built their own model they have built a wrapper and they pull on used to be only open ai but they've now integrated the others not that that's you know massive myth busting but i do think it's kind of important context you understand actually there's only three foundation models the exact ones you mentioned in terms of what they do it really depends on what you need to use ai for they are very similar in their capabilities and that's one thing to know it's if you're using ai just to help you in your day-to-day.

8:31Amelia Miller:It really doesn't matter which model you use. A lot of people appear to be like, you know, oh, Gemini is terrible at this and really good at that. It's a little bit better at one or the other, but for the most part, they're all incredibly powerful. They're all trained in a very similar way. So it's kind of six and two threes there. But if you did want to break them down into blanket nuances, we'll start with ChatGPT. That was the first that came onto the scene when OpenAI released it. And that's really what brought conversational ai to us laymen um it's great for everyday tasks i call it this a bit of a swiss army knife anything from copywriting to light touch data analysis to um research it can it can do everything quite well gemini is much better for analytics so more statistical things so if you work in like data transformation analytics or you you just want to create a chart create a table it's a little bit better for that claude has blown everyone out of the water recently they're moving at a pace that is genuinely unprecedented with the like capabilities of their models so they're the best at everything um i'm not sponsored by anthropic not paid to say that but yeah they can do everything extremely well if they were to have one strength i would say it would be the ability for them to have nuanced copywriting so if anyone is looking for a model that can write in a way that feels deeply human aka no ai slop no m dashes no and honestly blah blah blah uh i would recommend claude particularly that opus model so you have it's genuinely gobbledygook but you have the parent companies and then you have the llms and then underneath that the llm has different models that they release so you might hear of people talking about the gpt 3.5 model or the clawed haiku model i don't know where they they get these names from i don't genuinely go like random name generator um but that's kind of how it works but it can make the ecosystem feel quite confusing and overwhelming because you're like what's an llm what's a model what's what are all of these names getting thrown at me in the news why does nvidia fit into all of this um wait what there's physical chips involved it can definitely feel super overwhelming which is why you know we founded ivy to help break that down because having a having a foundational knowledge and a bit knowing a bit of the theory is really helpful in terms of the way you think about like interacting with these models so then let's imagine that barry 52 from swansea is listening to this and he hi barry at home if you're watching um and he heard you talking about um automating an email that goes out at three o 'clock and he goes but i've been on linkedin and every post looks the same and actually i think stuff like that needs to be human what would you say to him to make him feel more comfortable to hand over the keys to ai it's a really good question and barry is thinking about it in exactly the right way because a lot of people um are very quick to use ai to write outbound sales messages or linkedin posts or linkedin dms and it does not sound human it sounds like ai slop and there's this new term that's come out called work slop coined by uh this lab in at stamp uh at um yeah at stanford in the us they wrote a great like super interesting report on it but basically people that don't know how to use ai and crucially don't know how to prompt well enough will output awful you know awful sounding copy so to someone like Barry I would say you can absolutely use it but you must get your prompting right and that's why prompt engineering is still the number one skill that is needed to be able to use AI well if you can prompt it well what do I mean by prompting well it's how you interact with AI it's the copy you say the instructions you give to elicit the the output to get AI to do what you want and if that is a sales message or if that is an email replying to that 3am inbound the prompt needs to be good enough so that that message does not sound like it's been written by ai because the worst thing you can do for yourself or your brand or your company is to be seen sounding like you're using ai and i'm sure everyone listening can relate to that which is like i don't want this to sound like it's been written by ai or i've just received a linkedin sales message and it's full of m dashes and it looks unprofessional it looks it looks sloppy um but there really are like very you know sophisticated prompts you can do an example of our 3am message is i'm kind of giving away our secrets here sorry if anyone's um received this from me and thought it was i thought it was me at 3am it wasn't it was ai but you know i i have i've programmed nan to always start those messages with uh they they always get emailed four and a half minutes after they submit their inbound interest form why because it's not straight away that it's obviously automated and it's and it's uh phrased to feel like i've kind of typed it out on my phone and it says something along the lines of you know hey rachel um i'm traveling at the moment but i just saw your form pop up so i thought i'd shoot you a quick email would love to chat a little bit more about you know insert company name and and that's where the um agentic ai layer can get quite sophisticated here because you can take the problem they've written on that inbound and weave it quite naturally into that email but you must in your prompt give it the parameters to say you know only include a line of it here because we want this to feel very typed on an iphone would love to pop some time into chat i've actually got you know 15 minutes tomorrow here's my booking link i am very in control of what that email says ai is helping me with that one sentence um and subbing in the name and the company of course but if i hadn't given it a sophisticated prompt it might say you know hi rachel thank you so much for your interest m dash so hopefully you kind of understand what i'm trying to get out here it's all about the puppet master and it's all about how good you are at orchestrating these systems but it does all come back to prompting um and prompting is one of i mean it's what we teach the most at ivy because if you can prompt ai well you can be unstoppable oh god i'm so tired i've just been talking to qualified prospects all day without any wait time how have you been doing it because i've been cold calling people on my mobile and i reckon i've spoke to one person no no no stop that right now and use nooks.ai why what what what does nooks do it finds me the right people to call and it calls them and it connects with them over and over and over again so you're saying it's nooks.ai and i can just be speaking to qualified decision makers all the time no other better connection right wow looks at ai i think one of the assumptions people make based on what you've just said is obviously as a business you won't be hiring anyone now because you'll be just saying well ai can just take the keys to most things why would they be wrong to think that yeah again it's a really good question most most startups fail 90 of startups fail um because they run out of money and we wouldn't have been able to get to the point that we're at now where we are now you know hiring relatively aggressively if we hadn't been able to leverage ai to get to this point and and get to a point where we are generating revenue so i personally believe that ai is extremely powerful especially for startups and scale-ups to be able to grow to be able to hire um it it is potentially a different story you know very transparently right now for larger businesses who are potentially looking at ai as a way to freeze hiring and we're seeing that happen in the job market you know i'm sure no one's living under a rock but most large companies are are not hiring as much as they used to the grad hiring market has taken a massive nosedive because those are the roles that ai is automating so it's a bit of a it is an interesting question because it's a different story large companies they are using it to reduce headcount small companies it actually gives them a fighting chance to build a successful business to then grow that business and hire a great example is some work we're doing with one of our clients at the moment they're a tiny three-person startup and they're a dog food brand called poor drinks great product shut up and there's only three of them and their their founder is spending two hours a day in their customer service inbox manually replying to emails and is not a good use of his time he cannot drive the business forward he's leaving revenue on the table because he's not able to even do little things like if someone has something good to say he's not able to say great leave a trust pilot review do a linkedin post because he's just so time poor he won't reply or if someone's cancelling their subscription it's easier for him to just process it than it is for him to you know take them down a loop of trying to keep them and so what we're doing with with him is is helping him automate that customer service inbox still with a massive human in the loop he wants to press send on every single email but we're using we're helping him use ai to draft that and that means he can use his time to actually move the business forward to then in five years time be a 1000 person organization with a huge customer service team so it's a really interesting kind of dynamic but you know i'm not going to lie here and pretend and say oh it's amazing because ai can mean businesses can grow in the future some can some are using it to cut head count and that and that is happening and that is a fact and that's a very scary part of the ecosystem right now i suppose and if you could speak to the job market the people who are maybe you're looking to reskill or maybe just make some predictions about the next 12 24 months where would you say spend your time here and actually ignore here like what what predictions can you give people what we know is that ai fluency is the number one skill that's in demand in 2026 2026 is the first time this skill has kind of usurped any others that's followed by the the second most highly coveted skill which is critical thinking and that's interesting because oh it's interesting to me um you need to be able to use ai with the ai fluency but you need to be able to put that critical thinking on top of it aka not deliver ai slop uh prompt well review and refine those prompts and that's where that critical thinking comes into it so to anyone listening that is either looking for a job or is in a job and is worried about their job or wants to get promoted upskilling on ai is the single best thing you can do for your career it's here to stay whether you like it or not you know burying your head in the sand is kind of no longer an option so learn these tools leverage them get stuck in and show the world you're using them i always say this whenever we finish a learning cohort with our ivy learners like great you've learned it you've got to shout and scream about it to the world now you've got to get it on linkedin you've got to get it on your cv companies are screening for ai tools and ai terms on cvs so you must position yourself as that AI fluent candidate.

20:01Amelia Miller:Most companies are now actually putting AI fluency tests into their recruitment process. A great example is a tech company called Zapier. At every stage of their hiring process, they assess for AI fluency. And these assessments are getting a lot more sophisticated now. So there's no longer room to hide. Because access to technology and barriers to building technology are so much lower, it means that roles like marketing, brand, and creative are more important now than they have ever been. But that paired with AI fluency, it's always the winning strategy. So let's make this a masterclass in prompting.

20:39So I went to the dark side on the weekend and I got Whisperflow. Yeah, Zach's not heard. I've been banging on about it and I'm trying to get him convinced that it's the next thing. What does good prompting look like?

20:53Amelia Miller:A good prompt is a contextual prompt. and it's kind of as simple as that the more context you give and the read there is the longer the prompt the better ai will understand what you mean i like to liken your chat bot to i don't know if you guys have watched suits mike ross in suits who is you know basically a phenomenal intern if you will and he has a photographic memory if anyone hasn't watched suits that's that's who Mike Ross is AI is like that but it has to follow instructions so when you ask it to perform a task you need to give it the context of the task you need to tell it what the output should be is it a 10-page document is it an excel table is it an image is it a video all of this output is now possible with AI but you have to tell it you in a prompt should never leave any opportunity to guess because when ai doesn't know it will usually deliver something that is sloppy and that's why coming back to that point a bad prompt equals ai slop generic waffle so there's a few different areas of prompting but i'll kind of touch on the most important ones the number one tip i have for anyone this will actually change your life with prompting is this idea of a persona a prompt always start your prompt with you are a and then you insert a term that you want the chatbot to behave like in the context of this task so i'll give a practical example let's say you want to make an onboarding doc for someone that's joining your team you would say you know you are and let's say you work at a series A tech company, you are an expert HR professional at a series A tech company.

22:47Amelia Miller:Create me, Anne, and then you would go into the rest of the prompt. Why? Because you have given AI a role, and so it deeply understands from the outset, oh, this is how I need to behave. What would an HR professional do in this example? And it understands, oh, okay, I work in HR, so I'm going to produce an onboarding doc that looks like an onboarding doc. If it's vague and you just say, create me a doc for a new hire, it has no idea what to do there. So that's one example. Another one is tonality. So you have to say, this should feel professional. This should feel friendly. This should feel XYZ.

23:27Amelia Miller:If you don't give it the tone, it won't know how to react. An example is with all of my emails, I always say a friendly yet professional tone. Negative constraints is another top prompting tip. So what a lot of people forget is actually to tell AI what not to do. I use this a lot. You know, I say this should not exceed 200 words. This should have no em dashes. This should have no AI tells such as the word delve or, you know, unnecessary verbs or adjectives, which AI loves to put in um so yeah telling ai what not to do a great other prompting tip is this concept of it's called few shot prompting and it basically means you give ai a few examples so let's say you're using ai to help write your linkedin captions um which a lot of people clearly do and instead of just saying write me a linkedin caption about speaking on a podcast this morning i would actually say you know you are an expert viral linkedin caption writer for seed stage founders write me you know write me a linkedin caption about x y and z see below for three examples of my previous linkedin posts sounds really obvious um but that's called few shot so shot would be like an example in this case and you're telling ai this is what the output should look like again so that when it's giving you a response it knows how to behave it's seen those examples so it can emulate them nice and there's yeah a bunch of others which i won't go into detail on right now but one thing i will say about prompting and to people listening they're like well that's great you know this i know i need to do that but but time you know i don't want to type out paragraphs and paragraphs every single time i'm prompting because i prompt a few hundred times a day which is why we we've recently built and launched it is completely free but it's a prompt improver it's a extension it's called the ivy prompt improver and so you can type a really basic prompt or you can use whisper flow and this is what i do and it's like a glitch in the matrix i use whisper flow which is this voice dictation tool to be like you know write me a linkedin caption about xyz not not contextual probably quite brief i'll probably give it a bit more context than that you highlight the prompt and you just click improve with ivy and then ivy takes all of the logic that I've just told you all of that like you know theory and then improves your prompt for you gives it the right persona gives it the right tone and then you can just say use this prompt and your prompt will change live to a much longer prompt to save you a bunch of time because as people get better at prompting it then takes them more time and especially for people that use AI you know a lot in their day-to-day it can it can start to really eat into your time typing out these prompts that's a great idea just say that again where to find that it's in the chrome uh it's a chrome extension free it's called the ivy prompt improver nice wow we are all absolutely screwed okay so so that goes back to like level one you were talking earlier so means that tonight we're staying in london we are going to be sat there with claude we're going to have the the the ivy chrome extension and whisper flow i'm going to get him to download it and we're going going to be going hard on the prompt tim but that still feels like level one like you're just having a conversation with an llm what are the next levels if you were treating us like idiots of like okay you're going to run a business we're going to automate everything what should we do next yeah the next level i would say is the concept of an ai browser so this is where ai can actually live inside your screen and see the tabs you're in rather than because right now i'm sure most people listening have you know their chat gpt tab or their claude tab and they go to it and then they go back to maybe their email and they go back to the tab imagine that tab followed you around that's how i describe an ai browser so again every ai company have launched their own ai browsers so claude have theirs open ai have one called atlas perplexity have one called comet and then there are some startups that have built ai browsers why is an ai browser important because it can do two things it means you don't have to jump back to the tab an example I like to give here is let's say you're like cooking lasagna for dinner um and you go on google and you you know you want to use Jamie Oliver's recipe so you go on to the Jamie Oliver website and you're like what a great lasagna recipe I'm cooking for two people but this recipe is listed on the website for 12 people maybe if you don't use AI at all you'll do the maths on a calculator if you use AI via a chatbot you'd probably copy the recipe go to chat to pt on your other tab paste it and say hey give me this recipe for two people and it would give it to you an ai browser is already in that screen so you would it's and it's imagine it like a sidebar so it can see you're on the recipe so you don't need to copy and paste the recipe and go to another tab you can just say give me this recipe for two people and it's in the screen so it can understand and do that but where it can get really clever is it can perform actions on your behalf so ai browsers are so let's say you're looking at flights i do this actually a lot i'm on sky scanner and i've got an ai browser open and let's say i want to fly to rome next month i don't care about the dates i just want the cheapest flight and i want to be there for four days i could give that as a prompt into this ai browser i could sit back i would not be touching my screen and you would see your screen working and something is getting typed into the search bar the dates are getting changed you are not touching your screen and the cheapest flights within you know staying there for four days would be found but it's ai doing it all on your behalf um you could go and make a cup of tea go for a walk and you come back and the cheap cheapest flights would be there on your screen that's how ai browsers work but it can you know it can do that at a really complex level um it's great on linkedin if you wanted to let's say you were job hunting and you wanted to dm every single hiring manager in london in that works in marketing but you wanted to make the dms personalized you could have that running on an ai browser for your whole morning and you could dm 400 people while you're getting on with other work maybe you're working on your cv on the side so ai browsers are extremely powerful the next step beyond that is ai automations automations are nothing new we've had automation tools for years things like zapier make you know.com um but where ai can get really interesting is you can add an ai layer so in so an automation it's very simple it's like if this happens then that happens it's very rigid you have to set the steps you have to say you know when an inbound form comes in then send me a slack message where ai makes it really interesting is you can have an ai step there and you could say when an inbound form comes in use ai to like categorize how important this is and only the most important ones send me a slack message and then the next level beyond that is agentic ai a lot of people don't know what agentic ai means and kind of nod along at the pub when someone mentions it.

30:58Amelia Miller:And what that means is you don't set the steps. You just give AI a goal and it does whatever it needs to do to achieve that goal with the parameters you give it. So you might only give it access to a few of your tools, but you could in theory give AI, and when I say AI, you know, a tool like Claude or Claude Cowork, access to every single one of your pieces of software on your screen your bank card and you could say actually not only find me the cheapest flights to roam but book them um or you could say your goal is to fill my diary next week with like sales calls do whatever you need to do to get there and all you do is give it a goal and a gender ki is goal oriented and it and it understands what needs to be done to get there and it will do it and that can be a blessing and a curse because if you do not set the parameters and like the human in the loop approvals it can run wild like if you give claude co-work your bank details it might book a first class flight to rome you know you just you never know and again it all comes back to prompting if you have not given it enough context in your prompt it might spend thousands of pounds on your card or your company's card and that's why for any employers or people that lead teams that are listening to this these guardrails are one of the most important things to think about when you're rolling ai out across your company which i'm sure is on everyone's minds already because it's the number one thing companies come to us with which is like i want us to innovate i want my team to use this but I don't I don't want them to make mistakes I don't I don't want them to use Claude Cowork or use an agentic AI and it you know fundamentally make like huge mistakes whether that's a brand reputation mistake whether that's like a monetary mistake all of these things are very important to think about so you do need to set the right guardrails when you're setting these up and it's it's not it's it's not difficult to it's as simple as setting the approvals on the admin settings but if you don't know what you don't know you might let everyone run wild with a tool that could cause a bit of harm one of the interesting things that I don't use about this as well it feels like the market is quite easily influenced by maybe Joe Rogan maybe diary of a CEO a little bit so there was a period of time where it's like replet is gonna be like the next big tool and then I saw on LinkedIn everyone's building replet tools for a little bit and then before that it was like chat gbt for everything i don't think anyone even knew there was other products available and obviously now it's clawed i i imagine for you it's something you've had to kind of keep it on top of i kind of need to be a little bit ahead of like what the next tool is going to be um so what have you found works to keep you on top of where you think the market's going to kind of stare yeah great question i mean yeah the reason we built ivy is as i mentioned at the beginning we don't want to add to the tool overwhelm by building another ai product I also fundamentally believe that a lot of AI companies are now just getting eaten by Anthropic because it is launching.

34:11Amelia Miller:It's just come for the design industry. Any AI design company or startup or even large company like Canva is quaking in their boots. And a lot of startups were destroyed overnight when they launched Claude Design. They're now coming for law. What's next? What's next? um what we're betting on is one thing that will never change which is change so as everything moves so quickly how can we keep up but also how can we train all of the learners so the teams we train the individuals we train to also keep up to a point that's not overwhelming though because you can't teach everything and you shouldn't teach everything because most of the news that comes out is just clickbait or it's just an AI influencer going yeah this is incredible this new feature and actually it's not incredible and it's actually irrelevant and it's not necessary for your job um the way we keep up with it is yeah we obviously have a lot of like internal agents running to flag what's important we have a lot of filters set up to reduce the noise so that we are only learning and then teaching what is important and then in terms of how we like consume that knowledge and teach we are not a company that does 50 hour courses on excel and then you know you have excel on your cv for 20 years those days are gone what we're moving to instead is a world where we need to micro upskill and we're very bullish on that term at ivy constant micro learning to stay on top and that doesn't mean you're learning a brand new tool every day but if you use Claude when they release a new feature you do need to learn it because it's going to be incredibly powerful for your role the half-life of a tech skill is compressing so quickly it's compressing about two and a half times year on year 70 % of the skills we use in our jobs at the moment will be obsolete by 2030 and that again all comes into this point of we can't be stagnant anymore and when individuals are thinking about learning ai it's not like you can just dedicate a week to learning it and wash your hands which a lot of people kind of get frustrated at because they're like damn i want to just learn ai and i've learned it unfortunately it doesn't work like that you've got to stay on top of it with micro upskilling so that your skills don't become stagnant because that's one of the worst things that you can do for yourself coming back to that point of being that you know being that top employee being that top hire you need to show you're constantly on the bleeding edge which can feel really overwhelming to some people and you can feel like god I've got to constantly learn or like oh god a new tools come out or a new features come out but that's why we built Ivy to make it accessible and to make it interesting to learn and We aim to cut the noise out so you only get taught what you need to know.

37:08Amelia Miller:And everything else is noise. Another thing that would be interesting to hear your thoughts on is, so if we think about cybersecurity, right? So saying I'm a cybersecurity business, I think when people hear that and see the branding of that, they kind of, that's quite boring. And they go like, I know that's kind of a boring business straight away. And saying you're AI enabled or you're a.ai business kind of gets people now jumping to conclusions and assuming things do you think if businesses are upskilling and becoming ai enabled they need to suddenly start telling people we're an ai enabled accountancy practice or do you think actually have that in the back end and let it kind of speak for itself i think the former massively because the best talent out there right now is ai fluent and companies want that talent for from an employer branding perspective I personally believe one of the best things you can do is showcase to the world that you are training your staff and you are keeping them up to date with AI because it means your staff are going to be the best version of themselves.

38:14Amelia Miller:And it means you're also able to attract the best talent on the market. I think if you keep it behind closed doors, you're doing yourself a disservice. I think if you're using AI to lay off half your workforce, it's probably not something you want to be shouting and screaming about. But most companies aren't really aren't doing that. And most companies that are leveraging AI and training their staff are doing it in a brilliant way. That means they have more capacity and gives their staff more time back to do deeply human work and add genuine value to the business. and that is something that should be proudly spoken about because it's going to help you you know as i said grow faster attract better better talent um and we're seeing that and a lot of businesses are completely pivoting their actual product to become an ai product let alone an ai enabled business they're actually changing their complete product offering an example multiverse you know the apprenticeships they've completely doubled down on actually it's ai apprenticeships only or read.com read.co.uk the job board they're now read.ai and it's ai jobs so this is you know it's happening in all of the biggest companies and so for any smes it's not something to hide or shy away from i really believe it's something to embrace because if you aren't becoming ai native you you won't survive my favorite quote is from late last year from the CEO of Microsoft, Sati Nadella, and he said, every company is now an AI company or it will die.

39:51Amelia Miller:It sounds quite doomsday, but it is true. Companies that are burying their heads in the sand and not leveraging these tools, training their teams to leverage these tools will not be around in five years because they'll be usurped by a company and a startup that is being started right now in someone's bedroom that's ai native from day one amelia you strike me as somebody that's like not afraid of change or risk like you're willing to dive into the sea of uncertainty i want to ask you about dragon's den and what was the decision making to to go on dragon's den oh i wish it was a decision but it was it was quite a funny story because we actually didn't have a business when lydia my sister and co-founder had applied to it when we were it was ivy was an idea in our brain um our first product that we launched was actually purely a recruitment play which is i guess why i'm so interested and have a fair bit of knowledge about the recruitment market but yeah we didn't have a product and we hadn't heard anything we'd had one call from their casting directors with tech and we were so early we had zero pounds revenue so we're like obviously they haven't you cast us because they love fmcg brands that actually have made money and then we this is in 2024 kind of late summer we went to glastonbury festival got back from glastonbury not feeling you know our most energetic i had no voice and i get this call from an unknown number and it's he's like hi there i'm executive producer at the bbc we've had a dropout can you be on a train in two hours to film an episode of dragon's den at 9 a.m tomorrow morning and i'm like yes we'll be there and i turned to lydia and i said like we need to create a fake version of the platform because we're gonna have to demo it on a screen on dragon's den tomorrow and she said what and what followed were yeah the the most chaotic kind of 36 hours where we yeah we get on this train we didn't even have time to go home so we had to like run to zara to buy an outfit um lydia's writing the pitch on the train i'm stupidly watching um a compilation on youtube of the worst roasts on dragon's den and then yeah and then and then you go in and you have one chance to do the pitch so if you mess the pitch up you know they air that you it really is one take and then you're followed by we were in there for about three hours of you know hardcore questioning of which they edit and cut down um it was and it was very it was risky and we remember being on the train going this really could go one of two ways it really could and thankfully it went the better of the two ways and we got yeah four offers and then got steven bartlett deborah meaden on board as investors wow i mean i think if me and zag me and zach had 36 hours i think we'd be being choked out by Peter Jones or something that is that's incredible I didn't I didn't know that part of the story and so you had two offers from Stephen Bartlett and Deborah Mead and for people at home they probably think oh right that's it you leave with a suitcase of cash and then you probably like go and meet up with them on a weekly basis and they're telling you what to do with your business what is the reality of life post dragons then with the investors 70 % of deals actually don't go through which some people don't know um ours did go through it definitely did that you don't leave with the cash you go through a formal you know due diligence process to make sure you've not you know lied when you're there because all it is is q a they they don't see any of your data they you know they take your word for it so we went through yeah due diligence process or you know we did the deal did go through um so they are you know on our cap table as investors which is great I was actually much more pleasantly surprised by how hands-on they've been as investors I think I went into it you know any sort of celebrity investor I personally wouldn't expect anything of them they have a lot of portfolio companies they are busy people they are running their own businesses um but they've been great Stephen in particular like giving us so much time I can't even tell you like so helpful especially because brand is so important to us and we do a lot of organic socials the expertise that his team have in that area is yeah completely like unmatched I don't think you can find better brains for like short form content so yeah I've been like very very happy with how hands-on they've been as investors and we're down the road from Stephen's office so we're in there like we're in there last week chatting to his team because they're about to launch a really interesting thing called culture test which um helps businesses like assess a cultural fit of their talent um and it's yeah super exciting so we're doing some brainstorming with them so yeah there's loads of different facets and i think one thing i've learned though is like if you don't ask you don't get like if you're silent you're um you know deborah meaden and stephen bartlett aren't going to come to us go oh hey haven't heard from you guys for a while like what can i do you need to be very upfront with exactly what you need from them and you need to make it a partnership worth investing in by not just like asking and asking for oh can you repost this or can you do that actually you you want it to be like a great investor portfolio company business relationship where they actually are excited to see you and you can actually give to give back to them and help them not that the trade-off is even don't get me wrong but you know you want to be as helpful as you can so you got four offers four offers okay so what made you think deborah and steven we went in wanting those two from the outset deborah because she's just an absolute titan um a legend and has a lot of good corporate contacts she's also massive on in the social impact space and something that's really important to us is making sure we are retaining a strong social mission which is that no one gets left behind in the wave of ai some listeners rightly so might be thinking ai can never be used in the same sentence as mission driven or ethical and i totally like do appreciate that but as i mentioned it's here to stay and if someone's going to train people i'd rather it be a company that's like passionate about it and doing it in a way that's accessible to everyone we know that ai is a completely male heavy industry less than 20 of the decision makers in the world's 100 largest ai companies are women and when they are they're usually in roles like chief people officer and most of the people building these models are not women most of the data they're getting trained on is not um is biased data great example here is uh llms are twice as likely to diagnose liver disease in men than women because most medical research has been done on male people like candidates male patients in the past and that's happening across the bat across all um like ai algorithms that are getting built another example is amazon built their own internal AI algorithm to screen CVs and they use their own hiring data to train this this internal algorithm but because Amazon is a very male heavy industry the AI it's not the AI is biased not the AI is like going oh I hate women but it's code and so it looks at that data and it goes oh we like to hire men so every CV that had the word women in it for example women's chess club or women's university it immediately rejected those candidates because of the data that it was trained on um i know i've massively gone on a tangent here but yeah very passionate about keeping that like responsible and ethical side whether it's in our training whether it's the way we advise our clients whether it's you know the strategy we do with c-suites um rather than yeah potentially a company that doesn't really care and is just money-grabbing don't get me wrong like we're for-profit business where we want to make money we want to grow but I think you you need to train AI to the masses in a way that is mission driven otherwise it's not gonna resonate hmm I was just thinking then when you say that that like a philosophical point but the truth is subjective isn't like everyone what's from one person thinks is true another person might not and often for people maybe listening if you try and have like a debate with ai that gets to the truth you will both probably fail to do it so there is no fun it's kind of what's true to you and i think sometimes you can have like a feeling when you've been on it for a while like you've been on instagram for too long like you're a bit kind of wired from being on it too long because you're trying to wrestle with it to get to like a truth or a really simple answer but to your point because it's trained on human behavior and human knowledge it's going to be just as flawed on finding the truth as we are exactly exactly and yeah a lot of people forget that ai is trained on like the internet and so it's only you know it's only got the information available to it it's not you know there's nothing new it has it is not unbiased so any bias that exists in the world is amplified by ai that's a fact that is a fact whether it's racial bias whether it's gender bias whatever it might be unfortunately ai amplifies it and that is the scariest part because then you get into these feedback loops with ai where for example we know that more men use ai on a day-to-day basis and therefore it's getting all of the conversations it's having are with well most of the conversations it's having are with like males in the population and that means the feedback it's getting from all of these chats it's having is from males and so the tweaks to the algorithm are going to be more male-centric than female-centric um and again that can just spiral on all of those tangents obviously i'm using gender as an example there but that is one of the scariest parts of ai because we're at a point now where the you know these foundational models are being built like now is the time and there's no women in the room or in the data that they're being trained on we have a lot of these conversations around ai and sometimes i leave them as a bit of a doomer we go through the cycles of do we feel positive about it or not i think this conversation has left me feeling optimistic but also like i want to get really sick of prompting now i feel like you were saying things and i'm thinking maybe i'm not doing that to the level that I could or maybe there's some other things that we can implement in the business so it's been so refreshing and we always ask Amelia if someone's listened to this and enjoyed it what's one question that you would leave with them to ask themselves oh wow that is a great question Barry 52 in Swansea is waiting for it I know Barry sorry I'm leaving you hanging Barry I would say a question to ask yourself is perhaps if anyone's feeling like hesitant to start or even you know anti-ai resistant of it is you know to imagine that state of your being where you're not doing can i swear on this podcast like you're not doing the shitty parts of your job or even like the manual parts of your life and you can have a system that runs for you what would you use that time back for um so if you're a business owner or if we take poor drinks with doing their customer service if you could have their time that time back what would you do would you take up pilates would you go for a you know go play football with your daughter in the park would you in that customer service case actually use that time to send flowers to the dogs when it whenever it's their birthday and i think thinking about that and thinking about like what you could do with the time you gain back from ai if you are using it to really help you is like it's just such an interesting and fun way to think about like how helpful these tools can be and to imagine that end state where we're all living in a world where ai is working with us and for us and automating stuff that we don't want to do no one wants to manually type stuff into excel so what could you do with that time that is incredibly human um so i'd maybe leave people to ponder that and you know if you could click your fingers and that manual side of your life was taken care of what would you do with that time amazing thank you so much if people want to work with you they want to find you how can they go about that you can go to our website where iv.jobs ivwe.jobs or you can email us hello at ivwe.jobs and yeah we'd love to help as i said we do training for individuals as well as teams so if your team has just bought a bunch of clawed licenses and you're like ah don't know what to do um get in contact with us because we do yeah we do bespoke team training uh we also have a bunch of self-serve and live cohorts on our website that just kick off at the same time each month so you could just jump onto one of those um if you don't want a bespoke training and then we also consult companies so if you're thinking about your ai strategy from you know a higher level we're wanting to get some of these automations that I've talked about in place then reach out to us and we can yeah we can come in and build them for you amazing thank you so so much thanks for having me it's great