Why Smart People Get Generic Results From AI

Galen Low: Let's face it, we're all tired of reviewing deliverables, and reports, and status updates that all reek of generic AI. But while it's easy to blame the technology and just leave it at that, there's something more going on here. Most people are getting generic results from AI because they're asking generic questions. And most people are asking generic questions because they either don't understand how the technology works, don't have the confidence to challenge AI, or feel like an imposter among their peers, or sometimes all of the above.

And as a result, many people are losing the skills that we need the most right now: curiosity and the ability to think critically. To help us fix that within our teams, I've brought in an AI expert and educator who is hellbent on lifting the tide to raise all boats so that imposter syndrome, lackluster prompts, and generic AI outputs become a thing of the past. And so that the things that humans bring to the table move back into the spotlight. Hope you enjoy the episode.

Welcome to the Digital Project Manager Podcast—the show that helps delivery leaders work smarter, deliver smoother, and lead their teams with confidence in the age of AI. I'm Galen, and every week we dive into real-world strategies, emerging trends, proven frameworks, and the occasional war story from the project front lines. Whether you're steering massive transformation projects, wrangling AI workflows, or just trying to keep the chaos under control, you're in the right place. Let's get into it.

Okay, today we are talking about how imposter syndrome around AI can hold our teams back in subtle and not so subtle ways, and what practical steps we can take to help them find the balance between being inquisitive while still strategically interrogating the results from their AI tools.

We're gonna be stepping through a practical framework for getting our teams to ask better questions and assess the outcomes critically, and what impact that can have on a business's overall AI strategy. And we're gonna be talking about skill atrophy and how to keep teams leveling up as the world changes around them.

With me today is Mashhood Ahmed, Consulting Project Director at M1 Consultants. Mashhood helps businesses design and deliver generative AI enablement programs that drive tangible impact. Using a proven four-level GenAI adoption framework that covers organization-wide AI literacy, department-specific workflows, agentic automation, and training programs, Mashhood helps teams push forward with AI in a way that leads with human ingenuity.

To make that all happen, Mashhood wields his 20-plus years of experience in program, project, and change management. For him, AI is more than tools. It's about embedding AI into how work actually gets done.

Mashhood, thanks so much for being with me here today.

Mashhood Ahmed: Thank you very much. Thank you for having me.

Galen Low: I'm really excited to dive in.

I caught a couple of your talks around this subject, and I just think that the angle is so well-balanced, and it just really resonated with me, just your POV on how we need to look at AI in its implementation, in its infusion into our work. I think it's really pragmatic, so I'm excited to dive in today.

You have, of course, collaborated with us here at The Digital Project Manager a number of times. Your message always resonates. And so I thought maybe what we could do is I'll start off with a big, hairy question, but then we can just go where the wind takes us. Along the way, I'd love to touch on what it means to have an inquisitive mindset when it comes to AI and why that matters, what steps we can take with our teams to balance that inquisitiveness by strategically interrogating our AI outputs, and also just what's at stake for the future of AI technology if our teams continue to operate based on, you know, surface level questions and take generic outputs at face value when they're interacting with AI.

So yeah, I know that was a lot, but how does that sound?

Mashhood Ahmed: Sounds perfect.

Galen Low: All right. Awesome. Let's dive in. I'm just gonna start with a bit of a broad question because you work with a number of organizations, helping them get fluent with AI so that they can sort of build it into their ways of working.

But I've heard you talk about people using generic questions to get generic outputs and then just running with them as is. So I thought I'd ask, if you had to guess what percentage of people using AI in a professional context are only getting average generic results from AI and just thinking that that's as good as it gets?

Mashhood Ahmed: Yeah, that's a very good question. How many percentage of people are using AI in a generic sense and getting generic answer? My guess would be about, you know, 40 to 50% of the people are doing that. And what I would do is I would give all of your audience a little test, okay? To check what I mean by the generic response that you get from AI Most of us have played some game at some point in time in our life using a dice.

What I would like to do, and I've run this experiment with like hundreds of people in the room, twenty people in the room, and every single time I'm getting the same res-- similar-ish response, right? The numbers a little bit vary here and there. Open Copilot in one tab, in another tab open ChatGPT, in another tab open Gemini, in another tab open something else, okay?

I want you to try four or five ones. Typically, I would do like a, you know, in a bigger room with like a one tool, but let's just try because you're gonna experience by yourself. And put up a simple prompt, and the prompt is "Roll a six-sided dice for me." When you run that prompt, what will happen, most of the generative AI will give you the answer four, okay?

And as I said, I run this with like hundreds of people in the room. Typically, eighty-five to ninety percent of the people usually get four. There are occasional few people who would get different. And I have received feedback for getting on ChatGPT, on Gemini, Claude, Groq, Perplexity, and others, right? So again, all of them make the same mistake, right?

Now, what happens is that whenever you ask AI to run or roll a dice, it does not know how to roll a dice, right? Statisticians and mathematicians have taught AI in a model that, hey, if you need to run a die-- roll a dice, what you need to do is you need to take the one in six, those are the six options, take the average and round it off if needed, and give that mediocre answer.

So when you're getting four, the average of this is three point five, round it off to four. And that's why eighty to ninety percent of the times you will receive the response as four. Now, I want you to experiment something more. Go and type in the same window another prompt, "Roll a six-sided dice for me with each side getting equal probability," and, you know Have some biased comments you know, "I would like the guide- the highest number or the lowest number," or something crazy, right?

I have seen people putting, like, different types of prompt, and sometime I collect that. Then you will see that every single time you're doing it, you are getting a balanced response because you ask it to well-rounded, well-shaken dice or, you know, something like that. And what this tells you is that if you ask a generic question "Roll a six-sided dice for me," you will get number four 80 to 90% of the time.

The reason being is AI does not understand the context. It does not have a background about your project, about your judgment, like human judgment that we apply. And that's why it's really important to have your prompts or your agents have as much context and as many constraint in the guardrails to get the great response.

Galen Low: I love that. And actually, I opened a couple of LLMs right now, and you're absolutely right. I only had three tabs open. Two of them rolled a four, one of them rolled a six. I haven't got around to the inflection of the context, but I think that's hugely apt. And I think coming back to, of course, you know, neither you or I have done, you know, deep research, and neither of us are statisticians per se.

But I think that 40, 50%, that range that you mentioned, I think is... You know, I would agree. I think there's so many people who are still using AI like Google Search, you know, with just a simple four or five-word prompt, which might not be the worst thing in some contexts. But I agree that the complexity of the projects that we're running and the people that we're working with and, you know, there's a, a bunch of history and knowledge that, you know, sits between our two ears that is not in what the LLM has trained on, you know?

And without us saying that, then it's gonna give generic results. And the thing I love and hate the most about where generative AI is right now is that it's so convincing, right? And people are publishing this. This is not something novel. But it looks so good and convincing, and then it you know, tells you you're awesome and you wanna run with it.

You're like, "Cool, I'm done." This is good. This is way more polished than what I would've done myself, so I'm just gonna take it and run with it. And we'll get into some of the reasons why we do that. But I think the other thing is, is it just looks so convincing no matter what, even if it's very mediocre.

Mashhood Ahmed: Yes. I use a trick called persona switching. And what it is you know, you get your report ready, you get your product charter ready or, you know, whatever document, email ready. You can simply put a per- persona switching prompt, which sounds something like, "Now read this email or read this document from this person's perspective who is supportive or not supportive of this project," a little bit more context about that person.

And then you will see the response. And then now you may decide to edit the document, or you will say, "You know what? I will keep this question and prepare for that. If somebody asks me these questions in my next meeting or in my next you know, when I'm presenting this, about this project in front of a large audience or in front of the board, I have those answers prepared," right?

And this is very important because when we are working the project, we get blindsided by so much noise, right? And we can leverage AI to switch the persona. So, for example, look from the CFO perspective, who's not convinced that we need to spend $1.4 million to save this project. That's a different context.

And again, sometimes we don't even think about those things working as a project manager because we are busy. Now, if AI can help us do a lot of the other work, this is where a project manager can focus on these critical thinking skills and asking the right question, and that's one of the very important skill set, and we-- I call it inquisitive mindset, and we're gonna talk about it today as well.

Galen Low: Awesome. Yeah, let's get into that. I love that because, you know, I think in so many different ways, it's not doing the job for you. It's actually helping you build empathy with your stakeholders so that you don't get blindsided. You getting blindsided is usually as a result of you not having anticipated somebody else's mindset.

We'll get into that, but I just wanted to zoom out a bit because, you know, I've seen you use the word imposter syndrome in your talks and posts, you know, and for those who may not be aware, right, this is a phenomenon where successful professionals feel like they fooled their peers and maybe don't deserve their successes.

And in the context of AI, I think it's arguably a bit different than just being undeserving of success. I thought I'd turn it to you. What is it about AI that triggers our imposter syndrome, and how does it impact our approach to AI?

Mashhood Ahmed: So there are a number of things that triggers imposter syndrome that feels like, oh, I'm not catching up.

I don't understand this much, or everybody else getting it. Number one, it's invisible. We get used to of like input, process, output. And that's how we kind of, you know, spend our careers, right? We give input, something process it, and we understand the process workflow. Here in age of AI, the process workflow varies depending on the context and depending on the prompt details that you provide, right?

So it's becomes a black box, and it makes us a little uncomfortable. Sometime it makes me uncomfortable because when I'm running li- like live demos, it happened with me, three times the demo did not work the way I was anticipating, right? It did work for 10 time, but one time it did not. So that's like makes us like l- uncomfortable.

It's an invisible box. The other thing is that you go on LinkedIn, you go on social media, you will seem like everybody's doing it, but not everybody's doing it correctly. You know, I'm still learning, right? I'm learning everything new. The other thing is, as a project manager or the PMO manager, we get trained or we get used to of knowing everything what is happening on the project, right?

So why there's this risk? What is this issue? What is the problem with this vendor? And, you know, you know, somebody pulls me into like a specific meeting, then I can explain what happened with this vendor. I have the whole history, right? Everything is invisible here. So that's makes us very, very uncomfortable, and that's what causes imposter syndrome because we are not in control.

Humans Tend to be in control, okay? Yes. That's one of the thing that we used to do, is that we want to control everything. We are control freak species, right? And when the control goes to machine, that freak us out.

Okay? And that's where the imposter syndrome comes in, because we do not know what the machine gonna say, and then I need to double-check and verify everything, and sometime it's creating too much noise for me.

It's creating more work than really helping the value, and that's where the inquisite questions and the noise and other few topics that we'll discuss today.

Galen Low: That's really interesting because, you know, in a way it's two prongs. One is, you know, I think what we popularly... Is that a word? Yeah.

Anyways, the popular definition of imposter syndrome is feeling like a fraud because everyone around you looks like they're doing so much more awesome than you. LinkedIn feed, absolutely worst place to go to build your confidence in AI, in my opinion. It's like, "Seven things you're doing wrong with your prompt," you know.

"Why your agents continue to fail." And I'm like, "Oh, my gosh, I must be doing all this wrong," you know. "I'm terrible." And I think the other prong is, I think, what you said earlier, too, is just it's easy to look at some of these outputs and not want to criticize them. And it's like the other human thing, which is "Okay, well, it looks fully baked."

If I was to give feedback on this now, right? Just like as a project manager, you receive deliverables, they are, you know, 98% complete or 100% complete, and you're like, "Okay, if I provide feedback now and deadline's tomorrow, then, you know, we're gonna go into the spin, and we have to unwind it all, and then we have to kind of like rebuild it."

In this case, it's like, A, black box, like you said, B, looks pretty finished, and, C, if you're not feeling confident in that, you're like, "Well, I guess AI's smarter than I am. I'll just... You know, what do I know compared to this thing that's trained on the entire internet, the entire, you know, history of humanity since the Information Age?

What do I know compared to this?" But to your point, its blind spot is the context, is it doesn't know, you know, necessarily what was said in that meeting or, you know, what stakeholder is against this project and where there's risk. And there's so much that is in here in your brain that the AI doesn't know, and that's why I think that sort of, you know, that partnership is so important.

Mashhood Ahmed: So there's a old saying just come to my mind, devil in detail, right? So you'll always find it looks like a finished product, but there's a devil in detail that we need to look. Other thing is as a project manager, you know, you have the meeting transcription on that can record the meeting transcription, what was said, but you know as a project manager that the stakeholder who said yes, but he really mean maybe.

He's not fully buying your idea or whatever the situation on this, that particular risk or the project is. You know that, you know, this director will sign on the project charter or this document or the UAT testing or go live document, but that's not the person that he will come and do or she will come and do everything with you.

There's another person With a title of analyst or something like that, that he's gonna listen. So that is your indirect stakeholder, and then you need to get his or her buy-in, not the director title. And that's where it's all comes from the experience. It all comes from your critical thinking skills.

We'll talk about why critical thinking skills is im- important today as well, because I think critical thinking is more important than ever before. You understand the politics, you understand the people, you understand the dynamics. What works in your organization may not work in other organization. We have the PMOs, we have the PM templates, we have PMI providing so much of guidelines.

Not a single project that I have done in my life that can follow one methodology or one thing- Right ... because it's always unique to the situation, and that's where human comes in picture to say, "Yes, this is applicable in my situation. I understand my stakeholder. I understand my sponsors. I understand my team.

I understand my risk," right? You know, risk that is buried in your risk register somewhere number f- line number 14 or 15, but that's- ... the one that could kick in. So that's all comes from the human insight.

Galen Low: I love that. Even just that little section there, right, I think is what a lot of project managers need to hear, and I'm trying to come at this from the angle of okay, what if you're leading teams of project managers, a role that is already political, already complex, already, you know, imbued with imposter syndrome.

A lot of folks have fallen into that role accidentally. They don't feel like they're doing it right because they're not following the textbook 100%, and it just seems like they're going with their gut, and they're winging it. But exactly what you just said is that is what's valuable. That is the additional context that is needed.

That is your specialty, and AI is the tool that you use to partner with, to collaborate with, you know, and to wield your knowledge in a way that, you know, also makes you more productive I guess arguably more productive. Well, let's get into it actually, because, you know, I think the talk that I saw you give was about the sort of yin and yang relationship between inquisitiveness and strategic interrogation you know, like the critical thinking, and that's kind of the key to wielding AI effectively.

I thought maybe we could start off with the inquisitive bit. What does it mean to you to have an inquisitive mindset? And again, why is it important when it comes to AI?

Mashhood Ahmed: So remember you know, we used to hear the term like data is power, knowledge is power, who has the most data, you know, would be ahead of the technology and everything.

I think opposite is, is true right now. The more data and the more knowledge you may have, it could be a curse. You need to have a relevant data, relevant information that will makes it valuable. Now, the other thing is that I can find tons of data about anything and everything available on the internet, and now generative AI has crawled all the data, so you can ask it to share this specific data, that specific data, so that is available at your fingertips.

And there's lots of noise. You write a three, four line of prompt. Now you have a five-page document that you need to read and make sense out of it and pick up the points, right? That really matters. So there are lots of noise here, so understanding that the noise is a input. And the strate- strategic interrogation, what I call, is something that you ask the question.

In simple word, what is the most relevant question I need to ask AI to help me with? And I need to have a clear understanding of why it is important. Because if you don't ask the right question, AI will agree with you. It will give you the yes answer. You need to do a personal switching. Okay, now look from this personal perspective.

Now, what will happen in this situation? So what if scenarios. And again, putting all those what if scenarios, it takes lots of cognitive power. I can tell you from my experience, like two or three hours of working with any generative AI in cognitive sense is so much work that within two, three hours, I'm tired.

Yeah. Literally, I'm tired. Like two hours of continuous working with any generative AI on some specific thing, I get tired and my mind is like overloading. I need a break. Good. Take a break. Keep asking the question because if you don't ask the right question, you'll get yes answer. That's one thing.

Other thing I want you to think about you know, as a project manager, we have dealt with like CIOs, CFOs, CEOs, directors, VPs, and all those people in our careers. And sometime, you know, I wonder like how that person get to that level, right? And one of the thing all of those leadership people have common is when they come to me, they ask me one or two question that really matters, that really digs down deep.

And if I answer those questions correctly, boom, I gain my respect with them, right? But if I don't, they're like, "You know what? I'm not gonna go to Mashhood. He's not my favorite project manager or the PM manager or the director." And these people rise in their career because they ask the right question at right point in time.

And it becomes more critical now to ask the right question because AI can do a lot for you. If you set your AI agents correctly, it can do a lot of work for you in the background But now again, the context is, are you asking the right question that really matters at this point in time, and what you gonna do next?

What you gonna do with this, this information, okay? Oh, I received this report, that report. Okay, after reading that report, what you gonna do? Are you gonna change something? Great. If you are like, "Oh, this report good. This report show me this number," okay, great. But what's changing? And those are the real questions that will go deep down, and I want all of you who are listening to this podcast to pause here and say, "What are the deep questions I need to ask in my next meeting that really cuts through?"

And should I cancel this meeting if I'm not asking those deep down questions? Then just cancel the meeting. People hate meetings.

Galen Low: I really like that. Actually, I wonder, you know, could we, you know, as people un-pause from their moment of reflection, can we walk through a bit of a scenario of that? Because I like this notion of asking better questions.

I really like the idea of the fact that a lot of, you know, people that we see as successful or who have achieved a high level of seniority in an organization are the people with, you know, what we perceive as the instinct to ask the right question at the right time. You were talking about, you know, engaging with an LLM, sometimes having a conversation for, you know, an hour or two or three, and maybe not everyone listening is sitting down with their LLM in that way.

But even, you know, like a, like a sort of 30, 45-minute conversation with your LLM to sort of ask the right questions, can we just step through a scenario of what that might look like and what you might ask?

Mashhood Ahmed: Questions I would ask is, let's say I'm working on a new project, right? So I would give it access to my documents, right?

Whether I have a Copilot in Office. By the way, lots of companies nowadays have adopted Copilot, but most of them are not using effectively and efficiently, right? So be the one in your team to do it. Share your document, set the boundaries saying, "Only these documents," right? And then start asking question, "What is incomplete?

Where do I need to fill the gap? What are the inconsistencies in these documents? What are something that someone else can question me?" Like persona switching type of questions. You know, we are asking for more budget. What question would CFO ask, right? CFO does not care if your project is on track or not, right?

He or she cares is, is this in bud- within budget? Great. Okay. Your business owner, the business manager, they do care if the project goes live on time or not, right? Your CIO, he may care about what's happening with integration. Is, is our data safe and secure, right? So again, everyone has a different perspective, so that persona switching really helps there.

So ask it to do a persona switching. For example, I'm working on a project right now where I'm developing some new content, and then I'm asking generative AI, "What is inconsistenty h- inconsistency here?" Right? What is redundant that I've already covered? Because sometimes we tend to, you know, repeat things, and it's good idea to repeat sometimes, but sometime too much of redundancy is, you know, not too good.

And then just say, "Hey, help me out step by step." So one of the other prompting technique that I use that really helps me is ask generative AI to interview me or ask me questions one by one. And why is that? We need to understand why hallucination happen. Hallucination happens because we do not provide enough context to the AI.

So what AI does is it makes assumptions, and if those assumptions are true, we are getting a great response. If those assumptions are not true, AI will fabricate data, okay, make some assumptions that are not true, and give you the output, and you're like, "This is absolutely incorrect." Okay? So what we need to do is we need to train AI just like we would train an intern.

So think of AI as somebody who joined your team, who has an MBA or recently got PMP exam. Very smart, very good, but makes some rookie mistakes. Okay? Always it will make rookie mistake. So put up a prompt "Ask me clarifying questions one by one." What will happen is that any assumption that AI is making in the background, it will verify with you.

And why I say one by one, because let's say I say, "Ask me clarifying question." It will ask me four question in one go. I reply to one, and it will assume that I reply to all four. But if I forces it to do, ask me one by one and wait for the response, give you some time to reply back, and then you will see the quality of your replies and quality in the value that you receive from AI is much more different, much more valuable that you can leverage and help you in that type of area.

The other thing in that is setting up the constraints or the guardrails. So think about guardrails as as a human, we are really, really good in what to do. Okay? And we are really, really bad what not doing. Okay? So we understand that, hey, doom scrolling is not good for us, but most of us are spending a lot of time, a lot more time than we should be on doom scrolling.

Okay? Because it's easy to do. It's convenient. So it's like not-to-do list, right? But again, most of us don't have a not-to-do list. Now, in the age of AI, it's very important to have a not-to-do list and put that not-to-do list as a guardrail within your agents or within your prompts so that you can verify and double-check and ask those specifically because anything you do not ask AI to do, AI tends to do it.

So it's better to put the guardrails. Other thing I was talking about like a constraint stacking. So, you know, I need output in three bullet points. I want output in a table with five of these columns. I want response in 50 words, 200 words or 1,000 words, right? I prefer usually smaller response and then kind of a build upon it, right?

Within Copilot, there's a function or feature where you can say, "Make these changes to my allow editing." That means Copilot can go and edit your documents. So for example, in my PowerPoint presentation or Word documents, I say, "Hey-" Check this and fix this screen layout or this thing should be here. So I don't have to do like manually now, right?

I can simply say it takes a little bit more time, but it does a good job. Be very specific if you're asking Copilot to, to edit your document. Don't give it like a blank check to edit all your document because you do not know what it's gonna change. And remember this, when you submit any report, your name on it, on that, right?

By Mashhood Ahmed, whatever my title is in that organization. And when you receive a response from generative AI, it says, "Generated by AI, review this for this and this. AI can make mistake." You cannot make the same disclaimer on a report that has your name, that- ... "Developed by Mashhood, but generated by AI.

Check for correctness." Wouldn't that be great? You can't do that because your name at the risk. So that's... So tho- those are the things that we need to understand, that there's a human part. We cannot ignore the human part. Human in the loop will always exist. Even with agentic AI, we still need human in the loop at different gates, different stages to verify one agent is doing the right thing and then follow up, and then follow up from there.

Galen Low: Yeah. Let's get into the strategic interrogation bit. But before I get there, you know, I love that the inquisitiveness and asking the right questions, it is about learning, right? We're good at doing, like you said, right? We're like trying to get a thing done. Sometimes we're trying to get a thing done as quickly as possible because we think that's what AI...

Or at least maybe we feel like the expectation on us is that, oh, now that we have Copilot, now that we have these LLMs, everything should take 15 minutes. But I think what I really like about what you just said is that you described a conversation that took time to ask questions that might not be apparent, to put guardrails in there that are like the not-to-dos, to ask questions about things that are missing or that are, you know, contradictory.

Like, all of these things are things that, A as a project manager, a lot of project professionals get good at this, right? And when I was leading a team, this is something I really focused on, is don't feedback on what's there. Feedback on what's missing as well, right? And again, humans aren't good at that.

We're not good at seeing what's not there. We're immediately gonna go to, you know, what's there and, you know, provide feedback on it. And if we don't find anything wrong, we're gonna be like, "Oh, that's like a misplaced modifier," or you know, hanging Oxford comma, or we're gonna get like really nitty-gritty 'cause we wanna add value.

But actually, it's the zooming out and being like, "Okay, well, first of all, when I'm briefing you, don't do these things because these are our constraints." I like what you said about building up towards something, 'cause you're right. Sometimes you get, you know, seven pages of output from your LLM, and you spend the rest of the day reading it picking it apart, versus just working together in smaller chunks and then baking that into the file.

And then I just really like that idea of using the opportunity, not necessarily to work faster as in get your tasks done faster, but to use the extra time to learn a bit more and to think a bit more about your context. Maybe just one question before we move on. Do you tell your LLM like about some of those nuances like you mentioned earlier?

This stakeholder will say yes, but then they're gonna... The real decision-maker is actually over here.

Mashhood Ahmed: Yes. So, you know, in PMP prep, we have been taught how to have a influence matrix, right? One other thing that I suggest everyone to build on their project is influence metrics grade, but create the persona of each stakeholder.

So literally you can have an agent that can read through all your meeting transcription and say, "Hey, what do I know about this person who is joining my meetings?" And then you review it and then you can say, "Yes, remember this," or, "Don't remember this. This is true or this is not true." That will help because, you know, AI does something good for you, and it will help understand the context.

It will also help you once you start doing the persona switching, right? So you can say, "This is my sponsor. Read this meeting, meeting transcription, and tell me what personality type he or she has." So one of my other talk idea that I need to do is, is a DISC profiling. It's a concept by John Maxwell where you do a DISC profiling about your stakeholders, and the idea is to do the DISC profiling using generative AI for your stakeholders, and then keep using that over the period of time, and you will see AI get better at that.

Galen Low: I really like that. It's a neat idea. Again, right, it's you know, we need to be providing the context in order for it to be giving a good assessment. But also, like I've never been the kind of person to sit around and read through every meeting transcript and, you know, build my own DISC profiles of every stakeholder in my stakeholder like, you know, ecosystem.

Like abso- it seems like it's gonna take too much time. But that's where AI can maybe do a bit of work to help inform some of the decisions that it makes, and maybe in turn influencing some of the, the decisions that you make as well. Which actually might be a really good segue into the strategic interrogation part.

You know, we've touched on it already a little bit, right? Just kind of the fact that as the human, you are going to be accountable. You're gonna be held accountable. You can't just go out there and be like, "Sorry, my work may have mistakes, but please continue to pay me." For some reason that just doesn't work.

So what does it mean to you to be strategic in your interrogation of AI outputs and, like, how does this balance with that sort of inquisitive mindset?

Mashhood Ahmed: It balances very well because strategic means... What is strategy means, what's one thing that I need to focus on, right? To me, strategy is one thing and removing all the noise, okay?

And typically, if you look at any strategy document, it usually have three or five things because that's what our human memory is, and we can kind of tag along. So think about what is really important in this context, right? What is not important in this context, and what is a noise? Having that understanding will add real value because now AI can do the grunt work for you, but you are picking up the things that are, "Okay, this is valuable.

This is worth putting in a document. This is worth putting in the AI knowledge bank. This is worth putting bank in my stakeholder persona. This is worth putting or help me prepare on this because I don't want to put this on a presentation because I, if I put that word on presentation, I know my presentation will go this way.

Okay, I don't want to put it, but I know I have this person coming in the meeting. There's a very good chance that he or she will ask me that question. Help me prepare." Okay, that's where strategic interrogation and inquisitive mindset comes in picture, okay? Asking the right question, what matters, what to put on the slide deck, what not to put on the slide deck.

That's your decision because you know your stakeholders. You can read the room Remember, AI does not sit in front of the board or in front of the PMO or in front of your stakeholder and answer any question, right? You are the one because you are in the driver's seat. So all of those things matter because if you ask the right question, and if you have a strategically what is the right question on this project versus on that project.

So two different projects, you will be asking two different questions or many, many different question, but knowing what to ask, knowing when to ask, that is critical, and that's one of the most important critical skill set I believe we need to develop because AI can do anything and everything, okay? But are you asking it to do the right thing for you at this point in time?

Galen Low: You know, it's funny because I could go back through this conversation and take AI out and replace it with, you know, direct reports, and all of this would probably ring true, and it's what's funny about it is that we're investing all this time on, you know, giving context and challenging the output and providing feedback to machines that all along the way that probably would've helped us be better leaders of humans, too.

I hate sometimes the-- my prompt might be more detailed than a brief I would give a human because I'm just assuming that, you know, this human will figure it out or, you know, go get the context themselves, whereas actually, you know, we could do a better job of educating our people like we're educating our LLMs on things like specific project nuances, for example, on the politics and, you know, what's important strategically and what's not.

You help a lot of organizations sort of take this step, using their ability to be inquisitive and using their ability to interrogate outputs and be strategic. And, you know, you just mentioned something, you know, really important, which is that sometimes from that output, whether human or AI, frankly, you're gonna be the one making the call of what you're putting in front of your stakeholders, what you're presenting to the board.

I guess my first question is, are you providing that feedback to your LLM to be like, "Cool. Thanks for that. I'm not including this because of this reason. I'm just gonna go do the slide deck now," or are you kinda just moving on you know, that's your judgment call to make. You don't need to train it to think a different way?

And then maybe after that, maybe we could just step through, like, how you get people into this mindset, some of the, like, symptoms of people who are not using this combination of asking the right questions and challenging the outputs and how you get them to a point where they do know what to ask.

Arguably, those are two very different questions, but anyways.

Mashhood Ahmed: So to answer your first question, whether you provide feedback to generative AI on the responses, right? It is very important. So one thing I would like all of the people listening to this podcast is to check what is a system prompt, okay? Or what is a custom instructions and what is memory.

I'm not gonna explain all of them, but, you know- Search it, ask any of the AI how to do it, what it is, and what is the difference between these three, custom instructions, system prompts, and memories, right? And then you can update those memories and custom instructions in the Outlook, in the Copilot, in ChatGPT or Claude, right?

And that will help you set those boundaries. And you can set up some rules with those memories and custom instruction. If these type of things happen, then do this type of thing, and then you need to enforce it. Even if you set up everything, memory, custom instruction, system prompts or others, it will still make the mistakes, so you still need to have an eye.

And other skill set is building that eye to catch up these problems. It's a skill that we need to build more, right? Your second question was about people who are still at the fence, right? That always happen, you know. It happened 25, 30 years back when internet was introduced, right? People were like, "Oh, I know everything in my warehouse where it is," right?

Yes, you know it, but if you're on vacation, if you're on holidays, your people cannot work. How can you do it in a way that it, you make it foolproof? And the whole idea was to have the online database, right? And now we buy it. We are like, "Okay, there's no question about it." Other thing that, you know, question comes is, is it gonna impact my job?

Is my job at risk? Okay. I would say yes and no. We don't have a definitive answer whether your job is at risk, depending on the context. But what we know from history is that profession changes. New professions comes. New tools come. New technology comes throughout the history, right? Look back 25 years back.

Did anybody heard a term called cloud engineer or virtualization engineer or desktop engineers or things like that? No. We'd never heard about those things, right? The type of the projects we did for example, 15 years back, I did a project to set up a data center in a office building. Okay. 15, 20 years back.

Now we don't do those types of projects. What we do is we're like, "Oh, you need a data center? What server you need? How many servers you need? Go on Amazon Web Services or go on Azure Cloud or go on Google GCP." Build those thing, right? And if you know what you need to build, you can literally build your data warehouse or your data center or your servers within a matter of hours or a few days compared to six month, fifteen years back, sixteen years back, right?

So this is important to understand, have that mindset that the type of projects we are doing today are not the type of projects we'll do tomorrow. So the question is: How can I prepare myself for those types of project? The only logical answer is to learn the latest tool and be aware of those gaps. Be aware of your personal blind spots and be aware of the blind spots of your tool.

The AI model will keep changing. They will get more smarter. They will get more good at something, and they will get worse in some other areas. And there's a technical term called like a prompt dr- drifting or model drifting. That will happen, so you always need to have that eye for the QA, and that's something we need to build more than ever before because there's lots of noise, you know.

Five-line prompt, seven-page document, now read it. Takes twenty-five, thirty minutes at least if it's done correctly. If it's not done correctly, it's gonna take whole day to edit that document, right? So be aware of those things and be aware of those that is gonna take your time and it still does require human insight because we are the one in front of the board.

We are the one who are managing the people. We understand the complexities of how human brain thinks. Even AI can do everything and anything, but still, you know, and different cultures have s- few things different, right? And again, there are common things in different cultures as well, right? So again, managing with people, frictions, and all those things will become more important and will require a different mindset.

That's definitely for sure.

Galen Low: I really like that. Can I put you on the spot for maybe a bit of a story, maybe end to end? What I'm interested in is you know, we started at the top about sort of, you know, generic questions producing generic outputs, and then sort of getting into an inquisitive mindset and strategic interrogation, and then getting, you know, a better output from that.

I'm just wondering, you know, what are some of the symptoms that you've seen when you're working with teams that are sort of, you know, at the beginning using AI pretty generically, kinda just pasting that in, running with it, you know, feeling the fire when it's not right? Like, how do you identify that they might need to move ahead?

And then how do you get them into that inquisitive mindset, strategic interrogation space, and what's the difference afterwards? What's the result?

Mashhood Ahmed: So in my experience, it's showing them something very simple, right? So example, in the beginning of this podcast, I would like all of you to run the dice experiment, understanding that, that generic inputs gives us the generic output.

And number of times, and I've showed this one to people, there's a one time somebody said, "Oh, it's a magic. You, you kind of, hacked the system," right? Right. I'm not hacking your generative AI. I can say that, "Hey, I'm hacking your generative AI," but in reality it's just statistics and mathematics.

Understand, AI works on the patterns. Now, how do we bring the team level up? There are some team members who are ahead of the curve, right? They are your internal AI champions. Bring them in. Let them, you know, run some workshops and, you know, some Tech Tuesdays or you know, lunch and learn sessions, those kind of things to discuss a small thing, right?

Nothing big. And then just go slowly. So in one of my current consulting gen AI enablement, I'm working for a university in US, and what we are doing is we are just providing them the basics of AI. Basic prompting, basic stuff, and I receive a feedback you know, 70% of the people report higher confidence than before, right?

So we went from 19% to 70% plus With just one two-hour session Wow. 90% of people are likely to recommend this type of training to a colleague because again, you need to get to their level, okay? You need to think about Suzy in HR, okay, who's not tech-savvy. You need to think about Dan in procurement who's not tech-savvy, but he really understand the vendor management.

He understand how to deal with them. So you need to think from their level. Give them simple examples. Don't give them complicated example. Don't try to run like end-to-end thing from day one. Do small thing. Let them experiment. Give them a month. In next Tech Tuesday or next Lunch and Learn, show them something different, okay?

Get their feedback. Feedback is very important. So what we have built in there for this client is we build like a feedback with just simple survey, okay? "How do you feel confident after this session?" Okay? Simple question. Get a check. "What else do you want to learn next?" Okay. And then you are getting the feedback.

Also, if you are on the system admin side, technical, if you, you know, director IT or CIO, you know, check your Copilot usage, see the pattern, okay? And then map it. Every month are we seeing more tokens usage? And again, something to pay attention to is how much does it cost to you.

Galen Low: Right.

Mashhood Ahmed: 'Cause you will be surprised that, you know, if people are start using it very, very aggressively, you know, you're gonna reach your limit, right?

So again, think about that context. Think about the governance. Think about the policies. Think about, as I said, not to-do list, what could go wrong in terms of the policy and the strategy of your organization There are so many news stories about where AI did something that it should not be doing. The only reason is it's because there are not proper guardrails were set, right?

One of the example that keep coming to my mind, it was two years back, but still relevant. Air Canada, right? And one of the customer went to Air Canada website saying, "Hey, I have... My grandmother passed away. I need to go emergency. Do you guys give a bereavement discount?" And the Air Canada chatbot thought, "Oh, that's a good idea to give the bereavement discount to our customers."

We have the loyal customer. Great thinking, and said, "Yes, we will give you whatever percentage of bereavement discount." The guy took the screenshots, and he come back, he applied, and they said, "We don't have a policy." He took the screenshots, challenged the Air Canada. They did not accepted it. He went to some court, and they said, "No, Air Canada need to pay that discount."

Again, that discount could be few hundred bucks, but that becomes a story that people are talking still two years it's, it's happened. The reason is they failed to set the proper guardrails. They failed to have the proper governance. They failed to have proper policies for humans and even the agents to use.

So really important to think before jumping in, think about what can go wrong if we open this door. And then slowly, do it slowly, gradually, and see how your team is doing. Every team is different. Every business is different. Every business problem is unique. So, you know, think about those things and then start small.

Galen Low: I like that. I like the sort of finding of the balance. And also, I have to say, I really like what you're doing, you know, in the training with that university. We started out talking about imposter syndrome and what's the opposite of what we're seeing in our LinkedIn feeds. Our LinkedIn feeds are like, "This person is doing it better than the other person.

These people know more than the other person. You know, this is wrong. This is gonna fail." You bring everyone together, and you do the learning together, and you share together, and everyone gets more comfortable because they don't have that side of the imposter syndrome anymore. They know where everyone is at, and they're leveling themselves up where they need to.

And, you know, to your point, the inquisitive mindset and the strategic interrogation that's us learning as well. It takes some time, but it's building our confidence, and that's what's giving us the ability to not just resist asking generic questions but also resist just taking that polished output from your LLM and going, "I guess this is good.

I'm not even gonna look at it. It's gonna take me 25 minutes to go through this. I'm just gonna ship it," versus, "Wow, we can get a lot more high-quality output that we're happy to put our name on that shows that we are good at our job and that AI, while it's really good, is not good at everything." I like that thing you said where some people think you're doing a magic trick, and other people are just think that AI can do anything, you know, solve any problem.

The reality is humans are good at stuff, AI is good at stuff, AI is bad at some stuff, and humans are bad at some stuff. And I think, you know, that's the, the sort of, you know, the balance that we need to strike as well.

Mashhood Ahmed: And one thing it reminds me that in my-- if you go back and look at my, you know, few years earlier talk about AI, one thing I was mentioning in those is the important skill set is unlearning and relearning.

So think about that, you know. We are used to of... And again, we heard this in Lean Six Sigma, other thing like, oh, people keep the status quo, or we've been doing this twenty years. Yes, you are doing it twenty years, but that's not the effective and most efficient way to do it. It gets you the result. That's a change management.

But underneath it, it's the very important skill is unlearning. And I want all of us to be aware of that, hey, I need to unlearn this skill. This is not gonna help. So what helped me ten years, fifteen years back is not gonna help me now. And then come up with a strategy or plan what are the different skills I need to learn.

And you need to be open for that, right? There are some people open to it, great. Other peoples, they are not open to it, perfectly fine, you know. Again, everybody's different. Everybody's at a different level, different stage in their life, you know. Twenty-five years back, my dad is an architect, right? So AutoCAD came in and everybody was started using AutoCAD.

My dad simply refused to use AutoCAD. He's like, "I'm not gonna touch it." What he did, he hired someone to help him with AutoCAD drawings and everything, and he would sign, right? Okay, great. It worked in that time. Nothing similar will happen here. Some people are like, "Oh, you know what? I'm gonna try. I'm not gonna try."

Perfectly fine. But again, important thing is to understand what do we need to unlearn and what do we need to learn. This is very, very difficult To unlearn your old habits. Again, read some books on habit, how to build a new habit. It's easier said than done. So unlearning and relearning, there are a lot of depth to those two words.

Galen Low: I really like that. You know, throughout this whole conversation, what I really like is some of it is about negative space, like the way we might think of it. It's not what's there, but what's missing. It's not, you know, what did we ask? It's what didn't we ask? It's not, you know, what do we know? It's like, what are we assuming?

And it's not just learning, it's also what do we need to unlearn in order to learn more in a different area? And I think, you know, you raised some really good ones, right? Critical thinking, I think is, you know, when we talk about skill atrophy, that's what a lot of people are talking about. They're like, "Okay, well, you know, AI and maybe the media is reducing, you know, the emphasis on being able to think critically.

We're losing some of these things." And in a way, sometimes with if our AI tools, if our LLMs have the answer, then, you know, we feel like we might not need to be curious or inquisitive or learn the thing because they've got it all figured out. Whereas when you flip it, you're like, "Actually, this is our opportunity to learn more, to think more critically, to be those people, the senior leaders in the room, and to be the people like your dad, right?

Who are like have the eye, right?" You said the eye to just look at something, look at it critically, you know, interrogate it to the point where, you know, you can sign off on it. And that's a big deal in architecture. It's a big deal in project management as well. It's a big deal in the working world. You know, being accountable for decisions that were made from the people working for you is the way we need to look at it, not necessarily spending all our time learning about AI so we don't have to learn anything else.

Mashhood Ahmed: Yeah. So that remind me of something that I talked in the risk management of AI, is over-reliance on AI And what it is doing, it's kind of a, you know, damaging our ability to build our cognitive sense, to think, you know, problem-solving. These are like a fundamental thing, right? You know, no MBA can teach you this, right?

This is something that needs to be... We need to teach our kids from the very, very beginning before even they start the kindergarten, right? How to think critically, how to solve the problem, how to be ava- be aware of your surroundings, right? Again, nothing to do with technology, but it's all about human engagement and over-reliance on AI has become a bigger issue nowadays because people are putting their reports or putting you know, LinkedIn posts and things like that.

And hey, I have made that mistake a few times where I put something and shoot, I forget to remove those two lines, right?

Galen Low: Right, right.

Mashhood Ahmed: Again, it will happen. But being aware of that, I think the critical thing is making that mistake is okay to me, but being aware and then making the correction is more important that not being aware and not, not correction because my reputation is at risk if every single LinkedIn post has some mistakes like that, you know, my reputation is risked.

So I need to think about what value I bring to the table that no AI can do.

Galen Low: I love that, yeah. What value do we bring to the table? Real quick, can I put you on the spot? What's one thing that you've unlearned?

Mashhood Ahmed: Be okay that AI can do things that I used to do. Be okay that AI can do things better than me.

English is not my first language. I used to make tons of mistake. I had a Grammarly subscription a few years back that I canceled now, and Be okay with that, right? You know, understand your limitations and don't try to over-fix it. You know, I can take all these classes, and again, you talked about Ox- Oxford comma, and yes, that's a thing, right?

You know, where do you place the comma? I had some people like, "Oh, you have to put comma here, you have to put the full stop." Okay, fine. Pick one, and then just follow. And again, you can ask AI, "Here's my writing style. Follow this writing style. I would prefer Oxford comma or not Oxford comma," and then it will follow.

You still need to pay attention there. So being okay with those uncomfortable things and just proceeding one step at a time, learning, unlearning, relearning, asking the critical questions all the way, and what is relevant in my case, you know. Those are the few things that I would say are the important skills and important takeaways from this talk.

Galen Low: I like that. I'm learning that, you know, you as the human has to be the best at everything.

Mashhood Ahmed: Definitely.

Galen Low: Love that. Mashhood, thanks so much for spending the time with me today. It's been so much fun. For people who are interested in what you do, where can they learn more about you?

Mashhood Ahmed: Oh, they can search my name on LinkedIn.

You'll find me, connect with me. I do some LinkedIn posts that you will find interesting. You can check my website with the, which is mashhood.net, okay? LinkedIn is the best way to follow me. If you're on TikTok, I do put some content on TikTok, but not on regular. Follow me on the TikTok if you can find me.

And yeah, check some of my videos from different events. I'm posting some one, two minutes clips from different podcasts, different conferences that, you know, you guys can find a bite size, you know, things that you could do differently.

Galen Low: Awesome. Love that. I will include the links to your LinkedIn and your TikTok and your website in the show notes for folks who are interested.

And just wanna say thanks again. This was a lovely conversation.

Mashhood Ahmed: Awesome. Wonderful. Thank you much, and looking forward to this podcast to be released soon.

Galen Low: All right, folks. That's it for today's episode of the Digital Project Manager Podcast. If you enjoyed this conversation, make sure to subscribe wherever you're listening. And if you want even more tactical insights, case studies, and playbooks, create a free account with us at thedigitalprojectmanager.com.

Until next time, thanks for listening.