Staff Are Too Scared To Use AI, The Questios Designers Should Be Asking, and A Human Approach To Agents.

Staff Too Scared of the AI Axe to Pick It Up — The Register / Forrester

• Forrester's AIQ metric — a measure of individual and organisational readiness for AI — shows adoption is lagging badly, and the reasons are telling
• Two culprits: employees aren't trained well enough, and there's an ambient anxiety about job loss that turns people away from the tools altogether
• My take: anxiety is lack of clarity — people fear AI substitution because they haven't mapped what they actually do every day, let alone identified which parts AI could touch
• The exercise I'd recommend before any AI training: write out your full task pipeline as if you were handing it to an intern — inputs, outputs, sub-tasks, decision points, all of it
• Then ask three questions for each task: is it repetitive? Is it unfulfilling? Can AI do it well? Only when you get three yeses should you consider delegating it
• Most people will find AI touches maybe 5–10% of their work — and that realisation alone does more to reduce fear than any company-wide AI rollout

The Ground Is Shaking — Why Designers Must Flip The Script on AI — UX Collective

• Peter's article is one of the best things I've read on this topic — he frames the core question not as "what can AI do?" but "why are we doing this in the first place?"
• The concept at the centre: Vygotsky's "more knowledgeable other" — the figure who can see both where a learner is and where they need to get to, and who scaffolds the gap
• Silicon Valley's message to designers right now is: AI is your MKO — let it guide you
• Peter's argument, and mine: it should be the other way around — we are the masters of purpose, goal, and constraint — AI is the skilled executor, not the director
• Language is our current interface with machines, but not everything we conceptualise is linguistic — spatial thinking, embodied experience, tacit knowledge — AI can have theoretical knowledge about gravity, but it will never feel it
• The choice isn't whether to use AI — that's settled — it's whether you define the parameters or just accept the outputs — whether you build the floor or keep asking why the ground is shaking

A Human Approach to Agentic AI — UX Collective

• Christine's experiment: using a multi-agent AI system to write a book — editor in chief, sales and growth, voice, product, reader advocate — all as sub-agents receiving context and iterating
• I find this genuinely fascinating as an experiment in approximating human team work with AI
• But I'd push back on one thing: at what point does the context engineering required to replicate a human editor in chief become so large that you'd have been better off with an actual person using AI?
• There's an asymptotic relationship here — the more you try to replicate what a human does, the more documentation you have to keep feeding the model as the work grows
• My real question: how does the output compare to a human collaborator who is also using AI? That comparison is the one worth running

Transcript

SPEAKER_00 0:00

In today's episode, why are teams not using AI so much? What should UX designers or product designers do to adapt to this AI era? And finally, how can we build a human approach to agent tick AI?

SPEAKER_01 0:21

One, two, three, four, and I'm gonna get a lot of people.

SPEAKER_00 1:02

Should we? What are the implications and why are we not using AI so much for now? Okay, so I'm gonna cover three articles. The first one is from the register. The headline says staff too scared of the AI axe to pick it up for restored fines. So this is the main idea. I'm gonna probably butcher it, but I'm just gonna try to summarize what I read. So the article says that we are not using AI as much as what was expected in companies, apparently. And so there are there could be two reasons at least. The first one being that employees fear AI-driven job losses, and the second one being that they are we are not trained, well trained enough to use the tech. That is interesting. It claims that, so this is from the article, quote, it claims that among decision makers, 81% say AI co-pilots are important for assisting employees. Workers must therefore adapt it. It declares. The proportion of firms that say they offer internal AI training to non-technical employees grew slightly last year. They mention, of course, prompt engineering, that it's a key skill. You will see that it's kind of linked to the second article that I'm about to read, which is uh entitled The Ground is Shaking, Why Designers Must Flip the Script on AI. But this is the same idea that we need more training and we need more thinking around how to use AI. And then, of course, there is the idea that there is an ambivalent, there is an ambient, sorry, this is coming from the article, there is an ambient environment of anxiety and mistrust that hinders progress. So, quote, some of our employees fear job loss and it turns them away from AI altogether, end quote. Okay, and the register says that the solution is for organizations to frame workforce AI as an opportunity builder for employees and articulate the benefits from an employee perspective. Okay, so that is more or less what I could read on this article, which is very short. I really encourage you to read it. And um I'm kind of new to the Forrester, so I'm I'm I'm discovering what is being written there. Um apparently there is a comment section, I just discovered that. And um that is really really interesting to see the conversation going. I would love to read more about that. But in the meantime, let me share with you two sense about this thought. So, two things. People are not trained enough to use AI, and people fear about their job loss. That is really, really interesting, of course. And there is in the middle this idea of AI quotient, which is apparently the notion of the readiness of individuals for AI. And so this is really interesting because it kind of sorry, it kind of uh mixes the idea of readiness with which is kind of an average measure between probably this is probably a proxy between training and willingness to adopt. And so, like any technology, that is the proof that you can put an amazing technology in the hand of people, ultimately they will need to adapt. So we need to be trained on how to use it. And so this has a lot of implications, and that's why I love user experience so much, because user experience is everywhere, even on user experience professionals themselves or knowledge workers themselves. When we do our job, which is to conceive new experiences, we need to think about the end needs of our users, but it also is the case for ourselves if we are to use new technologies which make us supposedly more efficient. Well, this has a lot of implications. I'm not sure if you read the book. If you haven't, I highly encourage you to read it, which is from Miss Meadows. I don't know if I'm pronouncing the name correctly, the surname correctly. Donne sec. Yeah, I think she's called Yeah, Donnella Meadows, and the book is the title is Thinking in Systems. So by Donnella Meadows. So if you haven't read it, that's fascinating. I highly encourage you to do it. Um, it it really shares with you the idea that everything in this world is a system, everything impacts everything else. Like this is a main idea, and I'm butchering the idea, but you get you get the point. So, knowledge workers, the way we work, we we manage information in and out, we process it in the middle. So I like to see that as a pipeline. We have an input, then we process something, and then we give an output. But this is a system, and this information is is arrives from somewhere else and then goes to somewhere else. All of that is a system. If you place AI in the middle, which is which is thought to accelerate the process, let's say, or make you more efficient, this will have implications. This will have implications. Like I can think of thousands of applications like right now. The first one being if you're more efficient as a designer to produce a prototype, how does that impact product management? If you can make three prototypes instead of one, how does that impact? If product management is more efficient at making their decisions, how does that impact user experience design? How does that impact research? We need to think about all that. And if you can generate like tens of ideas like really quickly, what does that say about the value of an idea? This is linked to the second article that I'm about to try to summarize. But it's really the idea that, of course, this is significant, this impact that AI is having on our lives. We cannot we cannot ignore that. This is a fact. Then how we should use it, when we should use it, and why, this is another question. But it's here to say, this is a new technology and it's working. It's working to some extent, right? So we need to adapt to it. Again, I talked about this yesterday in yesterday's episode. The idea that if your neighbor, and by neighbor here, we can think about competitor, if your neighbor is using it, then ultimately you have no choice but use it as well to stay in the course. That that's how companies will think about this, in my opinion. Except if probably we see that ultimately um it there is a net negative for everyone, which I don't believe right now is the case, probably it's more neutral. That's a good question to to ask. And probably the the impact is much more how can I say, mitigated positively and negatively because of the all the ways you can use it, all the layers that it can go right or wrong. So, what I believe is a huge implication from this article is that people are lacking training about AI. What does that reveal? Okay, what does AI do? AI helps you to be more efficient, produce more quickly, summarize text really easily, produce more prototypes, and so on and so forth. But what does that imply? I have one element here today that I want to share that is one way you can start to use AI better today and tomorrow. It's not everything, but it's one way you can improve on your use of AI and your adoption. Okay, we know that AI accelerates a lot of the process. So it takes an input and it accelerates the output, let's say, and the input being a prompt, being a direction. And if you listen to other episodes that I that I shared, I really focused on this idea that if you take only one prompt and you ask it to an AI or an LLM, it will give you an average output. Which can be good in some occasions, but more often than not, it will not be good because it's it's not it's not custom to your situation. So ultimately, one great skill to have, not maybe maybe not skill, but at least one process to adopt is context engineering. So it's providing the AI as much context and knowledge as possible so that it replicates your way of talking, your way of doing. Okay, that being said, it requires you to think in system. It requires us to think to think in systems and not doing before thinking. So that's that's what is really interesting about AI, it's because it pushes our thinking. Sorry, it it it challenges our way of thinking. If you if you mix thinking and doing, and if AI does the doing, then you will feel like AI is doing your job. And I think this is what is kind of said in between the lines of this article, which is very short, but I think there is much more to say about this. So, and this is also the idea of the second article, which goes way, way, way more in detail, and I love it for that. So, anyways, let's think about AI right now, and I'm I will try to be as concise as possible, and it's not easy because it's it's so philosophical at some point. I like to see AI as right now as a really really skilled intern. Why skilled? Because it's trained on I don't know how how much data, but it's trained on so so much data from all around the world and from from so uh let's let's say so so so much data through time, also. Okay and so it's very skilled. And it's an intern because it doesn't imagine you are your company of one, and this intern joins, and it doesn't know how you work, it doesn't know your preferences, it doesn't know the way you like to speak, and so on and so forth, all of these ideas, right? It doesn't know all of that, so you need to train it, you need to show it the way. This is again the idea in the second article. So for that to happen, you need to be a master at knowing what you want. And you need to matter to be a master at knowing about your problem area and the way more or less that you want to that you want to approach a given problem. So you need to know about the context, you need to know about your tasks. One way of being clear at asking what you want from AI is being able to being able to draft the journey that you go through when you do your job step by step. And identify the tasks during which you would like to delegate that to someone else. I'm not even speaking about AI, I'm speaking about someone else. And then you only then ask the question. In these tasks, if it were to be done by an AI, would I be okay with that? So it doesn't need to be necessarily AI right from the get-go. It could be someone else, including or not including AI. But ultimately, you need to do this work. You need to be able to have a mapping of all the tasks that you do, and for each task, what are all the subtasks and what are the expected outcomes? What does success mean? What does failure mean? And yeah, basically, where to pick information, where to output new data, and so on and so forth. And that is, in my opinion, not only this not only related to AI, it's really related about your job, and then what is about what you are doing that is uniquely yours, and that you need and you will find such things. The problem is people don't have clarity on what AI can do for them and if AI will substitute them because they don't have this clarity in mind of what is their day-to-day task, and even those who have clarity, I I myself included, really, I think we could do a much better job at defining what we are doing every day and what is really necessary for us to do ourselves. And let me let me really like I think I think this will help a lot of people when those particularly who are anxious about AI, when they will realize, and you when you do a simple arithmetic calculation, and you're like, okay, I have maybe 10 tasks over 200 that I do every month that AI can assist with. So it's like 5%. So that okay, that that's really really just for the sake of the example, but that will help you identify what can AI help with, what can AI not help with, and it will make it clearer, so it will address the two things that this article is saying, which is lack of training, but knowing how to use AI and be trained to use it requires to know better what you are doing on an everyday basis and determine what you can ask AI to do. Okay, so my piece of advice for what it's worth is trying to delineate what you are doing, what are the tasks, what are the inputs and outputs, convert that into SOPs. For if you don't know, SOP stands for standard operating procedure, and it's like a piece of paper or a document that describes step by step, like if an if an intern was to do that, what are the steps to follow and what does each step contain? So, for instance, if you want to do a research plan, okay, step one, and you will see that's really a fun exercise to do because ultimately you will realize, even if I think that I did the job correctly of laying down all the steps, I always, always, always miss some of it. Let me give you an example. If I do a research plan, what do I do? I collect requirements from a stakeholder, so a PM, more often than not, and you can you can see directly that this has already some assumptions. I'm saying a PM. It may not be a PM. It may be a UX designer, it may be a product designer, it may be an executive. So we already starting with, let's say, branching right away. So you request you gather requests. How do you gather requests? Is it a meeting? Is it an email? Is it a survey? Is it a combination of all that? And how do you gather those requests? Do you take notes? Do you act on a transcript? Do you give them a survey? And then okay, once you gather this request, how do you how do you extract the need? Does that need to fill out some boxes? Like problem statement, research question number one, research question number two? If not, how do you translate the questions from the stakeholders to the need for research, and so on and so forth? Once you have done that, you need to re-write the research plan. So, research plan, how is that defined usually? You have a template, so that's something that could go into templates folder for your intern. Okay, so research statement, what is the usual template to write a research statement? You do follow a template, even if you don't feel like it, even if you don't think about it, you do follow a template. When we are trained professionals with 10-ish years of experience, we usually follow a template in our mind. That's what makes us so efficient at our job. It's because we have learned and learned and learned over time, and we are way more, in a sense, I like to say rigid, but it's for the good because it allows us to dedicate more time to the things that need thinking, debating internally, but that is achievable because for the things that we did a thousand times, we do it so quickly and efficiently because we learned how to do it. Anyways, and so this kind of stuff you can encapsulate it into templates, like okay, research admin is done this way, research objectives are done this way, and so on and so forth. So that's okay, another example. And then when you write a research objective, what do you do? How do you phrase it? Okay, what does that mean? And then the feedback from your stakeholders, and so on and so forth. So hopefully that gives you an idea of the work that needs to be done for you to have a sense of have a sense of where does AI fit into this picture? So, sorry, this answers the first question, which is people fearing that AI will substitute them. So, you need first to have this visibility, then in those steps that you do, what is repetitive, what is not fulfilling to you, down to the very, very last detail task. And once you identify that, third question is what is something that AI can do well? So it's like three conditions at the very least. The first one is what is repetitive, the second one is what is not fulfilling, and the third one is what is AI doing well? And only if you have a yes at these three questions when you identify a task, then you can probably have AI do it for you. And so let me tell you, like, we are anxious about AI. We, I'm also including myself in it because okay, who's who has never thought about this? We might be anxious about AI taking our job, but it's because anxiety is lack of clarity. It's lack of clarity. If we have more clarity about what is our pipeline, what is inherently strategic, and what can AI do or not do, we will be much better off moving forward. So I think that any training that that is about AI to adopt it in the company needs to needs to include this aspect. Anyway, so that was for the first article. The second one is really interesting, and I'm sorry in advance, I'm gonna butcher it. Uh, it's from Peter Zach sorry, I I don't know how to pronounce the last name. I'm so sorry, Peter. So that's from the UXdesign.cc on Medium, UX Collective. And so I read it uh only once. I need to read it again because it's it's it's a bit uh so it's really really interesting, it's insightful, and at the same time, it's a bit complex. Peter says at the bottom that he's working on an entire book on this topic, so that's why probably it's it's so rich in insights and ideas. So I'm I'm I will try to extract what I understood from it, but really high level because I read it only once. So sorry in advance if I make any mistake. But ultimately, that's really interesting in the way that it he positions he he he tries to step back and position uh let's say the the the the the user of AI, in this case designers, and their relationship with AI. So this idea of us adopting new tools and the fact that we are asking wrongly the question, when can AI do it? So as you can see, this is highly complimentary with the first article. Peter is saying we oftentimes ask the question, What can AI do? But we should ask a different question, which is why are we doing this in the first place? And so he's observing that nowadays people have AI, and so it's so easy and convenient to go to AI and talk to it real time about what we have in mind, but we don't even step back and ask the questions, why are we doing this in the first place? Like it's no different from talking to another human. And so there is this idea, I find it um really insightful. And then there is the other um topic which is called from Peter, so he mentions that there is a concept, quote, sorry, there is a concept in educational psychology that I keep returning to as I watch the design community negotiate its relationship with AI. Lev Vygotsky called it the more knowledgeable other. The more knowledgeable other is not necessarily a teacher in the formal sense, it is whoever holds the competence that a learner in any given situation is reaching forward toward, sorry. It is the figure who can see both where the learner is and where they need to get to, and who can scaffold the gap between the two. Okay. And really, I love it. Thank you, Peter, for saying that. So, right now in the current Silicon Valley Ph2 designers is essentially this. Sorry, this is a quote. AI is your MKAO now, so more knowledgeable others. So, what Peter is saying is that right now we are treating AI as our more knowledgeable other and we are prompting it, expecting to be guided by AI. And it should be the other way around. We are more so AI is more knowledgeable in a lot of ways, but ultimately we are the masters of the purpose and goal and what why we are doing what we are doing and the constraints that we should give the AI before asking it anything. So that's the idea, and I love I absolutely love this idea. So it's kind of showing the toddler the way, basically, uh, and we can do that with AI. Also, I think throughout the article, Peter talks about the fact that language could be right now. Sorry, I hope I did not I did not get this wrong, and I'm working, I'm I'm I'm speaking from memory here after reading the article. That let's say right now language is kind of the interface between us and machines, and so that we need to master more and more the way we speak to machines, and for instance, one way of mastering is mastering the English language, for instance, or any other language. But ultimately, this is not the only way because he states at some point, if I'm not mistaken, that not everything that we conceptualize is encapsulated through language. This is not necessarily the first layer for us to conceptualize things. Because sometimes we can think about um relationships in space, for instance, without talking about uh language. And related to that ultimate idea, this was the idea that AI right now can have a lot of theoretical knowledge, but not applicable knowledge, and not realistic. For instance, he it can tell you about gravity, but it will not feel gravity, and so that there is a huge difference. So that's also something that I heard in another podcast, which was saying that ultimately AI can never, never, ever embody experience that each and every human has. We have experiences that we live throughout our lives, and this is uniquely human. And AI can never have that. And that's why I'm doing this kind of things, like podcasts and sometimes writing articles and and yeah, trying to place this emphasis on that to communicate to people my opinion, my take, my experiences. Yeah, AI cannot have your experiences. If you ask it to write a book, it will write a book, but it will be it will be bland because it will not have all your experiences, whereas if you write it, it will it will speak to people because it will have authenticity and and yeah, your experiences. So I love this idea. And then it's the idea, of course, as Peter mentions, to architecture the constraints, have the constraint in mind when you talk to AI. And then I would like to quote the end of the article quote: This is the choice the design community is facing right now, not whether to use AI, I think that question is settled, but whether to engage with AI systems as the MKO, so more knowledgeable other, or as the prompter, whether to define the parameters or just accept the outputs, whether to build the floor or keep asking why the ground is shaking. And I love that. It's like who do we want to be when we talk to AI? Who do we want to be when we interact with technology? It's the same as for any other tech and and yeah, use cases. Who do I want to be when I go to social media? Do I want to scroll and be controlled by the algorithm or do I want to be purpose-driven? It's the same. AI is a tool, like everything else, it's a tool. And if we use a tool, we need to determine the best use for a tool. But it it's part of a bigger picture. It's not just a tool for the sake of the tool. Because then we will realize when we achieve the tool, when we use it, when we use the outputs, then what's left? And ultimately people will realize. There is no other way. I do believe that people will realize. And again, as I said in other episodes, my two sense is that AI will not like let's say, hypothetical scenario, I do not believe that in XYZ amount of years people will be left out of jobs. Like, probably if we are left without a job, it's because we have not been able to identify what we do that is how can I say that is uniquely ours. AI should be a part of our process, and in my opinion, should always be. Right? It should not how can I say we need to we need to go above and have this kind of helicopter view of what are how our systems plug one into another and how does that play out bigger picture and use AI in it because it helps us being more efficient, yes, but it's not the whole thing being more efficient. Anyways, I hope I'm communicating my point across. Peter does it way better than I do, so thank you for sharing that in this article. And then we have another article from Christine on the UX Collective again. So I think this is complimentary with the first article that I shared, and maybe the second, which is what how could we what would it look like if we used AI as agents to power our workflows in a really efficient way? So Christine comes up with some great ways to leverage AI to, for instance, write a book. That is an example that I found. Uh write a book, and so this works with several agents, and the agents are analogous to a team of people who would work for her. That's my understanding. So that is really in terms of user experience, that is interesting again, because I think this is part of how it could be done by leveraging AI. I'm just wondering okay, so really concretely speaking, just imagine you have Christine prompting the tool, and the tool assigns the tasks to several agents. So you have an editor in chief, you have a sales and growth person, you have a voice person, product person, reader and advocate, and so on and so forth. And this is all highly digital, right? And she provides data and and knowledge to all of these uh sub-agents who leverage it to iterate on the book. I'm really doing a really uh sorry, really quick, quick, quick summary, but there is much more to this idea. I love this process, of course. Um and um so concretely speaking, you just sit down, you talk to Claude or whatever other tool you use, and you prompt it, and then of course it has a lot of context engineering, what I shared in other episodes, and then your your requests and part of the request is being rooted to subagents who would take part of their own job. So it's like assimilating. Ultimately, my interpretation is that we assimilate the work that would be done, or maybe part, sorry, to be correct, part of the work that would be done by these team members to an AI. And so several things here. I'm fascinated at this idea because it helps us to ultimately as an experiment to see to what extent we can approximate the work of a human being to what an agent would be doing, okay, that's one thing. But then I would also challenge how does that differ? What is the output? If you write a book, how is the output like? If you use your editor-in-chief as a sub-agent, which is totally digital and is an AI model, how does that differ from having an in-person editor-in-chief, for instance? Who's using AI or who's not using AI? So that's the thing. I think that sometimes we take some we do some leapfrog or we take some leaps, going from let's say I have an in-person editor-in-chief to I have an in-person editor-in-chief using AI, or I have an AI editor-in-chief. And I would be challenging all choices, like, but probably, of course, there are some other things in the equation, like uh the costs and the time, and uh, let's say the ability to manage a team and so on. I'm not saying that Christine is not able to do it, of course, but I'm just saying, like, experiment-wise, in terms of output and revenue and all the like, and profit, and also not only that, these are vanity metrics. Also, how does that reach other people? How does that help them in their daily lives? I would be really interesting to see to what extent this is better than well, is better, defined better, but to what extent this compares to a team of people using AI or not using AI. That's probably it. And also ultimately we see the value, we see that this is is is uh let's say is possible because there is a lot of context engineering, but ultimately what if the context engineering grows and grows and grows? By that I mean you have a great editor-in-chief because you give it all the documentation that the editor-in-chief needs, but ultimately, I'm just wondering if we will not observe that through time we realize that this editor-in-chief needs more and more and more and more documentation because it grows over time. Like of an asymptotic relationship between what a human is able to learn in their brain, and what you are able to provide an AI, an AI model. Meaning you're trying, we're trying to replicate what a human being is doing, but ultimately, is that the way to go? Or is it not better to use a human who is using sorry, not to use a human, to have an employee or a call a collaborator or a colleague, whatever, who is using AI to be more efficient? I'm wondering, in terms of really net outcome. Anyways, that is it for today. So it's basically the idea that people are not using AI as much as they can, as they could, because either they fear it or because we don't have enough training. The second article was talking about how we should flip the script on using AI. Basically, the idea that we should be the more knowledgeable other rather than the AI itself. We should show them the way, uh, we should um include them in our workflow and not them being the workflow. And finally, a human approach to agentic AI. All really great articles. I hope you liked it. I hope you learned at least one thing. And um feel free to disagree, of course, and and react and let me know your thoughts. And see you tomorrow for another episode. Until then, take care, bye-bye.