The Enterprise Alchemists

Transforming Integration with Creative User Experiences

• Dominic Wellington & Guy Murphy — SnapLogic • Season 2 • Episode 5

Share episode

Matthew Holloway, Head of Global Design at SnapLogic, joins Guy and Dominic to unravel the complexities of designing user-centered enterprise integration platforms. Matthew shares strategies for presenting complex data to diverse user expertise levels, and some of what he has learned from how SnapLogic's groundbreaking tools, Iris and SnapGPT, are redefining data pipeline creation.

In design as in tech, it turns out to be critical to pinpoint the right problems before rushing to “solutioneering.” With the buzz around generative AI and its often-misaligned expectations, the discussion sheds light on managing these through the design hypercycle. Matthew emphasizes fostering collaboration through rapid prototyping, involving key stakeholders to ensure solutions are not only creative but also viable and aligned with organizational objectives.

Here are some of the topics discussed in the episode:

• User-centric information architecture and its complexities 
• The challenge of designing for varied expertise levels among users 
• SnapLogic’s proactive design approach and its unique tools 
• Insights on co-innovation and collaborative problem-solving 
• The journey through the generative AI hype cycle 
• Transition from prototypes to fully realised products 
• Emphasis on personalisation and user agency in design decisions 
• Understanding the importance of making critical information accessible 
• The future of design through generative AI personalization

Resources that were mentioned:

Find more details and a full transcript on SnapLogic's Integration Nation community site.

Dominic Wellington:

Hi and welcome back to the Enterprise Alchemists. I'm your host, dominic Wellington. I'm here together with my co-host, guy Murphy. Hey, guy.

Guy Murphy:

Hi there and festive greetings.

Dominic Wellington:

Indeed the last one of the year, but we are joined today by our colleague, Matthew Holloway. Hi, Matthew, thank you for taking the time.

Matthew Holloway:

Hi, thank you. Thanks for having me.

Dominic Wellington:

So, just for our listeners, who are not aware of who you are and what you do all day, do you want to introduce yourself briefly?

Matthew Holloway:

Sure. So I'm the head of global design here at SnapLogic. Previously, I had been at a series of startup companies. I've co-founded about a half a dozen startup companies in my career, but I've also worked at larger companies such as SAP Oracle. I've also worked at larger companies such as SAP, Oracle, Shutterfly, et cetera, running their design teams, and then I've also done advising of other startup companies and also coaching of design leaders at companies like Oracle and Google and Meta and Amazon, et cetera. I've been doing this for about 25 years and I have a background actually in industrial design with an advanced degree in cognitive systems engineering. I used to teach at Stanford and I've taught at the Royal College of Art and Malmo and Sweden and a bunch of other places as well.

Dominic Wellington:

Wow, so some stuff in there I didn't actually know.

Dominic Wellington:

We have a little bit of a theme on this podcast of inviting people who are architects, but different sorts of architects from us. You're obviously not an architect, you're a UX and design person, but the architectural link there is to talk about information architecture and how you present information to users. So the topic of enterprise integration is obviously extremely complicated. We deal with large amounts of data presented in all sorts of different ways from all sorts of different systems. How do you think about exposing that to users? And especially in the context of SnapLogic that has this notion of citizen integrators people who are domain experts but not necessarily integration experts. How do you think about the information architecture there?

Matthew Holloway:

It's a big challenge.

Matthew Holloway:

It's kind of twofold.

Matthew Holloway:

One of them is, as you pointed out, the level of expertise that our various users have in terms of doing these data integrations varies widely, and even for those people who are very experienced data integrators, what we found through our research is that they only spend part of their time doing these data integrations.

Matthew Holloway:

It's not like they're doing this every day, eight hours a day, five days a week, and so the complexity of these data integrations then has to become much more self-evident and understandable by them, so that when they come back to troubleshoot them, update the integrations or build a new integration, it feels incredibly familiar to them. And so for the new users, who aren't as knowledgeable, we have a different challenge, which is not necessarily the familiarity, but just the self-evidence of how you build this thing and helping them understand what the underlying concepts are, what's going to happen to their data as we take it from one system, possibly do some transformations to it, merge it with other data sources, etc. And place it into their eventual destination. For those users, we have a different set of challenges, which is to really help them understand the implications of what their actions are going to do before they perform those actions on their data so they don't wind up creating a mess, basically at the end of their integration.

Guy Murphy:

With team, how do you think about, then, the design framework to support this almost more proactive view of UIs, because that's very unique. I mean, I've had dozens of integration tools that are reactive to the design process, rather than the way you're framing it.

Matthew Holloway:

So we have explored different options within designer. For seven or eight years now, snaplogic has worked with artificial intelligence and machine learning. We had a system called Iris that was a recommendation engine that would help the more experienced users by saying this is the next thing that you should add, this is the next Snap you should add into this process, and now we have the SnapGBT capabilities where you can enter a prompt and create your pipelines. Prior to SnapGBT, we launched AutoSync, which was really intended for those business analyst type users, the citizen integrator types, and that model. We actually flipped it from right to left. We flipped it to left to right. So we wanted to start with what their goal was. So they wanted to create a dashboard of their quarterly sales, and so we started with that objective and worked backwards through the integration to get to the point where they were collecting the right data, but they knew what they were going to do.

Dominic Wellington:

Start with the end in mind. What a concept! Right.

Matthew Holloway:

It's a very user-centered kind of approach, and so we've played around with different approaches for doing that, and you know we keep experimenting with new ideas of how we can improve the overall experience that people have.

Matthew Holloway:

Taking a step back, talking about architecture, the information like the one that we have, where you have to basically place the components of your platform in a logical way, so it's like laying out the blueprints of a house, and so there has to be a high degree of predictability in these systems.

Matthew Holloway:

Oftentimes designers focus on consistency so it looks the same, so that button looks the same every place that button appears.

Matthew Holloway:

But for expert systems such as the one we're working with, it's more important to focus on predictability so that if I'm going to perform an action, I know what the consequences of that action are before I commit to doing that action, and that may require the button changes right in order to facilitate that predictability. And so this gets into kind of the UX architecture of how you lay out the system, where the components are, what their relationships are to each other, how you navigate between them, and then the overlay workflows that you're going to use as you move through those different spaces in order to achieve that outcome and starting with that outcome is really critical and not making any assumptions about what people are trying to do with our product based on what we think they should do with our product, but actually going and talking to them and finding out from them what they want to do and what their problems are and how we can help them solve those problems is what's going to generate the greatest value for our platform and for our customers so that's the perfect segue.

Guy Murphy:

So I've obviously worked with your team in several accounts and one of them I got to work with several of them were under the concept of co -innovation, where you use design thinking methodologies with them. Could you just touch upon and it was a new concept in SnapLogic, but I'm seeing it becoming more embedded Could you touch on how that's been working, what these concepts are and how that impacts again, medium, long-term design patterns for yourself?

Matthew Holloway:

Sure. So the design thinking method has, you know, has been used now for almost 20 years in terms of, you know, being able to focus on identifying the right problem to solve and then working back from there building out rough and simple prototypes of solutions that will solve that problem. And so we brought that to SnapLogic to do co-innovation with our customers. So, you know, an ideal situation we all get together in a room, there's lots of whiteboards, there's stuff to kind of facilitate the creative thinking process, you know, get people to relax and to really kind of get into a mental state of play, which isn't what you normally think of when you think of enterprise, software or a company.

Matthew Holloway:

But from a psychological perspective, it's important as adults that we give ourselves permission to be creative, because typically we don't want to be creative, we want to be analytical as professionals, and so there's tricks that you can do to get people to go into a more play state, and so we do that with our customers and with our own engineers and designers and product managers. And so we spend a day together, you know, really understanding what their problems are, modeling those out, brainstorming different ideas, building some simple prototypes, evaluating what's working, what's not working, doing rapid iterations on those things and at the end of the day, the goal isn't to come up with a final solution, but with a great deal of clarity around what the problem is. So once upon a time, somebody told me that ideas are naive but problems have wisdom. The interesting thing is that, you know, oftentimes we focus on the ideas but we don't focus on the problem as much.

Dominic Wellington:

Is that particularly virulent maybe in tech, where we tend to jump to solutioneering and not always what is the user actually trying to do?

Matthew Holloway:

Yeah, and we see that with see that with the generative AI hype cycle that's going on right now, where everybody wants to use Gen-AI to solve all of these problems, regardless of whether or not it's capable of doing that or not, and so they have these great expectations about all the magical things Gen-AI is going to do for them, which is why there's such a great deal of disappointment around some of the solutions that are actually being rolled out. I think the folks at RAND just did a study where something like 80% of the Gen-AI projects fail because there's a misalignment around the expectations people have for what it can do and what it's actually capable of doing, and they just they see what the outcome is and they walk away from it because it's just not being able to live up to their hype cycle.

Dominic Wellington:

Yeah, and that kind of feeds into. You published a piece on LinkedIn not too long ago and we'll put that in the show notes for listeners to be able to refer back to talking about how to bypass that hype cycle. And it is as you say. It's starting to be a problem because there's just so much hype out there that there are very real capabilities and the hype is so much bigger than those real capabilities. And I liked your notion of the design hypercycle trying to reset some of those expectations. So you get maybe a lower peak of inflated expectations, to use the Gartner terminology, but you also don't dive quite as deep into the trough of disillusionment and with any luck, luck, you avoid project failure. Embarrassment uh.

Guy Murphy:

Wasting budgets uh, bad things, could you dig into that a bit more, because I think we're all seeing this, where you get this disconnect um in many accounts and again it's interesting, we're starting to see the sort of not just a trough of disillusionment, but also a few clarity statements coming out of the people who've been what does work?

Guy Murphy:

yeah, as you say what, what, what is the capability to the engines? Does it actually solve the thing you think it's going to solve? Um, one of those telling reports I saw said and even if you've got that, do you actually have the culture and data to solve the problem through the new technologies? What's your point of view on that and is there a path out of this?

Matthew Holloway:

I think there is a path out of this.

Matthew Holloway:

Every new technology since I've been doing this, starting with things like wireless networking back when I was at Apple the internet, netscape, handheld devices, social media, all those types of things it's followed the Gartner hype cycle curve with varying degrees of peaks and troughs, but I think over time the you know folks from the design community at least have learned that by doing rapid prototypes upfront.

Matthew Holloway:

These aren't necessarily functional prototypes but they are proof of concepts that I mean they're not production-ready prototypes, where you work with the engineers who are building the technology, the marketing, the product management folks from a business perspective, and where you're going to get revenue and how you can monetize these things. And building out these prototypes rapidly allows you to kind of see what, what you're getting for your investment much, much faster and you can take that to your customers. You can shop that around inside the organization and other people can contribute to that and reflect on it by also making it tangible. You know a cultural perspective. Not every organization has a culture that accepts innovation. Oftentimes innovation is off in an ivory tower, there's an R&D team and there's a natural level of resentment for that, because those people are special.

Matthew Holloway:

There's a special digital team and they're not actually connected to what the rest of the org is doing right and so, uh, being able to bring that prototype out and make it tangible and show people um, everyone's like, oh, okay, that's what you mean, that's what we're going to get and that's how it's going to work, and now I see why this is valuable and how this is going to add to our growth opportunities. And then you know, being able to build on that prototype rapidly and iterate on it and refine it, I think allows you to have a much more targeted growth and acceleration and engagement, because you know what you're working towards, as opposed to just a vague thing about like I want Gen AI everywhere. It's like well, this is how we're going to use gen I to solve this specific problem with this specific set of metrics around evaluating its success, and we can work towards that goal in a very clear way. And then, when you deliver that success, it naturally leads to other successes and other efforts you learn from it.

Dominic Wellington:

You take the learning somewhere else. Yeah, one, you mentioned Apple. One of my favorite Steve Jobs stories is in the creation of the original Macintosh. One of the apps that they were going to build was the calculator. Simple little app, but Steve Jobs kept on throwing it back to the design team, giving vague feedback oh I don't like this, could it be more like that? And eventually what they ended up doing was they built a calculator design toolkit so that steve jobs could just play around with it and assemble the, the calculator that he wanted, which ended up being what the designers wants to build anyway.

Dominic Wellington:

Very close as I, as I recall the story, but the point was, by playing with it he learned learned what works, what didn't work, and he got to something which remained remarkably unchanged for decades in the macOS. It's quite a color at some point, but not much else changed with the calculator app, and I think that's key to adopting AI. So at SnapLogic, we have a few customers several customers at this point already who are using AI in production and we've learned from it. They've learned from it and now we're moving to the day two use cases. Okay, we did the easy little thing and now we go for the big strategic use case and how do you think that changes that transition? How do you think that needs to be thought of?

Matthew Holloway:

For these more complex use cases.

Dominic Wellington:

Exactly! Moving from the prototype to the minimum valuable or minimum viable product, to something real.

Matthew Holloway:

I think you know the second time you do something it's easier than the first time and the third time ideally gets easier still. You know, even with the rate of change that we're seeing in Gen AI, with the emergence of new technologies and capabilities, you know it's rapidly evolving. I feel like once you kind of understand the basic fundamentals of how these systems are different than what we've used in the past, it gives you an advantage to really treat this technology as a material, not as a sacred thing. So it's like designers think of software as a material, the way that you would treat wood or fabric or aluminum as a material, and so you learn how to manipulate that material to get the effect that you want.

Matthew Holloway:

In fact, everybody, I think, is probably familiar with the Eames lounge chair, the Herman Miller lounge chair. That came about from Charles and Ray Eames building the Kazam machine, which was a steam bender thing for plywood. So you'd put the individual sheets of laminate in there with glue between them and you could press this thing into different shapes. And there were different clamps and forms and other stuff in this Kazam machine so they could make it into all sorts of different shapes and they just spent months doing that to understand how to manipulate that material.

Dominic Wellington:

This new material yeah.

Matthew Holloway:

And that resulted. Initially it resulted in lightweight splints for World War II, but those got translated into their molded plywood furniture and then that technology has been used by other people for decades since then to do other things, has been used by other people for decades since then to do other things, and so I feel like that's an opportunity that we have with Gen AI as well is to really think about it as a material, not as just an ingredient that you add or sprinkle on top. But how are you going to manipulate the Gen AI capabilities? How do you manipulate the libraries and the algorithms, the prompts? We see that with our internal tools that we're building, like the prompt workbench that we're working on, it dynamically adapts the prompts to the profile of the data passing through the pipeline, and that really takes a reconceptualization that these things are manipulable.

Matthew Holloway:

You can shape them. They're malleable, you can force them into different alignments and everything, and so I think the more you work with this technology as you move into the day two, you're more familiar with that, and so people are going to think about this stuff as much more plastic and come up with much more innovative solutions for it. So theoretically, it should accelerate the development of those solutions, and it should also allow them to create more robust solutions to cover the heterogeneous workflows that most of our enterprise customers are dealing with.

Dominic Wellington:

Yeah, it's interesting to see the transition from some of the early assumptions that everything would look like a chatbot to now with agents. People are thinking about radically different interface methods where the AI is often even not visible. It's behind the scenes somewhere, as you say. It's just the material that's used to build the thing.

Matthew Holloway:

Yeah, well, actually, one of the other articles I posted on LinkedIn a while back was around generative user experience. I posted on LinkedIn a while back was around generative user experience, and the idea was that the elements that comprise the user experience on the screen are just no different than words in a paragraph, in a lot of ways right, based on your intent or based on the prompt that could be formed, on the user's behavior, or non-verbal cues, setting place, time, et cetera. Theoretically, you could generate a user experience that specifically addresses their needs, and so when mobile first came out right, so everybody was just trying to cram their application onto the phone and it became apparent really, really fast that putting everything on the phone was not going to work, that you had to.

Dominic Wellington:

You had to understand that even Microsoft could force that in with Windows Mobile.

Matthew Holloway:

And so you had to adapt your user experience and the functionality of your product to the device, based on the context people were in.

Matthew Holloway:

So if they're walking down the street, for example, they don't want to, well, they do now, but at the time nobody just wanted to keep looking at their phone. They wanted the phone to basically tell them which way to go, which is how we got haptic experiences on the phone, where it would vibrate to tell you to turn left or vibrate differently to tell you to turn right when you got to an intersection. So you could just really enjoy the space that you were in, walking through the city or something, while you're still successfully navigating to your destination. And I feel like we have the potential of doing more of that with generative AI to really dynamically create the experience, the best experience for you, which would be different than the experience for me or for guy, based on what I wanted to do and how I like working and my preferences for the tools that I use or the you know any of those types of things that was fabulous because you just literally preempted the next question I had, which was if you could look forward to three years.

Guy Murphy:

What are the design changes? You see, and everything you said is fascinating about that concept, um, and again, um. That concept of the dynamic, almost self-learning ui I can completely buy into as a very small sign-up. I'm into photography and I've been a Photoshop user for years. What was very interesting was, in some ways it was a functionally incredible product, but its design ethos was, I'm going to say, one of arrogance, and I'm aware Adobe is a customer of ours and I love their technologies, but for years they were renowned for an attitude of it's the user that works out how to use our incredible engine and then over the years, now they've got, they've learned the fact of, actually, how do you streamline the journey for the you or the photographer and bring the capability at the right times in the right places? So they've gone through that evolution, but man, obviously manually, with traditional design. Um, I guess this is going to become a turtles all the way down, this interlinking of engines to drive the process, engines to drive the human engines, to be the process themselves.

Matthew Holloway:

Yeah, I feel like you know, earlier we were talking about the minimal, viable or valuable product model, and I feel like we're going to wind up with something like that, but it's basically going to be what I think is a valuable product, and so you're going to have a set of capabilities you know, like in an app like Photoshop. You know, literally there are thousands of functions inside of Photoshop and any one user probably uses 10% of them, right, and so, rather than having to have an information architecture in that application and you know a set of workflows and the way you navigate and the tools and all that kind of stuff that contains a thousand different options, why not just have those ten and let me just really focus on those ten things? And then, as I manipulate that image on the back end, photoshop is like hey, you know what you could really benefit, I think, based on what you're doing, relative to what other people have done that I've watched, you might want to do it this way as well. So here's three new tools that you may be interested in, and I can show you how to do that and I can walk you through it, because the Gen AI has the ability to kind of, on the fly, do training based on what it knows is your relative skill level and the types of things you've been doing in the past.

Matthew Holloway:

And yeah, it sounds super creepy, but, uh, one of the startups I was at a few years ago, we were doing something like this with a desktop agent and after the initial kind of uh setup where you could, you know, blacklist items and whitelist other items, customers became very dependent on this as a way of making recommendations to them and automating tasks for them. As a way of making recommendations to them and automating tasks for them. So I have every reason to believe that people will be very accepting of you know these technologies, kind of watching you, and then you know, if you look at the quantum chips that are coming out and the ability to kind of put an entire Gen AI engine onto your device with, you know, very low power consumption over the next few years. It's not even a question of that data getting away from you because it's just going to be on your device. Some of the models may be passed back and forth, but anything that's personally identifiable data would be localized to your device.

Dominic Wellington:

Yeah, and that's the promise of these algorithmic and generative tools, and I saw it first. I was working in the monitoring space, and I say monitoring because it was right at that moment of transition between monitoring and observability, and the difference between the two is. In the old days the problem was how did you get enough information to know what was going on? You never had enough information. It was always about how can you get one more bit of information out of your infrastructure, and the transition to observability was a realization that wait.

Dominic Wellington:

These days, everything is instrumented. Everything is just spewing information. All the time is instrumented. Everything is just spewing information all the time. The problem that we have is how do we surface the one bit of information that we need right now from that unceasing torrent? And that is becoming the problem that we have in all sorts of spaces right now. In my phone, my, my front screen is all smart stacks and they never look the same at different times of the day or depending on what I'm doing, because I'm always surfacing different information, and it took a minute from when I first set stuff that way, but now it's almost telepathic. Everything I want is usually one tap away on the front screen of my phone. And that is the power of what we're promising to enterprises that they will have what they need.

Matthew Holloway:

Promising to enterprises that they will have what they need in the moment that they need it and will not be drowning in irrelevant information, right yeah, that that's the part that I'm very excited about, because I feel, like um, it will level the playing field for so many people inside of these companies in terms of really allowing them the freedom to think more creatively around what they're seeing from that data. And so you know, right now, you know you spend so much time wading through all this data, coming up with, you know, collecting it, analyzing it, putting it in a structure, being able to see stuff. You're exhausted by the time it comes to saying, all right, what does this mean? And going back to the psychology of creativity and problem solving, you do need a good night's sleep. To be creative, you have to be in the right mindset. You can't be depressed. You have to be optimistic about stuff. You can't be depressed, you can't. You know, you have to be optimistic about stuff.

Dominic Wellington:

You can't force creativity, yeah.

Matthew Holloway:

Right, and so if this information comes to you and it's just kind of like there, I think there's going to be a lot more opportunity for people to use it in a much more creative, innovative way than what they may have been able to do in the past.

Dominic Wellington:

Yeah, and that's also links with the previous conversation that we had with Brian Greene, and he was talking about how to apply more traditional software engineering concepts in the much more visual no-code paradigm that we have at Snap Logic. But again, it's going to be not about forcing someone to jump through a variety of hoops, some of which may be on fire, in order to do that. It's going to be about giving them the tools, putting them in their hands at the right moments in their own cognitive journey so that they're able to take advantage of those things

Matthew Holloway:

Yeah, it kind of reminds me of, you know, people who buy their kids Lego sets.

Dominic Wellington:

Buy my kids Lego sets? They're for me!

Guy Murphy:

Careful, careful of kids.

Matthew Holloway:

These friends of mine. You know they would buy the Lego set for their kids and they would help them build it. So they were like, the first time around they would build exactly what was on the outside of the box. Mom and dad were very happy. They built what was on the outside of the box. And then you go back a few weeks later and the kids have used that Lego set combined with three other Lego sets and maybe some Mindstorm stuff to come up with something completely different. Because it all just snaps together and it makes sense to them and they just see it.

Dominic Wellington:

Yeah, now the cowboy and the knight and the spaceman are riding the train and there's a dragon.

Matthew Holloway:

And so you know I'm hoping that you know these types of tools and this low-code, no-code, you know capability of kind of snapping things together and you know each of those components. Having an awareness of what it can do and what the other components can do will help people come up with much more creative solutions to these complex problems, allowing them to personally create the solution that works best for them. Like you were saying, with your screens, your setup works really well for you, but I'm sure if I looked at it I'd be like I'd be lost.

Dominic Wellington:

Yeah, yeah yeah, and that is especially in the SnapL ogic world, where we span so many different domains under the general umbrella of integration as data and apps. There's APIM, there's GenAI and the agentic components that we've added more recently and, of course, the administration of the platform itself. You need to have a unified toolkit. The Lego metaphor is great there so that people can find what they need, snap it together, get their jobs done and, as you say, for most people this isn't the whole of their job. They have other things they need to get to. They need to get in and out out, get what they need and move on yeah, that's, very true so, matthew, thank you for that fascinating as ever and thanks for sharing your insight.

Guy Murphy:

I did not think we'd be spanning from the design of plywood furniture through Lego but, as ever, these are important concepts and, yeah, I hope everybody listening really enjoyed and found this an interesting discussion, and this is impacting SnapL ogic. Your teams I've seen since you've come on board has made massive impact on the product where it is today, where it's going, so thanks yeah, lots more to come on that front.

Dominic Wellington:

I've seen the mock-ups, and I'm looking forward to it. Thank you so much, Matthew, for joining us. It's been a fascinating conversation uh. For the listeners, I'll drop links to uh those two linkedin pieces that matthew mentioned uh into the show notes and if you're listening to this episode standalone, do notes. It's part of a series you can subscribe wherever fine podcasts are downloaded. But until next time, thank you very much for listening and we'll be talking to you again soon.

People on this episode

Head Shot

Dominic Wellington

Host
Head Shot

Guy Murphy

Host