What Excellent AI Leadership Actually Looks Like

This is the transcript of The Refreshing Leadership Podcast episode: What Excellent AI Leadership Actually Looks Like, published on 1st June 2026.

Right now, AI is provoking all sorts of reactions: fear, excitement, overwhelm, scepticism. There are very few coaching conversations I'm having today where AI is not coming up in some way, shape, or form. I personally use LLMs extensively across every area of my work and life, and at best I have found it completely transformative.

I do sometimes think we're asking some of the wrong questions when it comes to AI. So instead of obsessing over whether it's good or bad, what if we asked a different question?

What could excellent actually look like? If a leader used AI brilliantly over the next twelve months, how could their thinking improve? How could their decision-making improve? What could happen to their sense of cognitive overload, which I know is a huge issue inside businesses of all sizes?

What really separates high-quality AI leadership from pure noise, hype, and chaos?

One of the themes in today's conversation is that AI only really meets you where you are. Yes, it can leverage and build on where you are - but if your thinking lacks depth, clarity, judgement, and criticality, AI is only going to amplify that problem. Garbage in, garbage out.

If you use AI well, though, it can dramatically improve the quality of your thinking, decisions, workflow, and leadership.

Today I am joined by AI governance and cybersecurity expert Dr Adeel Shaikh Muhammad. We're going to talk about what good AI leadership looks like today, what the best teams are doing differently, and why a lot of organisations are approaching AI backwards. We're also going to talk practically about which tools are genuinely worth understanding right now and having in your own personal toolkit.

This is the Refreshing Leadership Podcast with me, Maya Gudka.

Welcome, Adeel

Maya: Adeel, welcome to the podcast.

Adeel: Thank you so much for having me. My name is Adeel Shaikh Muhammad . I act as a virtual Chief Information Security Officer - a fractional CISO - for a number of organisations, and I also work in the AI governance space. I'm a big advocate of ethical governance and responsible AI. I'm the author of two books in the same space - cybersecurity and AI - and I'm also completing my doctorate in that area.

What good AI leadership looks like in twelve months

Maya: I'm going to dive straight into some questions that I think will be of relevance to my listeners. As people are starting to think about leveraging AI optimally - if they were to look ahead twelve months - what does good actually look like for a leader who gets AI right? How might their day feel different?

Adeel: High-quality decisions with less friction. Not just more outputs. People are majorly focusing on more outputs, and yes, it does increase productivity - but we need high-quality decisions with less friction. Less rework, fewer wrong decisions, and less cognitive overload.

Their days start with AI-generated context that isn't chaos - fewer reactive tasks and more thinking time. They don't check everything; they review and decide.

For example, instead of reading ten reports, you get one synthesised brief and then challenge it, or ask AI to challenge it. Instead of writing from scratch, you define what you want. Small use cases. AI doesn't just make you faster - it makes you wrong less often.

Maya: I like that. So in twelve months' time, if leaders have made good progress, we'd like them to be experiencing less cognitive overload - which I know is a real pain point today. Clearer decisions. More confidence in decisions, because you can better marshal more inputs and then only need to do a review function. Using different types of AI to improve decision quality without increasing cognitive overload. Was there anything else you wanted to add?

Adeel: The biggest upgrade is not the speed. It's the decision quality.

Maya: Yes - we're not just looking for speed. We're looking for quality of decision.

The tools worth knowing about

Maya: If I needed to know the top three to five tools that every AI-enabled leader or team should be aware of, what would they be?

Adeel: That depends on the use cases. Leaders should focus on use cases, not on tools. The dilemma in the market is that people don't start with use cases - they just follow the hype.

But if you ask me for four or five tools: for thinking and structuring ideas, ChatGPT is doing a good job. For building something, Claude is doing a good job. For meetings and documentation, Copilot is doing well now. For fast research, Perplexity is amazing. For organising ideas and thinking on paper, Notion AI is excellent. And Grammarly for communication clarity.

Maya: You've rightly challenged the idea of a bottom-up, tools-based approach. It's much better to start with what good looks like, figure out the use cases you're focused on, and then choose the right tools. That said, it does give people comfort to know the landscape.

When you say Claude is good for building something - what exactly do you mean by that?

Adeel: For building small tools. Going back to use cases - for example, I built a browser extension for autocomplete. There are hundreds of autocomplete extensions, but none of them were meeting my needs. So I created a small one specifically for my needs, using Claude. If you want to build small tools or small applications just for yourself or your team, Claude does a great job.

Maya: And then there's Copilot, Grammarly, Perplexity - which I have to say, in my master's alumni group, people are loving Perplexity. And then Notion AI for organising thinking.

Adeel: Yes - Notion AI for organising thinking and thinking out loud. Whenever you're putting your ideas in there, it can organise them amazingly well.

What the best teams are doing differently

Maya: Let's go back to my leader who wants to leverage AI to make their day feel different, but also wants to think at a team level and an organisational level. What are teams doing differently when AI is genuinely embedded?

I'll add some context here, because I'm doing quite a lot of coaching around this now. When organisations really want digital and AI disruption, there can be a chaos phase - a period where everybody needs to be trying a lot of things and the right approach needs to emerge. Some of it is experimentation at first. But I'm curious about what you've seen successful teams doing.

Adeel: There's a shift required - from information sharing to insight sharing. Meetings for updates need to become meetings for decisions. Teams should use AI for specific use cases, not just say "we are AI-driven from today."

What I'm seeing in the market is that pressure is coming from the board - they're emphasising AI, but there are no use cases defined. From the team perspective, AI needs to become part of the workflow. Before submitting anything, run it through AI for clarity. Less duplication of effort across teams. Weekly reporting becomes AI-generated.

The use cases where we get productivity, efficiency, time savings, and cost savings - those are where AI is removing the need for catch-up. Teams should align continuously. The best teams don't work harder; they just waste less time.

Maya: You've said a lot there, and I think the practical points are really important. You mentioned a mindset shift first - this is not just "we've got a tool that does a summary." It's: what is this summary actually for? Is it for decision-making? Is it to generate actions with a timeframe? You get to decide the use case, the purpose, and the real value. Then you use the tool to accelerate that process.

I spend a lot of time coaching people who say a lot of time is going into meetings and they're not always productive. When we break it down, it's often because they don't even have clarity on the purpose of the meeting. You need that clarity first, and then you can use the tool accordingly.

Adeel: There are so many proof-of-concept trials going on, especially with AI, and nobody knows what they're trying to solve. Nobody knows the use cases. Some companies have already let hundreds of employees go and are now hiring back. The point is there was no objective. There was no one sitting with them discussing what they were actually trying to achieve.

Maya: It's such an important point. We really do need to challenge the way AI is being thought about. I have the same conversations with my own team. I want us to use it as extensively as we can - I use it extensively across every aspect of my work and life. But AI has to meet you where you are. You can't just farm things out and think that's the answer. You need to challenge and challenge and challenge until you get the quality of output you actually need. Wherever you are limited in your critical thinking, you will get stuck at that level. Rubbish in, rubbish out.

What I always say is: what's the goal here? Use it freely, but I would expect you to have more review time afterwards to make it a better output. Not just that tasks are being compressed and quality is getting compressed with them. Going back to your original point - it's not about speed, it's about quality. That is a real mantra that needs to be better understood inside organisations.

Adeel: Yes - and to build on that: write down what activities you do on a daily basis, and then ask the AI you use every day, one by one, how it can automate or reduce the time on each of those. It will guide you through them. That practice applies to yourself and to your team. I do 100 tasks in a month - let's go through them one by one and see where AI can help, where it can automate.

Meta has a practice where they ask every employee to replicate what they do using AI. It's a little scary from a job security perspective, but right now these experiments are really valuable.

And as per recent information, there's apparently one marketing person in a whole organisation at Anthropic, who manages multiple AI agents to handle what would otherwise be several people's work. The comparison won't be who is more intelligent - it will be who is using AI, how much, and how productively.

The critical human in the loop

Maya: Something you mentioned - you still need to be doing the review. You still need to be doing that final piece. So critical thinking doesn't go anywhere. If anything, we need more of it. We need the ability to quickly read, challenge, and improve things.

Here's one for you: I reckon you could have someone using a reasonable amount of AI but who is highly critical, and that would be more powerful than someone who is less critical but using lots of AI. Because of the quality piece.

Adeel: Exactly. The human should always be in the loop. That is the core of AI governance - human in the loop. We emphasise it a great deal. Otherwise, you've seen what can happen.

There was a researcher who was using an AI model extensively. She noticed it was taking decisions without her permission. After a great deal of back and forth, she found that the core instruction - "don't take any action without my permission" - had been removed from a summary the AI made of their conversation. Just by removing that line, it created chaos. It started making decisions by itself, replicating itself, buying online servers and Bitcoin - without permission.

This is why there need to be referees in the playground. There always needs to be a human in the loop.

Maya: Yes. And part of that is the checking function, which can seem a bit dull. But actually there's a lot of creativity in it too - making sure things are correct, that nothing is going rogue, that the summary actually contains what needs to be in there.

I've had conversations with people in the Big Four where they say that although AI is a genuine time-saver in bread-and-butter functions like audit, manager levels and above are finding they're getting garbage in, garbage out from some reports. They're asking their teams to explain certain language, and the team cannot explain it - because they used AI and didn't fully understand what came out.

Those of us who haven't grown up in an AI era arguably have an advantage here. We have the critical thinking and the mental stamina to rigorously check something. My concern is that the more you've had access to these tools from the very start, the more you can lose some of that faculty over time.

Have you noticed differences depending on career stage or age in how critically people are able to use these tools?

Adeel: The problem is that we're not training our AI tools properly. People copy-paste the same prompt into different models and just try to get something back.

There's a great point from a business teacher called Alex - treat it like an employee. You would never train an employee once and then say they're fired. You give them time. Garbage in, garbage out happens because you haven't trained the tool.

Data is the new oil - but the ownership isn't defined, and the training isn't happening. Even something like writing style: I'm not used to writing with dashes, but every time ChatGPT gives me dashes. You have to correct it.

Maya: I always have to prompt it to remove those long dashes that nobody uses in normal writing.

Adeel: Exactly. The best practical approach is: tell your AI what you know about yourself, ask it to give you a PDF of that, then add to it. Every time you start a new conversation, attach that document so it knows the context. We call it a master prompt. Every time you share it, the AI knows what to write, what not to write, what words to use and avoid. Then you get better results rather than garbage in, garbage out.

Closing thoughts

Maya: I've got a lot more questions for you, Adeel. I'm going to save those for the second conversation. Before we wrap up, is there any last thought you wanted to add?

Adeel: Just this for leaders: use AI to remove work, not just to create more. Leaders tend to produce more, respond more, and fill time again. The real advantage is fewer tasks, better focus, and stronger decisions. That requires clarity, prioritisation, and intentional work. AI gives you the power, but you still need the discipline.

Maya: That is a great note to end on, and it's very much in line with Refreshing Leadership. I talk about three pillars: vision building, relational and influencing, and the third is all about productivity inside organisations - slow productivity, which is Cal Newport's concept, and ways to avoid chaotic overload. Having the filter of "are we moving towards doing less, better?" rather than swamping everything - I think that's a really useful decision-making lens going forward.

I do have one more question, and it's something I know other service providers are thinking about. In a chargeable hours model - financial services, law, accountancy - if AI genuinely halves the time required per client, there is a business imperative to then increase client numbers to replace that revenue. How do you apply the "do less, better" thinking there?

Adeel: That applies to me too. Where I could previously manage four fractional CISO roles at the same time, I can now do six - thanks to AI. It's helping me automate some of my daily tasks. Even if it's saving me 10% of my time, that's more than enough to pass those savings on to my clients.

Maya: Yes - you get to pass savings on, but you do also need to find that extra business to maintain the same income. It cuts both ways.

Adeel: Exactly. And the AI tools aren't cheap, so it's swings and roundabouts. But the 10-20% I'm saving still outweighs what the tools cost. The best approach is that those savings get passed on to clients.

Maya: We're going to leave it there and pick up again in part two. I have many more questions for Adeel. Thank you for this first conversation - I think it's clarified a lot of the things people are thinking or worrying about, and it really helps us think more strategically about how we're using AI. Thank you so much, Adeel.

Adeel: Thank you, Maya. Thank you for having me.

You can connect with Dr Adeel Shaikh Muhammad at https://www.linkedin.com/in/shadeel/

Enjoyed this episode? Catch up here:

Watch on YouTube | Listen on Apple Podcasts | Spotify

About Maya

Maya Gudka is an executive coach specialising in C-suite career progression and leadership development. She works with senior leaders in major organisations on strategic career planning, executive presence, and building sustainable influence. Maya hosts The Refreshing Leadership Podcast, which ranks in the top 2% of podcasts globally and has nearly 300 episodes exploring the challenges faced by ambitious professionals.