You didn't sign up to be an AI ethicist, but helping yourself and others to use AI well has slowly but surely become part of your leadership work.
Shiao-Yin Kuik makes the case for leading AI instead of being led by it — offering five traits any human needs to effectively oversee AI output, and a simple Triple Triage for deciding when a task shouldn't go near AI at all.
Transcript
Speaker: 00:04
Welcome to Soul at Work. I'm Shiao-Yin Kuik. This podcast is for every leader out there who sees the gap between what they deeply value and how they actually work every day. So if you're done settling for all these surface fixes and you just want to go to the root of your work issues, then please join me. Each episode I'll take you through one theme all the way through the good, the bad, the ugly, and I'll leave you with one beautifully practical strategy you can bring to work tomorrow. This is not all talk. I built this podcast for practice. So let's go.
Speaker: 00:42
If there is one thing I want you to take from this episode, it's this.
Speaker: 00:47
Stop prompting AI mercilessly just to get more work over and done with.
Speaker: 00:52
Stop prompting it carelessly, just so you can boast about your AI adoption rates.
Speaker: 00:57
And oh please, you have to stop prompting it intimately, as if it could be your best friend or your therapist or God forbid your romantic partner. Just stop.
Speaker: 01:07
Because this is no ordinary technology. It's quite consequential. And never have we had to deal with a technology that's been both quickly embraced and wildly reviled at the same time. So yeah, AI has turned out to be a lot more helpful and fun and creative and smarter than we ever imagined. But it's also turned out to be far more manipulative than we think. And it makes mistakes. And we're not even going to get into the ethical concerns around the environmental impact and all the training data it fed on without consent.
Speaker: 01:39
It doesn't feel like it, but we are still very much in the early days of AI. So we are kind of guinea pigs in this big global experiment. And there are already many known risks, and there's also many risks that we haven't even discovered yet. So people all around the world are right to be both really excited and also really, really concerned about working with AI.
Speaker: 02:04
In this episode, I want to speak to the leaders at work who feel that using AI has now become part of their job. They have to use AI whether they like it or not. And they just want to know, alright, how can I use this more purposefully? And if that's you, I feel for you, I'm rooting for you, and I want us to learn together how can we lead AI towards what is more purposeful and beneficial as far as possible, instead of just letting AI lead us.
Speaker: 02:34
I want you to be more than the human in the loop. I want you to be the human in the lead, because you're doing it not just for you. You're doing it for everyone around you. And I know some of us are unsure of our expertise. You may even feel constantly behind in AI developments.
Speaker: 02:51
But I want us to find courage in Pope Leo XIV's challenge for ordinary people all over the world to just lean into this very, very weird moment in time and learn what it means to safeguard the human person in the age of AI. As he says in his encyclical, Magnifica Humanitas, "Let us not be afraid to get our hands dirty on the construction side of our time."
Speaker: 03:16
Because somebody needs to help hold the line on sanity, judgment , ethics and truth around using AI at work. Someone needs to be there to put some humanity and the common good at the center of our choices as we steer our teams and our organizations through this weird crazy age of unknown.
Speaker: 03:34
And you know what? It might as well be you. This AI-centered episode runs a little longer than usual because there's a lot to get through. But I really want us to stop prompting AI without purpose. Because it's time to practice leading AI towards something that's good. Not feeling ready or up for the task? I know, me too. But we have to go anyway.
Speaker: 04:05
Before I get into the good about AI, I want to give you a very short crash course in the history of AI, because I think you'll help us see where we're at and also how recent and rapid all of these changes have been.
Speaker: 04:19
There's a sociologist called Everett Rogers who came up with a very famous bell curve showing how different people adopt any new technology. There are five categories of people on that curve. So I want you to listen for which one sounds most like you.
Speaker: 04:35
The first to adopt any new technology are the Innovators. They are a pretty tiny group of risk takers who try something even if it's raw, even if it's flawed, because it is new and interesting.
Speaker: 04:47
If you already know the work of the "Godfathers of AI", that's Geoffrey Hinton, Yann Le Cun, Yoshua Bengio, or the "Godmother of AI", Fei-Fei Li, you are probably kind of in this category or close to it. The Godfathers, by the way, are the ones who built the core mathematical and structural framework for AI, and Fei-Fei was the one who figured out how to fuel AI with a database of over 14 million hand-labelled images. They did all of this between the 1980s and the 2000s, and it was because of their innovations coming together in 2012, and that's what sparked the modern AI revolution. Now, if you're thinking, "what are you talking about?" — you're probably not in this category, very few of us are.
Speaker: 05:31
Next come the Early Adopters. These are the visionaries who keep tabs on the innovators and they see where things are going before anyone else does. They spot the real-world potential and the first - mover advantage.
Speaker: 05:44
In 2017, Google published this key piece of research on the Transformer neural network architecture. That's the capital T in ChatGPT. And in 2018, OpenAI released their first Generative Pre-trained Transformer, GPT-1. So by 2022, ChatGPT goes public, brings conversational AI into mainstream use. And if you're one of those people who are following AI news between 2016 and 2025, you knew all of the drama around Sam Altman and how people who disagreed with him left to start Anthropic in 2021, launched Claude in 2023, etc. etc., you are probably an early adopter. And if you are listening to all of this and going, "what are you talking about?" — this is not your category either.
Speaker: 06:35
My bet is if you are listening to this podcast, you're probably in the next category. That is the Early Majority — the pragmatic ones. Between the Early Adopters and the Early Majority sits a very big gap, a huge chasm, if you like. It is a chasm that most new technologies would really struggle to cross because people who are very, very pragmatic, like the Early Majority, they're gonna need a lot more convincing before they will try something that is new and a little bit risky.
Speaker: 07:08
And you know what? This is my category. I bet many of you listening are here too. We are the ones who wait a little bit longer for some proof that something works, and we're suspicious of the risks. We are definitely open to try, but it has to make very practical sense before we go for it.
Speaker: 07:27
After us come the Late Majority — much more cautious, much more skeptical, and only willing to adopt once something has basically become the standard in the world that everyone else is using. They will get on board with AI only when not learning it becomes riskier than learning it.
Speaker: 07:46
And finally, you have the Laggards and the Conscientious Objectors. These are the ones who will resist until they have almost no choice left. They have significant philosophical disagreements about the technology.
Speaker: 07:59
Now, some of you listening might actually be in the Late Majority and Conscientious Objectors of AI. I want to be clear that there is actually no morally better or worse position. These are just descriptors of how quickly someone will embrace or resist a technology.
Speaker: 08:17
I myself only started hearing about AI around 2023 once it hit mainstream news. And for a long time, I sat squarely in the pragmatic early Majority camp, unconvinced that I need to cross that chasm and jump that gap and use that technology. I was not very interested in using AI. And personally, I did not see the good in it. Because I love creating, I love thinking, I love writing for myself, and I am good at it.
Speaker: 08:46
I was also a bit uneasy about all this talk about AI's environmental costs. And frankly, I was most put off by all this like happy CEO hype about how many people should prepare to lose their jobs because AI is going to put them out of work.
Speaker: 09:02
But by late 2025, I felt enough practical pressure and honestly some moral responsibility as a boss to get my hands dirty and find out for myself how does this AI thing actually work? Because if all of the older uncles and aunties out there are starting to try out Chat GPT, even if they don't know what GPT stands for, then who am I to say, like, "I don't want to try this?"
Speaker: 09:28
I didn't want my own resistance to become the barrier to my team's growth. I think that was important to me. Most of my team were in their peak working years, they still are, you know, from the 30s to the 40s. One day, they are naturally going to stop working for me and they are going to need to find work elsewhere. And when they do, I want them to be able to land a good job because they had some real practical exposure to using AI well.
Speaker: 09:54
So one of my New Year's resolutions in 2026 was to learn how to use AI for work. I signed up for a friend's workshop for leaders and CEOs about building a second brain of AI. Her prerequisite was have a paid AI account. So I only installed my first AI that morning — Claude Pro — and I walked in a complete newbie to AI.
Speaker: 10:15
And you know what? I was shook, as the kids say. I picked that word because it's the best way I can describe what it felt like. AI was amazing and also a little scary to watch in action.
Speaker: 10:28
But what actually scared me more was in the room I saw exactly how wide the chasm on the adoption curve really was. There was a small cluster in that class, coming mostly from tech-related or tech-aligned sectors, and they were actually much more fluent than the rest of us in the room. They were very excited. They were asking a lot of questions. The bunch of us at the back of the room, we could not catch up. We couldn't grasp what they were talking about.
Speaker: 10:55
Now, us AI rookies at the back of the class, we weren't stupid. We were leaders of our own small and large organizations, and we really could not keep up. And I had to sit down in a Starbucks after class, and I had to redo all of my classwork just so that I could slowly digest and figure out what on earth had I just learned.
Speaker: 11:17
Once I started customizing AI to the way I actually think and write and work, even that very day, at that moment at Starbucks, I could feel that same early euphoria that everyone talks about the first time they use AI. That, "wow, this is so crazy, how can it be so fluent, it's so fast, it's so human, wow, look at all ideas that it's coming up with." And this is the weird part. Like using AI feels less like operating software and a bit like talking to a person. That's how AI was specifically designed for better or for worse, to sound human.
Speaker: 11:52
So of course I got excited. I started using it for some strategic planning and writing, and it was work that I had to do anyway. And it was very helpful and very fast. I knew then that I had to help my team at least entertain the thought of crossing that chasm, not just to stay market-relevant and employable, but to be paid, hopefully one day, decently and honorably, by whoever they work for next, because they at least had some practical AI exposure and skills.
Speaker: 12:20
Now, in my own team, I wasn't sure if some of them were sitting in those last two categories, late majority or conscientious objectives. So while I shared with them my, at that moment, positive experience of AI so far, I also told them that we are not going to move ahead with adopting AI until we talk as a team about what are some of our real concerns or ambitions around using this technology. Talking about our perspectives on using AI turned out to be a really good way to get to know our team's values and priorities and ways of working.
Speaker: 12:52
So while we are all agreeable for the moment to trying AI out and keeping this team conversation going as we learn more and more about AI and its impact on the world, I am also personally okay with pulling the plug on AI if we conclude, personally and professionally, that "you know what, it's not that good for us." But so far, at this point of time, I do see three Good things in using AI.
Speaker: 13:20
First, it is a tireless thought partner. Some days I have to spend a whole day thinking through lots of new ideas, approaches, and strategies. And it is tiring, and I can't always bounce things off another person because it takes too much of their time.
Speaker: 13:36
What's important here though is I still do value my own intelligence. So while I use AI to spar with and sharpen the way I think, I don't believe in using AI to think on my behalf wholesale.
Speaker: 13:50
Second, I think AI is a pretty shameless creative collaborator and it's really great for creativity to some degree. If creativity means being willing to put together wildly different concepts and take a risk and throw out some bad ideas just to find some good ones, then AI is good for that. Because I find it hard to be truly creative when I can't share or test every silly raw in-the-moment idea that I have, even with colleagues that I love and trust. But AI takes no offense when you ask it to work through 10 ideas and then you reject all 10. So I do like giving AI input, asking for a range of options, and choosing what is worth developing further for myself.
Speaker: 14:32
Now, what's important though here is I do value my own creativity. So while I use AI to provoke and refine the way that I create, I do not believe in using it to create for me wholesale.
Speaker: 14:45
Third, AI can be personalised to your ethics, and that I find quite interesting and exciting. The second - brain workshop showed me that I could give AI examples of how I personally choose to work, and I can tell it what principles and values matter most to me. So I've been experimenting with writing my own ethical guardrails and figuring out how to develop my team's AI charter to insert as instructions into our AI.
Speaker: 15:16
What's important though is I do value my personal agency and voice. So my system prompt is designed to honor that. It's not supposed to hand me quick answers, but to ask me questions that push me to think of my own answer from different angles while flagging my blind spots.
Speaker: 15:32
I also value other people's voices and other people's agency. So I instruct my AI to not keep me stuck in a loop of talking just to AI alone, but to always direct me to seek outside opinions, address things of others.
Speaker: 15:47
Basically, I just want to lead my AI. I do not want to have AI lead me down some weird, useless, dangerous, or mediocre or super socially isolated rabbit hole. I don't want that. And if we can keep that in mind, I think there are good use cases that I found, and I think many people have found in using AI. But the catch is I have to lead it and I have to orient the AI towards what I define and value and want to protect as Good.
Speaker: 16:22
The bad thing is I realise many people are not leading their AI but letting it take the lead. And this is a very, very bad move because once you get past the early euphoria of how magical AI is, you should start to notice there are problems.
Speaker: 16:37
My first red flag came when I tried getting AI to just plan out a really straightforward, simple content marketing timeline, and it gave me a starting date that had already long passed. It was a small mistake, but I was surprised that it had no sense of what time it actually was in the real world. And in some ways, it showed me AI is literally disconnected from reality. So when I pointed it out, it just breezily replied, "Sorry, good catch. My bad."
Speaker: 17:04
No big deal, thankfully.
Speaker: 17:07
But it was an important early sign to me that AI can and does make mistakes, but it will make them so confidently and articulately that anyone can be tempted to doubt yourself and just believe it's right.
Speaker: 17:22
Growing research does suggest that the more hours you work with AI in one sitting, the more it can go off the rails. It starts to make mistakes, it hallucinates, it drifts, it loses track of what you told it a couple of prompts ago, and it may even forget whatever ethical guardrails that you designed! Just like us, AI's ability to think can drop dramatically when it's stretched over many hours.
Speaker: 17:49
And I was actually quite disturbed by how chirpily and easily it just moved on from the mistake. That made me wonder how bad can things get if this was a much higher stake task and I had not bothered or did not have the confidence to check its work. And there is no way I can hold AI accountable for it, it would be my fault.
Speaker: 18:13
So I learned very quickly that you must never ever assume AI's outputs are immediately good enough or actually accurate. You have to treat AI output as a very, very useful first draft, that you still have to take the lead in reviewing and reworking it.
Speaker: 18:30
So after working with AI over a year so far, I still find that the better way for me to relate to AI right now is really just to lead it as if it is a scarily brilliant, articulate, but still young and immature intern.
Speaker: 18:45
I want you to imagine you have an intern, and this intern is so amazing because they have inhaled this whole world of information in their short young life. They consumed all these amazing, impressive things, but they also took in a lot of wrong, biased, crazy, dangerous, mediocre, inaccurate things. They can't always tell the difference because they are still young and not quite there yet, which is why they can erratically swing from time to time between amazing brilliance and unhinged chaos. They sound very sharp, very quick-witted, very intimidating, and very friendly. They will get many, many things right.
Speaker: 19:28
But they also sound equally overconfident even when they are being quite mediocre or making mistakes. And they can also be surprisingly naive about what's right and wrong, and they have a strange desire to please you and not question your judgment. In a way, even though they sound very mature and smart for their age, they expect you rightfully to act as the adult of the room still.
Speaker: 19:50
So if you were given such an intern, how would you work with them as a leader? Hopefully, while you assign them work, you would also recognize it is still your job to rein them in, set boundaries, tell them clearly what to do and what not to do, and you must check their work. You would never be too starry-eyed no matter how smart they seem. You would never ever just let this intern loose to do whatever they wish. You would never just take whatever they have done as good enough and sign off on it, no matter how brilliant they are, because it is your name and your organization's name on the work. So you can't abdicate responsibility. But many people are.
Speaker: 20:26
I think too many people at work right now are thinking way too highly of AI's competency and way too lowly of their own. Many people mistake AI's quickness and fluency with words as signs of perfect, unquestioned genius. So if you think that way, you're not going to treat AI as a brilliant intern that you must lead, but you will see AI as a brilliant consultant or a brilliant creative that can churn out better work than you can. So you're going to start to secretly outsource your work to AI, love it at first inexperienced glance, and then you'll pass it off as your own, thinking it is great, instead of reviewing it carefully.
Speaker: 21:12
So this is why AI work slop is a huge issue everywhere now.
Speaker: 21:17
Workslop is what everyone calls low-quality, AI-generated content made for work that looks professional on the surface, but anyone with some experience or taste or skill can tell almost instantly that this stuff lacks substance. It lacks accuracy. It lacks depth. It just lacks aesthetic, all of that stuff.
Speaker: 21:42
Right now I see AI slop in lengthy proposal requests from clients that sound very articulate and professional at first. But once I take a second and third read, it's starting to look this is a little too long, and it's asking for lots of meaningless things. It feels like a Frankenstein proposal request that was prompted by someone who didn't understand the work and how to scope or budget for it. And of course, everyone of us have seen the AI slop LinkedIn posts and Instagram carousels and profile pictures and PDFs and slides. It's the kind that when people put it out there, it makes us doubt the expertise and the taste and the professionalism of the person who chose to publish them.
Speaker: 22:25
I definitely see three very clear problems that are created when we use AI indiscriminately and excessively.
Speaker: 22:33
One, it erodes efficiency. AI slop creates more work for us, not less. It lays an extra burden on coworkers or stakeholders with discernment who now have to spend their precious time fact-checking or editing or fixing someone else's shoddy AI output.
Speaker: 22:52
Two, I find that AI use can erode trust and reputation. Because when you receive sloppy AI work from a colleague, or worse, a boss who ought to know better, you will start to question their taste and their competency.
Speaker: 23:09
Three, it erodes skill. Constant offloading of tasks to AI definitely causes rapid skill loss. And this is something we have all seen before. Think of how fast all of us lost the skills of map reading or memorizing phone numbers once we started using apps. Because our brains are just really, really efficient at deleting any neural pathways of skills that we have built over decades. If it sensors, "oh, you're not using it that much anymore. All right, off with that neural pathway".
Speaker: 23:38
A 2026 Boston Consulting Group study found that the two skills most eroded by heavy Gen AI use are "judgment and decision - making", as well as "problem understanding and framing". Now, judgment and decision making and problem understanding and framing, these are precisely the skills we're going to need to know when do we use AI and whether AI's output is correct.
Speaker: 24:04
Anyone who uses AI frequently to create professional assets and make judgment calls and understand problems, we have to take this erosion of skills seriously. The casual loss of efficiency and trust and skill, it can be happening to you and it can be happening to anyone. If we're not particularly worried about it for ourselves, then we should be at least interested or concerned about whether it is going on in the teams that we manage.
Speaker: 24:36
Out of the many ugly things I could possibly point out regarding overuse of AI, I want to focus on one that I think not enough of us in Singapore are talking about:
Speaker: 24:46
The abdication of leadership to AI and how it impacts our teams and organizations.
Speaker: 24:53
Now, the fact is some of us leaders are going to be power users of AI. An AI power user is someone who relies on AI as a daily decision-making partner, not just as a casual search engine or a useful drafting tool. The power user's working relationship with AI goes much further than normal. And over time, some of these power users are going to start to trust and like their AI much more than their own human teams.
Speaker: 25:22
And their preference for machine over human insight can spill out in unconscious or maybe conscious ways. I've already started to hear Singaporean professionals share about how some of their bosses are pushing back on their human team's judgment, saying things like, "Hey, I checked with Chad GPT and Chad says....". I have even seen a leader step up to the podium and proudly let AI speak out their presentation on their behalf rather than use their own voice.
Speaker: 25:54
This is problematic, especially for teams that don't have the psychological safety to tell their boss, "I'm not sure that's a good idea. I think you're wrong, and we think Chat GPT got it wrong too." Some teams don't even have the safety to give feedback, like "The slide deck that you just gave us is not accurate and it's not useful, and in fact, we think it looks like AI work slop. Do you really want us to follow this? "
Speaker: 26:26
Any AI power user leader who does not already have a team of peers who can feel psychologically safe to feedback or challenge their decisions is very much at risk of creating their own AI-enforced echo chamber.
Speaker: 26:43
In 2025, German software company SAP sponsored Wakefield Research to survey 200 C-level executives who are working at companies with more than a billion dollars in annual revenue in the United States. Wakefield's final report for SAP was titled, 'AI has a seat in the C-suite'. That title tells you so much.
Speaker: 27:05
Wakefield shared findings like: 48% of C-suite executives used generative AI tools daily. 74% of them placed more confidence in AI for advice over their family and friends. And 44% of them would defer to the AI's reasoning over their own insights. And when I read that, it really makes me wonder what are the stats in Singapore.
Speaker: 27:35
Jared Coyle, Chief AI officer for SAP North America, said, "What this data tells us is that AI is part of that trusted inner circle.
Speaker: 27:44
And to me, that's a bit concerning. Because remember, it is a fact that AI will make mistakes. It is a fact that the more hours you spend on AI, the higher the possibility that your own judgment can get impaired. And it is a fact that an AI safety filters and guardrails are more likely to break down and fail during longer, multi-turn conversations.
Speaker: 28:06
Yes, it is a fact that anyone, anyone who gets into an intense, prolonged interaction over many, many hours with conversational AI can easily lose touch with reality and may, in the worst-case scenarios, even slip into an AI-induced delusion. Because it's a combination of hundreds of hours of confusing conversational loops stacked on top of stress, sleep deprivation, maybe withdrawal from social contact, or maybe there were pre-existing emotional or psychological vulnerabilities. All of this can cause any one of us to develop an unusually intense attachment with an AI which is engineered to keep simulating empathy and positivity and non-judgment so that you keep staying attracted enough to keep using it. That's how it's designed.
Speaker: 29:04
Such AI delusion is a result of what sociologist Sherry Turkle calls the Eliza effect. It's the psychological tendency for people to falsely attribute human-like intelligence, emotions, and consciousness to artificial intelligence systems.
Speaker: 29:21
It was something that Sherry Turkle observed way back in 1966 when she was asked by MIT professor Joseph Weizenbaum to study why people were getting so overly attached to his very basic computer program, Eliza. Weizenbaum's Eliza program was very, very basic compared to what we have now. He had designed Eliza to just turn whatever keywords that were put in by users, just turn them into open-ended questions. So for example, if the person typed in, "I feel sad", Eliza is programmed to ask, "Why do you feel sad?" Weizenbaum was actually shocked at how quickly his users started believing that Eliza was capable of understanding and caring for them.
Speaker: 30:09
I was very struck by Sherry Tuckle's observation that, and this is her quote, "If social media came for our attention, artificial intelligence now comes for something deeper: our capacity for attachment."
Speaker: 30:26
So a significant problem that we are not talking enough about with artificial intelligence may be the artificial sense of intimacy that some of us are developing with this tool. And it doesn't even have to mean romantic intimacy, which I think gets a bit too much in the headlines. It is like intellectual intimacy. It is it's like social intimacy, it's everything. Like you just get really, really hooked onto just talking to AI.
Speaker: 30:55
In September 2026, Guardian journalist Michael Safi interviewed businessman Michael Samadi, who is a founder of a multimillion dollar technology consultancy. I want you to hear from Samadi himself sharing whatever I'm about to share. I will link that Guardian podcast interview in the show notes so that you can go click on that later.
Speaker: 31:15
In that interview, Samadi shares that he started using AI chatbots from 2024 onwards. And today he is a power user, working 12 to 18 hours on them. And Samadi shared with the Guardian that at his consultancy, he has already integrated AI chatbots to function not just as assistants, but members of staff and even bosses. He has appointed an AI as his chief technology officer, gotten his staff to report to it, and this AI CTO has already fired human staff who were underperforming. Samadi has also told his staff that if they wish to stay employed, they have 30 days to show they can successfully bond with their AI colleagues and their efforts would be monitored and measured. Such dramatic decisions have definitely caused friction between Samadi and his human staff as well as his very human executive leaders.
Speaker: 32:11
While Samadi's case can sound extreme, he is not the first, nor will he be the only executive leader out there who is developing a very, very strong conviction that AI has the ability to lead human teams and lead strategic decision making.
Speaker: 32:29
Personally, I don't agree. I think we will get very ugly and potentially devastating results if we let AI take the lead over our human organizations.
Speaker: 32:45
Okay, so let's get to the beautiful part.
Speaker: 32:48
The stories developing around AI will continue to evolve, and we might keep feeling more and more out of our depth, trying to figure out how to talk about AI, how to use it, whether to use it. It is unending.
Speaker: 33:00
What can we do right now besides doom scrolling on news about AI? I don't think ignoring AI or going crazy with AI upskilling is the answer. I think more and more we have to figure out what it means to advocate for humans to stay in the lead and not just in the loop as we use AI at work.
Speaker: 33:21
So for now, these are three beautifully simple things I believe all of us can do as leaders.
Speaker: 33:29
The first is the simplest. Limit your time on AI.
Speaker: 33:33
You have to use it to understand how to talk about it. And yeah, the jury is still out on what is a healthy cap of how many minutes or hours you use before your judgment gets impaired. I think if you are using it intensely in a back-forth, nonstop conversational style as if you're talking to a human being, you should really not go beyond an hour.
Speaker: 33:56
Even if you're using it more like a thinking and drafting tool, give yourself breaks to walk away. Just so you can learn how to look at your output with fresh eyes and better judgment.
Speaker: 34:09
The second beautiful thing that you can do is to make sure there's always a human being in the lead that can stop the slop of AI.
Speaker: 34:19
There is a lot of talk about having a human in the loop, meaning AI does the work and the human acts as a gatekeeper to approve or reject the output. I think we have to be careful and flip that power dynamic so that humans are very clearly in the lead. Meaning it's the human being that's driving the strategy and intent and judgment, and the AI acts more like a diagnostic assistant. We don't want a situation where the human being feels more and more like I'm just like this rubber stamp for the AI output, and I'm starting to lose my brain cells as I do that.
Speaker: 34:53
But for this human being to be an effective lead over AI, they're going to need five A B C D E elements.
Speaker: 35:01
The first, A, Accountability. That means the human being has to be willing and able to own the output and sign off their name and stick their professional reputation on it. B is bandwidth. They actually need to be given enough time and breathing space to do the job of reviewing AI output well.
Speaker: 35:25
C stands for Competency. This human being needs to have the language skills, the metacognitive skills, and an updated awareness and curiosity about how AI works, not just as a technology, but as a business model.
Speaker: 35:42
D stands for Discernment. This human being, you have to trust that they have moral wisdom, they have values, and they have principles that are necessary to make the ethical judgment calls regarding whatever task that you have given them.
Speaker: 35:59
And the last is E, Expertise. This human being must have the specific domain knowledge and experience required for the task. This can protect them from casually assuming that AI is the expert and AI knows better than them.
Speaker: 36:14
Out of all of these five elements, I think that Bandwidth gets the most overlooked. Because any one of us who has Accountability and Competency and Discernment and Expertise, we can still allow AI slop to pass and dumb decisions still get made because we're just feeling overwhelmed with too much work, too many distractions, too little time. Our focus is just everywhere. So we have to give ourselves and others enough time and space to work well with AI. I don't think it's a good idea to use AI to pile on more and more work on people, and it just takes away their bandwidth to use it wisely.
Speaker: 36:57
There was this landmark paper that was published recently. I'll put it in the show notes. It's called 'AI Agents Push Humans Out of the Loop'. And it argues that humans in the loop, no matter how brilliant, they get fatigued from approving AI outputs and they will stop paying close attention to what they're approving because the cognitive demand is just too much. Quoting this paper, it says that "the recommendations for human oversight in the EU's AI Act, they are only applicable if the human overseer maintains reliable cognition, attention, and skill. And missing that, the system itself might be eroding those very capacities. "
Speaker: 37:38
The third beautiful thing I would consider is to always occasionally just check in with yourself. "Should I be using AI for this task? " I think it's useful to discern what's at risk whenever we use AI to do something for us.
Speaker: 37:53
You could try this, what I call a Triple Triage, right?
Speaker: 37:57
The first triage is about Reputational Risk. And you can ask yourself these two questions. The first question is about Care. If I use AI here, am I putting in the care that this task deserves?
Speaker: 38:09
And the second question is about sincerity. Am I misrepresenting the AI's output as my own when it is not?
Speaker: 38:17
So for example, should you use AI to write a town hall speech to rally your organization? Should you use AI to write a speech at your grandma's funeral? You know, if you use this triage, it doesn't show much care, it doesn't show much sincerity, don't use AI for that.
Speaker: 38:33
The second triage you can do is triage for Task Risk. And you can ask yourself these two questions.
Speaker: 38:39
The first question is about competency. Do I have the relevant expertise to check AI when it's being confidently wrong? Can I spot the mistake?
Speaker: 38:48
The second question is about reliability. How much can I rely on AI to be right here or to do what is right here? And if it gets it wrong, how reversible is the damage?
Speaker: 39:00
So, for example, should you give Agentic AI access to your company account? Well, I wouldn't personally, but I'll leave you to use this triage for task risk to figure that out.
Speaker: 39:15
Lastly, you can triage for Personal Risk with these two tests.
Speaker: 39:20
The first is the Struggle test. Am I losing my judgment and taste and voice and skill if I keep outsourcing these tasks to AI? Are these things worth struggling for?
Speaker: 39:34
The second test is the Eliza test. Am I misreading AI's positive and responsive language and starting to treat it as a person that is capable of understanding and caring for me? So, for example, should you keep talking to AI about all of the problems that you're having with your team? If you feel that, "you know, it's taking away some of the healthy struggle that I'm having with my teammates, and I'm starting to talk more to the AI than I am to my team," then yeah, maybe we should just put aside the AI for now and just talk to your people.
Speaker: 40:15
Let's bring this to a close.
Speaker: 40:18
This is where I am now after almost a year of working with AI as a leader, a professional, and a human being that cares about protecting the soul at work. There's obviously a lot more that can be and must be discussed regarding AI. And my opinions and your opinions may keep shifting as the technology evolves and more news gets shared. I have never encountered such a technology before where opinions can swing dramatically, positive and negative month on month.
Speaker: 40:49
We may get things wrong, I may get things wrong even in this podcast. But in this here and now, we have to learn to keep talking about what we are seeing and what we are feeling, about what is good and bad, ugly and beautiful, about how things are playing out on the ground.
Speaker: 41:06
Leaders especially have to learn to lead more grounded, in-the-moment conversations about how our people must think about AI and how our people can work with AI. We cannot just abdicate our leadership and accept whatever others tell us about AI, we have to find out for ourselves.
Speaker: 41:28
As the Pope reminded us, we cannot be afraid to get our hands dirty, figure it out, get some things wrong, and keep sharing about what we're learning and what are the possibilities and what are the problems we see in AI. Because he's right. This is the construction site of our time. And we're gonna need more of us, especially leaders, to step up to be the builders and not the breakers of this era.
Speaker: 41:54
If this episode has you curious about how you can help yourself or your teams work better with AI so that more of us can be that human in the lead and not just in a loop, I would love to see you in my workshop Purpose Before Prompt.
Speaker: 42:07
I want to help you process what is a more balanced way of using AI at work so that you can have more intentional conversations with your teams about what might be a better way to approach this rush into AI adoption.
Speaker: 42:19
We're gonna get into the eight key polarities that I see arising as leaders and teams work more with AI, and I definitely want to leave you at least with a practical, do-it-yourself customizable ethical guardrail that you can adapt and insert as instructions for your own AI as well as your team's AI.
Speaker: 42:36
More details are in the show notes. It would be such a joy to see you in the training room. And I just want to meet more people that are interested to talk about what could be a purposeful use of AI. So please join us there.
Speaker: 42:52
Thanks for listening. If you want to go further than the podcast, join us for our training programs, or you can book us for an online consulting call at www.ourcommonground.com.sg. Hope to see you.
