Information Matters with Jacqueline Stockwell

061 Bridging the Gap: How to Sit Between Tech, Business, and AI with Duncan Boyne

• Jacqueline Stockwell

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 24:08

Send us Fan Mail

Are your business users actually opening your data dashboards, or are they getting lost in a sea of rows, columns, and overcomplicated metrics?

In this episode, host Jacqueline Stockwell sits down with Power BI and Fabric specialist Duncan Boyne to bridge the "messy middle" between heavy data engineering and real-world business decisions. Duncan shares his unconventional approach to data storytelling—including why he asks clients about their favourite addictive smartphone apps to design better dashboards.

We also dive deep into how data and compliance professionals can stop waiting for official IT corporate rollouts and start using generative AI tools like ChatGPT, Claude, and Gemini safely today to become "dangerously efficient."

Key Takeaways

  • The Golden Rule of Data Storytelling: If a stakeholder looks at a report and has to ask "What does this column mean?" or "Why is that blue?", the dashboard has already failed. True storytelling prompts actions, not formatting questions.
  • Designing for "Doom Scrollers": Why understanding whether your stakeholder scrolls TikTok, reads LinkedIn, or browses Amazon dictates the layout and visual architecture of their reporting.
  • The Power of "Silly Questions": How asking the most basic, seemingly naive questions can break through the "curse of knowledge" and solve massive operational blind spots (like how a simple seasonal temperature shift ruined an engineering company’s product failure metrics).
  • Becoming Dangerously Efficient with AI: How to use LLMs right now for mundane tasks like accessibility testing, color blindness theory verification, or creating date tables, without exposing live or sensitive customer data.
  • Enterprise Guardrails vs. Shadow AI: Why rolling out AI requires the exact same rigid IT planning, testing, and sandboxed UAT environments as a new CRM system.

Links & Resources Mentioned

Support the show

SPEAKER_00

Hello and welcome to today's show. I'm Daphne Stockwell, CEO and founder Leadership Through Data. I inspire and motivate information leaders across the world. Now, today's guest on the show is someone I met in London at Experts Live at Code No. I watched Duncan from the room and I knew instantly that I had to grab him and get him on the show because he told me he was a Power BI specialist, a consultant CPIO, and the founder of the Norfolk Power Platform User Group. And he's an international speaker. And his name is Duncan. Duncan operates in what I would love to call the messy middle. He sits right between the heavy engineering teams and the business stakeholders who are just trying to make a decision. Now he works with businesses running complex systems like Sage and Microsoft 365 Dynamics. Now, taking complete scattered siloed information and turning it into really pretty dashboards that people actually open, understand, and use. And no overcomplated manuals are required. But what I found really useful and what blew me away as his session last week was his take on AI. While the rest of the corporate world is sitting around waiting for official co-pilots to be rolled out by IT. Duncan is out there using tools like chat, G GT, I can never say it, Claude and Gemini today, to make his workflows and his data storytelling what he calls dangerously efficient. And this is exactly what he wants to talk to me about today. So Duncan is here to hand over his personal real-world AI pumped pack. We'll see how we get on with that one. And we're going to look at how to take chaotic stakeholder rambling and turn it into clean data requirements. Now welcome Duncan to the show.

SPEAKER_01

Thank you. Thank you. That is a hell of an introduction. Thank you. I appreciate it.

SPEAKER_00

You're absolutely welcome. So you said that the problem today isn't about the lack of information, but that nobody has made sense of it yet. In the information management world where I sit and my listeners, we focus heavily on structure, life cycle, and security. But if a user opens a dashboard and can't instantly see that story, the data is effectively kind of dead, unreadable really, there's no point seeing it. How do you define data storytelling for an audience that might be used to thinking in rows, columns of retention schedules?

SPEAKER_02

It's a really, really good question because when you get to data storytelling, you are doing it for an audience. Sometimes the audience is broad, sometimes you're doing it for a full business, and every single person in there needs to be able to view it and understand it. And that is really difficult to visualize from a you're trying to make everyone happy. I can do very specific reports that make the finance team happy because they like looking at Excel and they want to continue looking at Excel. You can make really flashy things for the sales team because they like things with glitz and glam. You can make stuff that is very logical and follows a strict regime if you're doing it for an engineering team. For people who already work in data and are used to looking at it from a SQL Server perspective or very Python perspective or very rigid structure, you build it towards those needs and you make it look like a finished product for them. I'm very big on making reporting look like systems you already use. Like you're already bought into your intranet, you're already bought into the the processes and and programs you already want. If you if you are a end user who is in Dynamics 365 all day, I'm gonna make it look like it's part of a system. And it's gonna make sense to you because if you've got to learn something new, if you've got to look at visual architecture in a different way, I'm using your brain power when you don't need to use your brain power. You're supposed to be looking at a report, seeing something that you can make a decision on, or seeing something's going well, and that's the end of it. I'm I'm very big when we talk about data story time. I don't want people to ever have to ask a question of what's in front of them, other than, oh, why are our sales done this much? Why are we not shipping here, or why is it a six-day turnaround? I don't be like, what does that column mean? Why is that bit blue? As soon as people are asking those sorts of questions, you've lost the purpose of that report or dashboard. You're taking away from its usefulness. Like I've never seen someone go onto a computer program, not understand something, and then still be happy with it. Because it's it's not a thing. It's a very it's a very humbling thing when you think you've done very, very well on a dashboard, it's very pretty, you think all the day it's correct, and then you give it to the person that's intended for, and they're like, what even is this? Like, what am I looking at? Because as soon as they say that, you've missed what they're looking for. I had a previous career in banking, and then I did a very, very short stint as a graphical designer where I wanted to do that. I did all of two months of it and then did not pursue it. But one thing I learned from that, and one of the very first things they were taught was the brief from a customer is a square on a bit of paper. They're then gonna try to explain that to you, and the square you draw is gonna be off-centered, it's gonna be a little bit bigger, it's gonna not work as well. And the better listening skills and the better questions you ask, that square is gonna get slightly closer and closer and closer to what they're doing. And from a perspective of how I like to work, when I'm talking about design, design is the first thing I get signed off with a customer. Their data is always gonna tell the same story because it's that data. But the visual buy-in is something we're gonna get in front of. I'm gonna send them a wireframe, I'm gonna say, This is the idea I've got, how it's gonna look and you're gonna use it. If they're like, that feels really cluttered, or I feel like you haven't put enough information on this page, we can add to it then. If I build the whole report, put all the data on it, then put the shiny stuff on, and then give it to them, and they start asking me why that blue is used at the top, or why there's four charts at the bottom where we only want three, I've lost that trust, I've lost that buy-in from them. I know I'm rambling a little bit.

SPEAKER_00

No, it's absolutely and I think it's such a good point because information leaders within our um community, we have the same issues with buy-in. And actually, what you're describing and what I relate to is you're having those conversations with them, like, what do you want? How can I provide it? So you're you're building on that relationship with them. So one, it's not wasted work for you, but also when you present something to them, they're like, Oh, okay, you can instantly make that decision rather than why does this colour graph here or look like this? So I think it's um I think the way you've described it has definitely kind of related it to the way that our listeners look at the world. So, what I think is really important, you're making a point of avoiding dashboards that look beautiful, that answer the wrong questions, because actually I like pretty pretty picture things, but if it's does it's a pretty picture without a purpose, it's absolutely no point. So, sitting in that kind of middle ground where you sit between techno technology and business stakeholders, what is the framework? Do you have a framework that you use for digging past what the users are saying they want to find and actually what they need?

SPEAKER_02

Yes, and it's often it's often questions that they're not necessarily ready for. So a lot of the time people will ask for something specific. They were like, we want to look at the three months rolling about sales, or we want to look at forecasting for opening new branches in this country or in this region. What I usually do in those calls, and I I take down everything they've asked for and what they want. I then ask them a series of questions. From a design aspect, I always ask them what is free phone, what are free apps on your phone that you use too much? The reason I ask that is because Instagram, Facebook, LinkedIn, they are all built and they are built by very, very clever people, clever design teams to make you stay on that application for a reason. There's a buy in, there's a visual draw to it that activates something in your brain that makes you want to spend an hour on it doom scrolling. So understanding what the customer actually gets stuck into or what they actually use helps understand what they want from a design perspective. If they are a person who looks at very structured social media like Facebook or LinkedIn, where it's mostly text and photos, then they've got this buy-in. If they are somebody who likes spending an hour doom scrolling on TikTok, they like short bursts of stuff. So they need something that's going to catch their eye. If they are big into reading news on their phone, they're going to read it like they'll read a newspaper. The visual architecture of how they're going to read it is doing that. So I want to know those early buy-ins. I don't always get it right. Sometimes they'll tell me an app and I'm like, well, that's a game. I don't I I can draw some things from that, but like I can't always get it. And some people are like, well, I don't use a phone or I don't do that. My follow-up to that one is always, what are what are free websites that you regularly use that you enjoy the navigation and look of? Like, I've never heard anyone say they like using eBay. eBay is built for what it does, but it's a clunky website and it's ugly. But some people are like, I actually really like how Amazon works or or Sheen. Like, I like having that sort of flash in front of me. Following up from that, I start asking questions of what benefit is this report going to give you? Am I saving you time? Is this because it takes two people from your accounts team, four hours, to reconcile this information? Are we are we saving time here? Are you looking for insights you haven't had before? Because yes, you think you've got the data, but you don't actually know how to get it to show you that information? Or is it because someone above you is asking for it and you're now doing it? I I I quite like challenging questions about what people want from their data. There's a talk I'm giving in Manchester in two days' time, which is a it's just the one where I worked with an engineering company and they have a great, great product, an absolutely amazing product. The whole company is built by engineers, the CEO is an engineer, the FD is an engineer, they're all engineers. They had very, very big customers that did 95% of their business with them, and then they had little customers that did a little often. The long and short of it is the big customers always bought during fall or uh during spring or autumn, and the the failure rates on the bits were absolutely fine. During summer and winter, when they took the parts from a controlled environment in the workshop and took them to the warehouse, it's also controlled. They went outside for a little bit first. And the extremes of the heat and the cold made the failure rates on the bits worse. But because it was only smaller customers that are rolling during those times, the failure rate on the report looks minimal. And it was actually gone from like a 1% normal failure rate to like a 15%. And it's I asked a very, very stupid question of what's different during the times of year, and they're like, Well, actually, we have pretty bad heat waves here. It's like, okay. During those heat waves, what do you do? It's like, well, we have a skeleton shift, we normally have the loading bay full up, and it's like, that's not temperature control. Like I I asked silly questions to some of the most intelligent people I've ever met and gave them a light bulb moment because I'm asking questions up, because I don't understand. I'm I'm a I'm a computer nerd, I don't do machinery, I don't do other than the stuff on my desk. There is always a question to be had of someone, and it's the curse of knowledge that holds a lot of us back. If we presume everyone knows what we know, or they presume what we know, then there is never those questions of actually I don't understand this, or why do you do it this way? And it's those questions that actually help with reporting because oh, actually, we don't want to look at the three-month rolling or forecasting. We're trying to look where we're underutilizing our staff, or actually where we're sitting on stock for too long and it's going out of date, or however it's working. Like it's yeah, you have to ask silly questions to get good answers.

SPEAKER_00

Yeah, agreed. And information management, we assume that everybody knows what we're talking about, so either RMS implementation. So I'm very much on a speaker, uh a real simple level so people can understand, and that's exactly what you're describing. You're you know, you're asking those questions because you don't know the answers to build better stories and reporting for them. But it's also about having the understanding of what they want, so to get them to buy in, having those conversations and really getting to know them, I think it's absolutely essential for us all. So let's talk about code node. So you spoke about how AI isn't here to replace developers, but to make us dangerously effective. So while many corporate leaders are waiting around for IT to roll out official corporate co-pilots, you're actively using different um methods like chat G G P T I can never say it, chat GPT, Claude and Gemini today. So, what is the lowest hanging fruit for an information professional to start using these LL LLMs tomorrow?

SPEAKER_02

There's several different ways of looking at it, but I honestly think the number one thing I always recommend anyone who's going to use AI as part of their job, obviously, you never want it on live data. Like if you if you've got if you've got a system, like a well-baked system, you've got dynamics, you've got Sage, you've got any of the big boys, there's demo data out there. There is demo systems, you've probably got UAT, you've probably got demo. Like, if you can anonymise some data and ask it questions. Genuinely say, like, we've got this sort of data, where are the the footholes? I often ask ironically, like the big thing in in in my world is that everyone's trying to push for AI on top of their data. Whereas I know from a BI perspective, companies still struggle with bad data. If you slap on AI on top of bad data, it's going to give you bad answers. So actually use that the other way around. Like ask a question of your system. If you've anonymized data and you can do it that way, or like that's the best way to do it, ask where the holes are. Ask it where it feels like there's missing parts so that you can investigate from that side. But actual quick wins for an end user or someone who isn't using AI as part of their job, ask it something about your processes. Say, if something like this happened, I would do this, or we currently do this, what are some alternatives? Like it's the whole I always find that when, like, I go back to the silly questions or stupid questions. Whenever someone new joins a company, there are ideas that come from them in that first six months they're joining, and only better than anything to anyone who's worked there for a long time gets, because they will ask the silly questions, like, why are we still doing this when this is available? Or why are we doing it this way for this? And that's all learned experience. Like each generation that comes out and each person that uses tech differently will learn in a slightly different way, or will read a different article, or will understand a different point of view. Using AI for your current job should be an investigative way and an assistant way. AI is out here to assist us in our job. When I the phrase being dangerously efficient is one I love. Lots of people, and this is the same as what happened when computers come out and and and so on and so forth. When they say about saving time, like, oh my god, this thing is going to do 90% of my job, you're still gonna do the same amount of hours because you are gonna fill that time with something else. I want to be dangerously efficient at the things that are mundane for my job. The creating a certain table, like a date table. In Power BI, I always create a date table to align all my dates. I always want it to check for accessibility. Like I'm colour blind, I struggle with colour theory, I often have to rely on other people or AI to tell me what colours are correct for each other, and I need them to test it that other people who have sight problems can read what I'm doing. Because it's important. I'm not building a report that only 98% of the population can read. Everyone needs to be able to look at it and take something from it. So AI can check that for me. Like, why would I not have something that's been trained expertly in these fields and has the best information available to it to check those things? I'm not gonna ask it to look at all my customers' data and make a report for me. That's dangerous. That's that's irresponsible. But getting it to ask, why have you stuck six KPIs on there? Because your client brief says that they're looking for four high-level figures, but you've given them six. Are you trying to confuse them? Are you trying to guess what they want? Like, I often tell whichever AI agent I'm using or code agent I'm using to question everything I'm saying and make sure that I'm justifying it correctly and that I've got proof of why I'm doing it. Like I'm not gonna, I'm not gonna let it just. There's a the thing with AI, and and and you'll see it a lot of the time, whenever you say something to it and you disagree with it, it's like, oh yeah, sorry, I sorry, I caught that, or like, oh, you're right about this, and it's like, I don't want that. I don't want someone to just be a yes man for me. I need someone that's gonna be critical of me, who's gonna give me raw feedback. We part of being at Experts Live, you yourself hosted a panel on autism and AHD or neurodivergency. I'm I'm autistic in ADHD. I really like blunt conversations about work. If you tell me this didn't work, this didn't work, this didn't work, we need this, this, and this, I'm gonna do it, and it's gonna make sense to me. If we fluff about like, hmm, I'm not sure about this, or that could work, or I'm not gonna get it. AI can deliver feedback to me and tell me in very plain English for me, and it helps. I can't always get that from a colleague, I can't always get that from a stakeholder, but AI is gonna push back and say this. Yeah. AI is is a wonderful, wonderful invention. I always oh god. AI has obviously been around for a very, very long time in different ways. Like anyone of any generation that's played a video game before, there are characters that are designed and made to have artificial intelligence. They've they've got very strict guidelines, like they're only allowed to do certain things, they haven't set move apart, they have set spoken things. That's still that AI is still been giving a little bit of knowledge. The what we currently are looking at in the big LLM world, so ChatGPT, Claude, Gemini, Croak, the Elon Musk racist one, is massive models that have been taught on humans and all information out there. The one thing I would always say with LLMs is that think about it like it's Wikipedia. Wikipedia is a brilliant site, but it's also just written by humans. And a lot of the time humans are a bit stupid, or they're a bit mean, or they try to be funny and they say things that don't make sense, or they say things that are false. AI's been trained on all of that. AI has been trained on Einstein, but it's also been trained on Stalin. It's been trained on the best of us and the very worst of us.

SPEAKER_00

Yeah.

SPEAKER_02

So you need to tell it to be smart, you need to give it the you need to be this person, or you need to be this. So be structured, but just give it a go. I ramble, sorry.

SPEAKER_00

Love it. So you mentioned sort of like AI, and you obviously work in Power BI and Microsoft Fabric in sort of like short sentence. What is Microsoft Fabric to the listeners?

SPEAKER_02

Oh god, I should be able to meal this off, right? Uh fabric is like the future of data management, data storage, data flow and analytics. Microsoft are heavily investing in it, it's a very, very cool product. I can very much see in the next few years that Power BI is just going to be integrated into Fabric and no longer called Power BI. They're very much brother and sister, I would say.

SPEAKER_01

Possibly adopted brother and sister, but they're they're related.

SPEAKER_00

So yeah. So for information and records managers listening, um, who might be terrified of this shadow AI or messy, ungoverned data environments as you've um described, so we call it redundant, obsolete, and shrivel. So all about information that uh that AI runs off if you haven't uh cleaned your data up. How do we use these tools for speeding without throwing out data compliance and security out the window?

SPEAKER_02

It's all about guard rails. About turn it's actually about turning AI off. And like, when does AI have so many positives to it and so many privileges, and it's clever enough to do things that we have to turn it off? At the moment, it is still an infant that has been given a motorbike and still all at once. That's the only way I can describe it. Like it's it's it's doing really well, but like it's 10 seconds from crash at any point. How can people use it in a real-life scenario for actual real data? You need structured guidelines, you need enterprise AI, not well, the dude from marketing has made a clawed skill, let's use that for our business. No, you do not, you're not doing that. You're not doing an MCP, you're not letting people push stuff to production that's got API keys and stuff. You need the same level of IT guardrails that you had with all of your other software, but you've just got to look at it as a new bit of software that you're integrating into your business, a new CRM, a new way of doing it. It needs planning, it needs use cases, it needs testing. Like get some test data, like I said, spin up a demo, let a person in sales have a play with AI and say, right, who are my best buy prospects? Who haven't I spoke to? Get at those little winds of information where you're taking away some of the thought process. We as humans are consuming so much media now and so much information that we never used to. Like we are constantly overloaded, people are burning out, people are overwhelmed because we're constantly reading information, we're constantly looking at our phones, reading news story after news story. Like we used to read a newspaper for half an hour, and that was your news for the day, unless you watch the news at nine o'clock. Now you get 10:15 emergency or headline news sent to your phone every day. If you are taking AI and using it to take away some of the mundane things or to try and make your day quicker, that's what you want from it, and that's where you should be trying it first. From the broader perspective of like running out for your business, you need to do thorough testing like any other system. It's yeah.

SPEAKER_00

Testing and pilots. So the ultimate takeaway, so let's just think about it. We've talked about a lot today. What if there is an information leader listening right now who wants to make their data work harder but feels a little bit stuck, is the single biggest mental shift or prompt they can steal from your toolkit today?

SPEAKER_02

That's a great question. Like, what has seen like an AI prompt, like lost one thing you can give AI?

SPEAKER_00

Yeah, yeah.

SPEAKER_02

I would say give it if you're in a position to and you can remember what you've done, say give it the information of what have I tell it exactly what you've done this week in your working week. Like, say I had three meetings for four hours on today, I did two hours of coding yesterday, I did this, I did this. Ask it for my job role and for what I do for this company, am I being efficient and where should I be focusing on? And see what it says because a lot of us join and do a job, and then it's never the job we actually end up doing, and seeing where our value is because I feel like using AI to get over imposter syndrome is a big thing, like get it to understand your worth and get it to understand what you're actually good at so that you can be good at the good bits of your job and promote yourself like that. I there is I've got a whole prompt packed, I will give the link. There's stuff for BI, AI, all sorts in there. But just get it to critically think about the stuff you're doing and why you're doing it. So it can ask you stupid questions like why are you spending an hour in the morning telling your emails when you can just ask Copilot to give you anything to have an action to? Like, why are you why are you letting an email come in and then six seconds later trying to reply to it? Like, an email is not an instant conversation. If I wanted to talk to you like that, they'd call you. An email is for when you are available, not for actioning now, and that's important. So get it to ask you stupid questions, is my advice.

SPEAKER_00

I love that. I love that. Duncan has been absolutely amazing. How can listeners reach out to you if they want to know more?

SPEAKER_02

LinkedIn, I'm a LinkedIn demon. I spend far too long on the on the thing.