Ep35 Making AI work in the real world

Artwork for podcast The Web Usability Podcast

Episode 35 • 12th November 2025 • The Web Usability Podcast • Lucy Collins from Web Usability

Ep35 Making AI work in the real world- Transcript

Lucy Collins

Welcome to the Web Usability podcast, where we explore what it takes to make the web a more accessible and enjoyable place for everyone. Whether you're a website owner, developer, or just a curious mind, we're here to share insights, tips, and stories that can help you improve the user experience of your digital world. I'm Lucy Collins, Director of Web Usability and your guide on this journey to better usability. If you need a transcript of this podcast, just visit our website at www.webusability.co.uk.

Now, let's dive in.

Lucy Collins

So today I am joined by Dr. Bruno Oliveira, Senior Lecturer in Strategy and Entrepreneurship, AI Solutions Expert and Director of the Masters in Entrepreneurship at the University of Bath. Bruno's mission is to transform the way that businesses operate and how students learn through AI, combining his role as an academic consultant and perpetual learner to create the perfect environment for this. His background spans founding startups, advising across public, private and social sectors and developing custom AI systems that make a tangible difference in the real world, some of which we'll hopefully discuss today. What makes Bruno's approach unique is his ability to bridge theory and practice, helping organisations and learners not just understand AI, but actually apply it in a meaningful way. Today, we're going to be exploring how to make AI less overwhelming, more practical and ultimately a little bit more human. And I'm going to be taking an awful lot of notes.

Bruno, hi, welcome to the podcast.

Dr. Bruno Oliveira

Hello, Lucy. It's a pleasure to be here. What a great introduction you made of myself. yeah, happy for the conversation. AI is such a great topic.

Lucy Collins

Amazing. Well, I mean, you have done so much. Obviously, I have spent a good amount of time stalking you across various social platforms. And there's so much in there, know, academic, but also founder, start. I just really don't know where you have the time to do it all. I feel it's hard enough running a business without actually also empowering the future generations to do these sorts of things. it is very impressive.

Dr. Bruno Oliveira

Thank you. Yeah, I think it's I think it's part of my role as well to as an academic that you test some of the things that I preach in the classroom and be very hands on. And I think that's one of the differences. I have many colleagues that are brilliant in the research and developing an advancing theory. But I focus a lot on this bridge between being an entrepreneur hands on practice, applying things, but also having the conceptual academic understanding and then trying to say, how can we apply this? How can we actually help business professionals and entrepreneurs use this in practice in a way that empowers them and it's not too jargon heavy and difficult.

Lucy Collins

Yeah, absolutely. And that's, you said that nicely about bridging that gap between theory and practice and the fact that you kind of do span that divide. So what does that mean in the context of AI and the conversation we're going to be having today?

Dr. Bruno Oliveira

There are a few different layers there. One of the layers is understanding, and I think that's one of the most important, what AI can do for you, because I think that's where you need to start. If you don't know even what AI can do for you, it's very difficult to know how you're going to use it. You need to know what it can do for you and then how to apply it in practice and make it actually have an impact in your workflow, in the activities that you are doing, in the value that you are creating. So that's the first understanding of what it can do and how can you use it to make it happen. I think the second layer is

If you understand some of the management concepts across marketing, across entrepreneurship, then you are able to also use that theoretical understanding to apply that when you use AI. Because AI is going to be much more powerful depending on your context and how you steer it, how you guide it. And that's something that we can discuss it today. think sometimes people are not fully aware of their role in really guiding AI because you can have a fantastic output from AI if you guide it in the right directions. But if you are very simplistic in the approach that you take with AI, your output can be very simple as well and not high quality and not something that you perhaps want to use for business or more kind of professional work.

Lucy Collins

Yeah, and I'm certainly guilty of that, of giving it a kind of very basic prompt and then being frustrated by the output it gives me because the output is basic. I mean, that is just me reflecting my own problem back at me, which I always find a bit amusing, but still I get frustrated with ChatGPT when it doesn't give me what I want. Can't you read my brain? So, I mean, that's, I think, takes me nice to my next question, which is why do you think so many people find AI abstract or intimidating or, I guess, in my case, frustrating as well? You know, why, why do we have this like emotional response to AI and particularly at the moment where it is everywhere and in every conversation that we're having?

Dr. Bruno Oliveira

I think one of the issues, it's very difficult to understand what AI does. It's almost like a mystical black box that you throw things at it and it comes out with answers for you and sometimes they're better, sometimes they're not so good. I think that's the first layer. Obviously, you said, it's being talked about and hyped about so much and sometimes if you don't understand the tools, the power of the tools, which tools to use for which tasks, how to leverage the tools in the right way, then you can get very frustrated very easily. I did an AI workshop at the University of Bath with some managers and executives a couple of weeks ago. And I realized that a lot of people were still using the free AI versions out there, free ChatGPT or free Claude or free Gemini. From a business perspective, that's quite complicated because unfortunately this company's AI is very expensive. The good AI models are very expensive to run. So when they give it for free, they want you to use it, but they won't give you the best models.

And I think that's where the understanding doesn't exist at the moment. If you use free CHPT, most of the time they are going to give you a small model or an instant model that is very, very underpowered. So depending on what you want to do, if you want maybe some cooking recipes, it might help you. But if you want to deepen ours, this deep strategic understanding, writing skills, if you want high quality research with accurate sources, with accurate information, then you need the best models. And usually the best models are behind the paywall. And I think that creates a lot of frustration because not all AI is made the same. And definitely the free versions are usually not made the same. And that's the first point of, think, that we need to highlight to everyone that it's important to know which models you use for which tasks.

Lucy Collins

And I think that probably feeds into the overwhelm and this slightly intimidating feeling that I feel about AI, is, but which tool do I use? And you named already a handful of things there. And that is but the tip of the iceberg, I mean, we're talking about, I guess, generative AI chat tools primarily here, but then there's all so many AI enhanced tools of every, you know, every walk of life that it's just, it's just completely overwhelming. Where do I start?

Dr. Bruno Oliveira

Yeah, and I absolutely understand it. As someone that loves AI and follows AI closely, it's still difficult to keep breadth of all the changes that are happening because every week you have new tools, new features, even if you just want to focus on the main models. And let's talk about the main models, maybe Entropic Cloud, OpenAI, Chat GPT, Google Gemini and XAI Grok. Perhaps these are the four top of the line Western AI models from the big labs out there. Even then they are releasing features and new models are almost like weekly, monthly, definitely things are happening might be different feature, not a different model or a new model. So it's quite difficult. And then, as you said, there is apps that can do everything for you. I think the best way to start is, okay, start with these general models and start to see how you can start to leverage some of your repetitive workflow tasks with these models first. I think there are a lot of quick wins out there for business professionals and businesses that do not require extensive learning and cost. And I think that's where you should start.

Dr. Bruno Oliveira

Yeah, actually, it probably takes us quite nicely because one of the things that you do is support businesses in this AI readiness piece. So what is it that you look for? What are the key indicators that an organisation is ready or perhaps not yet ready? Although I struggle to understand how anyone could not be ready to integrate AI into their ways of working. Yeah, that's a fantastic point. This is not obviously a black or white assessment, but obviously the first point is I believe everyone is ready for AI. The question is, to what extent?

to what level of integration in their workflow, in their business processes, in their existing software stack, they are ready for it. And this is the kind of assessment that I can help to do. Okay, where are you at at the moment? Do you have almost no AI or manual use of AI here and there occasionally? And the first thing, okay, let's get these quick wins out of the door because there is a lot to be gained in quick wins. And quick wins might simply be you start drafting 90 % of your emails with the help of an AI that knows your instructions, knows what it needs to do have your context, your files, your company files, your previous answers or whatever. From my experience, just email alone, you can probably save 10 times the time and have five times more quality. So it's not only a time saving, but it's also the quality of your emails is going to go much higher. You can have a meeting summariser to summarise all the meetings and to have actionable points already done for you. Nobody needs to do that after the meeting. The AI can do that in the background, and you can automate those processes. And I think that's where the power is. You can have content creation, blog posts, media posts. You can have research because you need research if you are creating content. You can have an AI that will going to give you research on topics and ideas for those posts. So there's a lot of things that you can start, I wouldn't say fully automating and the initial stages may be something called SME automation where you create a few AIs. Normally they call it projects. If you go into CHATGPT, if you go into Claude, if you go into Grok. They have projects and you can really nicely set a few projects for the different tasks that you do repetitively on a weekly basis. And just this alone, I think it's going to already significantly transform your productivity. And that's the starting point. And then obviously we can go higher up to more automation. Like you hand on a lot of it to AI automatically.

Lucy Collins

Yeah. So, you sort of almost suggesting that you look at the different bits of your business and kind of break them down, it sounds like into individual tasks almost, and then that forms kind of the basis of where you might look at applying AI. Would you say that's the kind of starting point? how does someone get into this if they're just going, my God, it's everywhere and I don't know what to do with it?

Dr. Bruno Oliveira

Yeah, for sure. think that's the first step is, okay, what do I do repetitively, daily and weekly that perhaps I could do better, I could do faster with the help of an AI?

I think that's where to start, of a review of your daily workflow and what you can do to kind of smear automate at least those tasks. And I think that's already going to be a huge help. It can be, for example, even data analysis that you want to have your some metrics, financial metrics you have for your business, some goals that you want to achieve. You have an e-commerce store online and you want every week to review your metrics, your sales, your average or the value, your returns, cart abandonment, all those things. And you can have an AI that does that for you automatically every week. And that lets you look, there is a problem here, this metric this week or this month. Those things can be set up, SME automatically, but obviously we can set up automatically to happen on a schedule every week or every month. On my AI workshop, I demonstrated the kind of an advanced dashboard that I created without any coding because I'm not a coder.

Just prompting an AI and it put all the metrics that are important for the business. I asked it to create a traffic light system. So if a metric is within our target goals, make it green. If a metric is not within our goals, make it red. If it's like borderline, make it yellow. And the AI was even summarising the key points of the analysis. Okay, this is the strength, this is the weaknesses, this is the things that you need to do or you need to look at more in detail.

These things are not possible. Dashboards like this will take weeks or months or require expensive software that usually was not flexible to your needs or goals because you couldn't adjust everything. And now you can adjust it to your needs by yourself. And you can do it's me automated where you drop the file with the data to the AI and say, okay, now update this dashboard. Or there are ways that are takes a little bit more effort, but you can automate the process. And basically then you don't have to do anything and then you can just go there very week or month.

Lucy Collins

I mean, yeah, that sounds remarkable. And that's just using Claude or...

Dr. Bruno Oliveira

Yes, for this particular one at the moment, I think in terms of coding, in terms of creating, this is called like an artifact that you can then interact with. It's fully interactive. You can go across the graph, it shows you the data points. If you have sales month by month, you can just go through the graph and it's going to show you it's fully interactive.

And it's Claude, think it's the easiest one at the moment to create these kind of tools quickly, just with prompting. you just talk like you're talking with a colleague or with an employee or you tell him, please do this for me. And then there are errors. say, look, there is an error here. Can you sort this? You don't like the design? How can you change the colours? Can you make the colours? So, you basically on a conversational basis, you have the power to do it. Obviously, depending on how advanced it is, what you want to do, it can get a little bit more expensive. Claude is a little bit expensive on that sense compared to other options out there because coding requires a lot of work on the background. So you might go through your pro plan very quickly. So that's why they then have max plan, which is 75 pounds or 150 pounds. it's not everyone will need that. it's interesting.

Lucy Collins

There's kind of two things that you've said that have struck me, which is the first one is start with the workflow, start with the problem, start with those repetitive tasks. And I said this before we started recording, but it has felt like a lot of the conversation has been led by the other end for a long time, which is use AI, use AI, use AI. But it's a guess, but how? And what's nice about what you're saying is, we'll start with those repetitive tasks. And yes, AI will be part of the solution, a probably big part, if not the whole part of the solution. But it's still starting with the problem tasks, those repetitive tasks where you can save time and free yourself up presumably for the more creative or interesting work or whatever else it might be. And then the second thing that you've said a couple of times is that it might be semi-automated, which makes me think of this kind of human in the loop phrase that I've heard quite a bit when we talk about AI. it's something we got challenged on a little bit recently about, okay, well, which of these jobs would you be happy to completely outsource to an AI versus which are ones that you as a human still want to have sight of?

Like one of the examples was if somebody contacts my business, who are they dealing with? Is it an AI or is it a human? And for me, it's always a human at the moment. That might change. But for me, my values are that they engage with a human. when it comes to some of those more automated tasks, helping me write some of my social media content, dare I say it? Yes, I am going to use AI, not to completely obviously do the work for me, but to speed up the process. And I think those two things, starting with the problem and knowing where the human is non-negotiable, feel like two really important kind of lines in the sand to have when it's starting down this AI integration path, whatever you want to call it.

Dr. Bruno Oliveira

Yeah, yeah, for sure. I think that what you can do is automate a lot of the processes that you focus on the higher level kind of architecture, vision, strategy, guiding. I think they still need some guiding to get to the most high quality output but you can simplify your work very much. you are very right that part of the overwhelm is that sometimes when I talk with people, think, oh, but what software do I need? Because they think they have to go out and acquire a lot of very expensive software and change a lot of their existing processes and their software stack, which is sometimes for some companies quite challenging. But you can start with a subscription on a pro model like ChatGPT for £20, so around 15 pounds, for Claude, 15 pounds. And that alone, if you start with one and you start setting up your automated projects for the tasks that are repetitive, that alone already is super helpful. As you feel more comfortable and you understand the power of AI, then you can go to the next level a little bit step by step and do more things. But starting there, doesn't need much cost. I do recommend a pro subscription because I think that the level of models you're going to get is going to be far superior. But apart from that, you don't need much more to start with. And now they're building all sorts of integration that you can integrate. For example, in Cloud, you can integrate with your Google Drive, with your Dropbox. they can do files now. They can create Word documents for you. And it's a lot of things that are emerging all the time. But you need to start somewhere. And I think the easiest thing to do, and I think it's perhaps one of your questions later on, is use projects. think that's the easiest thing you can do. Define different projects, set a role. Because when you define a project on an AI model, you have an opportunity to say instructions. So you could set a role for this project. This is what I expect you to do. This is your role and this is what you need to do. And then you can add files that give context to that role. Because I think one of the issues that I know we're experiencing, and this is something I've heard from other business owners is, how do you ensure that AI is being used consistently across the business as well?

Lucy Collins

You know, I feel like we're all kind of going up this learning curve and obviously staff are doing it at different rates. Some people are more open adopters. Some people are a little bit more rejectors of it. And it's about bringing everybody along this journey in a consistent and to a certain extent standardized way so that you as an organization are using things in the right way. Does the kind of project-ization of content allow you to do that more consistently as well?

Dr. Bruno Oliveira

Yes, yes, I think that's obvious depends on the side of the organization, but usually it might help for them to have a kind of a training workshop that everyone is there. So you have half a day, I think is enough to kind of set up the basic parameters of how to use AI. One of the easiest ways is to get business accounts because business accounts can share projects, they can share elements together. So you can go to GPT or Claude business account for their team. don't know, 20, you have a 20 team business, you can have 20 accounts that share. And you can create projects that are shared. So you create it once and everyone can use the same project and everything is there in the project, all the different shots, the information. And then it means you can put your knowledge base. imagine you are creating a project just for Instagram post writing. You can give the instructions, and you can use an AI to help you to provide the instructions if you don't want to write everything from scratch. And then I think what is powerful is you can go to check your, maybe your 20 or 50 best Instagram posts ever from your analytics. You extract those posts, you put on the knowledge base, you put the analytics and the metrics, the likes, the comments of the different posts, and the AI was going to see it. This is for this brand. This is the posts that have worked in the past, and it can mimic your language, it can mimic your to keep the brand consistency. So this is adding context to the AI. And this is this is quite powerful. Obviously, depending on how advanced the companies are, there are companies now that they develop their own internal knowledge base and they have their own local models that have access to all the company documents, HR policies, technical guides. So if you have a question, instead of having to bother someone sending an email, I don't remember how to do this, can you help me? You just ask the AI and the AI responds straight away.

So that different layers of integration to get the team together, but it will be good to have the team to have an understanding of using the basic tools, the projects, sharing those elements. There are a lot of companies that use Slack and Slack is adding more and more integrations. You can have, for example, Claude as a member of Slack and Claude can pitch in with ideas if the team is feeling stuck. He can summarise conversations. If you ask, Claude, can you summarise what people would do?

what was the decisions or action points from a meeting that's what happened two weeks ago or a project or where are we with this project? And it can summarize everything in the chat. Those things can be very powerful to get the team working together and using AI.

Lucy Collins

Yeah, fascinating. mean, goodness, there's so many things I'm thinking, we should probably be doing that in our business. I'm definitely going to be taking some of this back to the team. You said, I mean that It was really nice hearing that example, that Slack example there, you obviously shared the dashboard that you created as part of your training. Have you got any other examples of kind of businesses that have successfully actually implemented, you know, an interesting AI workflow or something that might be useful or interesting to share with the listeners?

Dr. Bruno Oliveira

Yeah, I work a lot with e-commerce. One business was a jewellery artisan selling jewellery via e-commerce. As a very small team, they had less than five people lot of it is content creation because then they need to get some organic traffic so that they can keep selling their products. Emails, customer support, and they were quite overwhelmed because they were using AI but very rudimentary. They were going to chat to GPT asking, okay, can you do this for me? And every time they had to explain the task very repeatedly. That can help a little bit, but after a while you get frustrated because every time you want it, you need to explain, okay, this is the situation, can you do it? And it might not have context of previous situations that happened. So we tried to start with the quick wins. So reviewed everything that they did and started to create those projects. So we created a project just for customer support. We put all the knowledge base, frequently asked questions. We got access to their last six months of customer support. We put all that data into the context. So the AI already knew everything that has been said before, questions that have been asked and how it could be answered. So every time you had an issue that relates to customer support, you go to that project, you paste it there and the AI already has a lot of context and you might not even have to intervene and it's going to give you a perfect answer for you to send straight away. Just that alone is a step forward. So we set that up. We set a marketing strategist project that would help them to get ideas for omni-channel marketing, topics that you can do for the different channels. And also we set up a research assistant. So there is a project that you can tell the AI, for example, every week, do research on the trends in fashion, in jewellery, what new ideas we can potentially use to connect with our customers. So, all those small little things that started to help. And then an Instagram post project a blog post project and we can start to connect those things. So every time you need something, you go to the marketing strategist and the research assistant, you put that together and you already go to the Instagram post creator and blog post creator with the topic and the guidelines that the other AI gave you to do it. This is like what I call SME automation because it still requires you to copy and paste and go around and do quite a few things, but it's quite helpful. Obviously the next stage that is possible in e-commerce you can have a chatbot on your website that has access to all the previous customer service responses, frequently asked questions, your instructions, your product base, your orders, and can reply in a very fluent way to customer queries. In this case, when we released this, they had a lot of customers in the US. So one of the biggest problems was that overnight the team is sleeping. So when US customers are coming for a chat, nobody replies sometimes the customer only needs to know that you are there to reply and then they feel, okay, this company is responsive, I can order with confidence because it's jewellery. And just by doing that, having an AI that knows everything and can reply to most customer pre-sales questions straight away during the night, increase the orders in 25 % from you as customers because they felt more confident that someone is replying to my queries so I can go on and make the purchase.

I think it can be very powerful. This is the second layer of automation that you're already giving the AI a lot of power to respond directly to customers. You can put some guardrails that you say, okay, if you feel that this is too complex, do not answer and say, I'll have to pass this to a human later on. So please, I'm very sorry, but I'll have to come to you later. So, it's possible that not all answers will be dealt with, but you can have 75, 80 % of the most frequently asked questions addressed and increase customer confidence in your brand. And that can be very powerful. I think that's a second layer of automation that I would, which it's more automated than projects features that we were talking about.

Lucy Collins

Yeah, but it still requires you to kind of start with that underlying foundation of, well, what are the jobs that you have in your business, having it really well structured, having the process clear, like, actually, it's really interesting, because I feel for a lot of people, actually, this sort of AI work might help them crystallise, like, well, what are the different areas of your business? And where are those repetitive tasks? The stuff that you kind of just do a bit mindlessly, and you don't really realising that you're doing it five times a day, suddenly, it gives you an opportunity to go, hang on, let's reflect on that. And maybe we can improve it. So I just think it's, it's a really nice way of starting. even if it is literally just breaking up those job roles into projects and then it's off the back of that going, okay, well, how could we actually then move from semi automation to more full automation rather than thinking I must find a tool that's going to do something to help my business? Where do I start?

Dr. Bruno Oliveira

Yeah, yeah, I think you're very right in emphasising that, but it always goes back to to understanding what you do in your workflow, what is repetitive. And the beautiful thing is that when you start to get a bit more familiar, with some of the tasks, with what AI can do for you, you start having more ideas and seeing more opportunities of, okay, what else that I am actually not doing at all because I don't have enough time. But if I have an AI that is doing it for me, that would be nice to have. And it would increase the value that we will create, it will increase the insights that we will generate but not increase our team's time because we don't have team's time. Particularly for small companies, it's very likely that many times there are a lot of tasks that are that could be done, but they are not done because you don't have the time and you have to focus on other things. So you could, for example, have the AI every week analyze all the customer support chats and emails and tell you here are the insights, here are the main bottlenecks. Maybe if you sort these, you might increase your sales or you might give your customers a better experience. So you can create this automated every week or every month, which you probably wouldn't have time in a small team to be analysing tons and tons of chats data and understanding the patterns because it takes quite a lot. An AI can do that in less than a minute or maybe if you use a very good thinking model in a few minutes, it can see the patterns, it can tell you here are the problems, here are the issues. So you can start doing things that you never did before because you didn't have the resources to do so. that's the second layer. First layer, what is repetitive? How can we automate that? Or how can you use AI to improve our efficiency and increase our quality? The other thing is what are we not doing at all because we don't have the resources to do it, but with an AI doing it for us, we can actually do it and increase the value that we are creating as a company and the efficiency that we have as a company.

Lucy Collins

Yeah, love that. So this is possibly a slightly selfish question on my part, but at Web Usability, obviously we're very focused on digital inclusion and making sure that technology serves all users and we're not leaving anyone behind in this. So how do you see AI supporting inclusivity and accessibility online?

Dr. Bruno Oliveira

Because of its multimodality, I think it can be very powerful because It can work for multiple people in multiple ways. If you need to engage with an AI via voice, via audio, you can do that. You don't need text, you don't need images. And that's really powerful, not only if you have a health disorder that requires you to do so, but also if you, as a learner, you are an auditive learner and you learn the best and you develop your ideas the best by talking with someone or by listening to something.

With an AI you can do both, you can actually tell the AI to tell you stories, to tell you about some concepts, to tell you whatever, or if you actually have a better because you interact with someone, you ask questions and you answer and this helps your thought process, you can do that with an AI these days. They are very good to have conversations if that's your preferred modality. If your preferred modality is vision, you can engage with images, you can engage with text, you can engage with video.

So I think from an inclusivity perspective is extremely powerful because of this multimodality. And obviously, even more if we talk about different cultural barriers, language barriers, very soon it will be very flawless to have a conversation from different languages and it's automatically translated for both people at the same time so they can actually have a conversation. So that's not a barrier anymore. If you're talking with someone in China and they don't speak English, you can speak English, they can speak Chinese everything is going to happen that will create a lot of inclusivity as well from from a cultural and language perspective. So I think that the opportunities are huge at that level.

Lucy Collins

I agree. And we've had some fascinating guests on this podcast previously, people with lived experience of disability who shared some of the AI tools that they're using and you know, how actually how empowering a lot of those have been and for a number of the reasons that you already mentioned. I have great hopes for the future. that actually, digital accessibility, we're going to be slowly breaking down a lot of the barriers that people experience in the kind of traditional online world, thanks to AI. it's just making sure that, particularly data sets and things like that are well-representing those groups of users who presumably are a little less well-represented in the standard sets. But even so, as you say, think the different modalities, the different ways of interacting with it does open up doors, hopefully, for a lot of people. Are there any particular technologies or trends that you are excited about right now? there anything that's kind of giving you a bit of a buzz about AI as someone who's obviously fairly deep into this world? What's exciting you at the moment?

Dr. Bruno Oliveira

Yeah, I think the beautiful thing about the AI is that I think we're just getting started. I think that there will be so much more improvements and new features that are coming. The biggest transformation that is upcoming that we already see some signs is the transition from AI is an instruction based AI that you tell him, do this task, write me this email, summarize this meeting, create this report to an AI that is autonomous as what they call AI agents. That's it's a buzzword that has been around for a while. They are still not very good, but I'm sure we will get there that instead of saying, do this task for me, you say, please achieve this goal. And they will interpret the goal, analyze everything that needs to be done. So for example, you are having a problem in your website that you have a too high return rate or a request for refund. You can tell, can you please sort this out? I want to decrease this in 5 % or I want to decrease this in 10 % and the AI will start to have the ability to understand first how to diagnose the problem and then what it can do to solve the problem. And you might do very little intervention throughout this process.

There are a few glimpse of this already happening. The deep research feature that exists nowadays with OpenAI, with Gemini, it's kind of early stages of AI agentic workflow, which basically you give it a task, please research this topic or research my competitors or research the trends in AI and it's going to go out there, going to visit a hundred or two hundred websites or social media and it's going to summarise everything for you and give you actionable points. That's kind of early stages of AI. What we saw last month or a month and a half ago, Claude released Claude for Chrome. So it basically can take over your browser and can do things for you in your browser. That can be very powerful. If you have, for example, data that you need to copy paste from a spreadsheet somewhere or from a website into a spreadsheet. You can go before you go to sleep, can say, please do this. And in the morning you wake up, it already have done it. It's not perfect at this moment, but it's already being able to do some of these tasks for you. But I think this is the direction we are going much more into the future where they will be able to do those things for you. Obviously in coding, this is already happening that they deploy AI agents, they are building an app, they tell what features they want to see in the app. They can define the strategy and the plan and they can leave an AI.agent working for seven hours autonomously creating the code, building the app, and then you come back later, it could be very close to being done. So these things are becoming more and more possible. And I think this is quite exciting to see a future where AI agents can do stuff for you, which means as a small business, you could have a very small team that adds high value human understanding, but then you have AI agents to do a lot of the repetitive tasks that usually were quite challenging.

Lucy Collins

And there will still be a role for humans in all of this?

Dr. Bruno Oliveira

It's a good question. I do think so. do think that humans can still bring value as architects and visionaries of the system. It's very difficult to predict in 10 years. But the other day I was actually listening to the CEO of Box.com and he's one of the top companies in Silicon Valley providing enterprise cloud services. And he was very optimistic. He was saying, that he actually thinks that maybe big companies like Google, Amazon, they won't need as many people because AI can do a lot of those jobs. for smaller companies, what might happen is you will have the opportunity to do tasks that you could not do before because you just simply didn't have the time or the resources. And now, because AI is taking a lot of the repetitive work, then you will be able to create much more value because you will have more time to do so. So he actually thinks that there will be quite a significant role for humans to play.

It's difficult to say. I don't want to make predictions, obviously. They talk about AGI. And if you have a super smart AI that can do a lot of things for you, then I don't know, we'll have to think what is our role as humans in the process. I believe we'll find a way through the process. It might not be smooth all the way, but usually we have been able to find solutions to new advances in technology.

Lucy Collins

Yes, we are fairly good at evolving to changing situations. I mean, I'm all up for something that does my job and allows me to hit the beach and drink a pina colada. So here's to that sort of future. So a final wrap up question, if listeners were to take one action from this episode to make AI more practical in their own work, what should it be?

Dr. Bruno Oliveira

Go out and make some projects in your AI of choice, because I think this can really separate your work very significantly without you having to every time explain to the AI what needs to be done. think that's the main thing I would say something I don't think I mentioned throughout is I'm a big fan of voice input because you can usually speak much faster than you can write. And I'm a very fast writer. I grew up with computers, but I can still speak much faster. So I really enjoy sometimes even to draft an email to say, please reply to these emails, say this, this and this. And it already has previous context as well. So in 30 seconds, I think take what I want the answer to be or steer it in the right direction and it's done. And even

being a fast writer for emails, it probably would take five, 10 minutes for a long email and now it might take two or three minutes. So those things can really make a difference in the workflow. So I think that would be my main advice, but I think we should always go back to your point that you emphasize that start with your workflow, what you do repetitively, what are the tasks that you're doing, do an analysis of your workflow as a starting point to start automating, then choose one thing.

and try to create a project for that one thing and go from that. You probably get excited with the results at some point and start doing more and more gradually.

Lucy Collins

Yeah, brilliant. Bruno, thank you so much for joining me today. So much good stuff in there. I wish I had a pad of paper in front of me. We should have just used, in fact, you know what, I'll just use an AI summary of this podcast recording to take my actions forward. But thank you so much. It's really great to, as you say, of bridge that gap between theory and practice and provide some really, really practical tips and advice there. So thank you. Do you want to finish off by telling people where they can find out a little bit more about you and your work?

Dr. Bruno Oliveira

Yeah, fantastic. It was a pleasure to be here. First of all, I have been doing a lot of work on gusto-mind.ai. So that's where all my work about AI is being concentrated. And yeah, the mission is to try to simplify things for business professionals and to basically take advantage of AI as separated because prompts can also be very powerful, give people ideas of what you can do with AI, how you can automate your workflow and how you can use the best prompts for your project.

Lucy Collins

brilliant. And we will include all those links in the show notes. So thank you again and thank you for listening. What a pleasure. Thank you. Thank you for tuning in to the Web Usability podcast. We really hope you enjoyed this episode. If you have any questions, comments or topics you'd like us to cover, reach out to me on lucy at web usability dot co dot uk or connect with us on LinkedIn. Please don't forget to like and subscribe so you never miss an episode. Until next time, keep making the web a better place one user at a time.

Share Episode

Shownotes

Summary

Sound Bites

Chapters

Transcripts

Follow

Links

Chapters

Video

More from YouTube