Episode 1

Piloting Generative AI for Creative Operations

Listen now

Listen to Mark, the Global Head of Creative Solutions at HubSpot, speak on how AI is revolutionizing creative operations and how Hubspot is changing the game with a whole set of AI tools.

  • The potential of AI in creative operations and how it’s more of a co-pilot than a replacement for creative teams. 
  • Challenges like visual anomalies and biases in AI-generated content and ethical AI practices.
  • The future of AI in digital asset management operations, where AI can classify assets and create more meaningful taxonomy.
Mark Brady
Global Head of Creative Solutions
Arjita K
Director, Growth and Business
Development, Artwork Flow


Arjita: Great. So hi Mark, welcome on the Creative Operations Podcast by Artwork Flow. Thank you for taking out the time to be with us today on this podcast and would love to know more about you. Maybe if you could give a quick introduction about yourself before we dive in, we do a deep dive here.

Mark Brady:  Hey, Arjita. Thanks very much for having me on the podcast. I'm super excited. Sure. A little bit of background about me. My name is Mark Brady. I am the Head of Creative Solutions at HubSpot. I have a background in both design and software engineering, and I currently lead a team of a dynamic team of engineers, designers and program managers. And the work that my group does within HubSpot is really tailored to enhancing HubSpot's creative capabilities for marketers and creatives through self-service platforms. So we're really looking out for the latest technology in terms of innovation and technology that can really help our stakeholders get things done quicker.

Arjita: Okay, great. That's very interesting. So, you know, as you mentioned, you had a mix of background of creative as well as software. So how did that, you know, how did that come together and what did your career path, you know, look like? Like, did you start as a creative function? Did you start in the technology function? How do you know that worked out for you?

Mark Brady: That's a great question. I'd love to tell you that I very carefully and meticulously planned it this way, but I absolutely did not. I really ended up where I am now by accident. I started with general engineering in terms of what I was studying, so that was mechanical and electrical. And while I was studying that, I was building computers for a local computer shop. I didn't really like that, but I finished that.

But what I was really passionate about was design and graphic design. So after I finished that, I studied graphic design and multimedia design. And that was where my passion was. But unfortunately, when I finished that and I came out, the market was flooded with graphic designers and they couldn't get a job. So I kind of slipped back into that software and computer side of things. And that led on to a 10 plus year career of doing software consultancy for companies like the European Commission.

And then I actually joined HubSpot about five years ago, nearly, but I joined on the product and engineering side to help them out with the content problem. And while I was in HubSpot, I moved around a couple of different places like FinTech, business intelligence. So it was very much on the data and software engineering side and a somewhat serp-enduity opportunity arose where our marketing team were doing a restructure.

Mark Brady: And they saw that I built some process change in marketing or in the engineering group, and they asked me to do a talk. So I talked to them about some of the changes that we put in to make the software process more efficient. And that turned out into a job offer. And then I moved from the product and engineering side over to marketing. And I now find myself in creative operations. One thing that is really cool about what I do right now is I've always been fascinated and intrigued by creativity.

Arjita: Okay.

Mark Brady: But I also have a passion for data and engineering. And I think where I am now, it's the perfect blend of both of those worlds. It's the creative side mixed with the software engineering and analytics side. So it's a very cool position to be in.

Arjita: That's really great. And I sort of resonate with that. I don't have formal training in design, but I am an engineer in terms of background. So the work that I also do at Artwork Flow, it has a lot to do with business and the technology we are building because it's a SaaS platform, but I've also started a design business in Bizongo. So it's been like a unique, like a unique mix of experiences. And I'm glad that your degree in graphic design after so many years actually gave you this, you know, sort of like an amalgamation of all the good things.

Mark Brady: I am nowhere near as talented as our actual design team, but I can at least talk the talk.

Arjita: Yeah, that's great. And so Mark, what is, so, you know, just for our listeners to understand here, right? Being on the technology side of the creative, what does it actually mean for an organization like HubSpot?

And I am an ardent fan of HubSpot. I've been using them for three years. I read their blogs. In fact, some of their infographics and blogs I've saved as bookmarks because I find the content really good. So I'm sure to manage that level of quality, I would love for you to break down what it means to deliver creative at that scale.

Mark Brady:  Great question. So when you think about HubSpot, we're a massively globally dispersed company and we have teams all over the world. But our marketing function, their sole purpose is to market the great products and features that HubSpot develops. And there's many different facets to that. We have a HubSpot community, we have our blogs, and then we have blogs for each one of our specific product hubs as well.

So you mentioned that you've saved some infographics. It's highly likely that the infographics you saved would have either come directly from our creative team or come from one of the tools that my team provides to our marketers. So on the technology side, specifically for what I do, my team is focused on self-service. So we help marketers get things done quicker without the need of necessarily waiting for a creative partner to design something for them. So we have two main facets of that, we have tools that we build in house with my team. So landing pages, we will build them like Lego block modules, so they can build a landing page really quickly. So that's one element. And then we have managed platforms. So we manage HubSpot's Canva Enterprise account. And within Canva, we have all of our templates in there to allow marketers to create infographics really quickly.

Arjita: Okay.

Mark Brady: We also have a digital asset management system where we store all of our creative assets so our marketers can use them in terms of imagery. And then also if we're working with vendors, we give them access to a creative kit so they can access our brand guidelines, who we are, what we do, how we represent ourselves, and get access to assets like our logos, our icons, and all of those things.

Arjita: Right, interesting. So all of this entire gamut of, you know, publishing, content management, everything comes under the umbrella that you manage.

Mark Brady: A lot of the publishing is done within our marketing group, but our creative team is like a small subset within marketing. So that's how we operate. And then my team also keeps our ears close to the ground for like the latest in technology to find new ways that we can provide better value. And one of the things that we recently explored was a lot of AI pilots, particularly around mid journey, for example.

Arjita: So, you know, that's the next main theme of the chat I wanted to have with you. You know, how is AI changing this entire landscape and how are you implementing that in the tools that you use, the processes that you develop, the tools that you develop in-house? Like, what is the application actually, you know, what are some of the processes that have changed for you? Has it increased productivity and where are you seeing an actual usage of AI?

Mark Brady: Great question. I still think professionally, I still think we're quite early yet to kind of see true benefits. And the reason why I'll tell you that is we recently ran an AI pilot where we provided mid-journey to some of our creators and also some of our marketers. And that pilot ran for about two months. And we ran into issues such as visual anomalies.

So you would have photorealistic models who might have like three or four legs or two rows of teeth. So obviously that's not imagery that you'd want to necessarily publish on a.com page. We also know that AI has a lot of issues around biases, stereotypical imagery. And if you're looking at more, I guess, text type AI interfaces such as Jasper or things like that, we also know that there's hallucinations where it can just make up fictitious facts. So there are a lot of issues. And I think we're quite early in the AI adoption story to kind of start really using these technologies in a public facing way. That being said, when we ran our AI pilot for Mid Journey, what we found was that it's absolutely fantastic as a co-pilot for creatives. So if you want a creative to really think outside the box or to reframe a problem,

It's amazing at ideation and discovery. So thinking about new concepts and new ideas as a starting point. And it's also very effective at storyboarding for video teams as well.

Arjita: Right. So that's interesting. You know, so like things like, you know, mood boarding, just doing like a quick brainstorming on the brief. So those kinds of things, rather than focusing on the final output, just the ideation phase itself and just to get that creative block away. Some of these AI tools can be beneficial. So you've implemented that as a process in your team or, you know,

Is the adoption of that mid journey tool good? Like do they use it on a day to day basis? That is now.

Mark Brady:  Our creative team continues to use it. So part of the pilot, we wanted to kind of test the feasibility if we were to adopt Mid-Journey and kind of roll it out much wider. What are the implications? Like what are the implications on our brand? So one of the other challenges that we've seen is that it's very difficult to produce brand images or brands that resonate with who we are as a company without a significant amount of pre-prompting. So you almost have to train mid-journey, for example, and like, this is our brand, this is our colors, here's how we represent ourselves. So you're now looking at massive prompts to get a small image. But that being said, we still find so much benefit from the ideation and discovery process that our creative team at least is going to continue to use the platform throughout the rest of this year, but we're going to limit use within our marketing group because we realized that in order to create publishable output in terms of not having stereotypical imagery or any visual anomalies, it takes quite a long time to get there. And on average with our test group it was anywhere between five and 10 iterations to get to a usable output. So over the period of the pilot, we generated about two and a half thousand images and we only really used about 50 in the public domain. So that tells us that we're spending more time trying to create the imagery than actually using the publishable output. So it's not really very time effective from that perspective.

Arjita: Yeah, got it. And I'm sure like for a large brand like HubSpot, you know, maintaining your brand compliance and just, you know, the look, the feel, the tonality of the brand matters at the highest level. So you would definitely, whatever you would use has to follow those like certain specific guidelines.

Mark Brady: Yet again, but it depends on what it is that you're creating. So we have a lot of sub-brands within HubSpot. So we have the Hustle, for example, which is a bit of a grungy feel to it. And it's kind of more conceptual art. AI is really good at that. But when you look over some of our sales blogs, it's more professional type imagery with typically human models in an office environment and mid-journey can struggle with those types of images. So it really depends on what you're creating.

Arjita: Yes, that makes sense. 

Mark Brady: Obviously, if we're doing blog articles on AI generation, well, that's a really good use case for using AI.

Arjita: Right, interesting. Did you see any like the 50 images that you finally ended up using, you know, in terms of like consumer reviews or just usability of those blogs? Was it good, as good as the other human generated image or was there a difference in that?

Mark Brady: It's a great question and that was also part of the reason why we ran the pilot because we needed to test the effectiveness of using this technology. What we noticed is in some of the A-B tests that we ran that we were getting a lot more click throughs with the AI images. I think one thing that's brilliant about AI imagery is it's fascinating and it lures you in to take a closer look. And I think we noticed that certainly in some statistics of the pages and content that we published.

The only downside I would say is, as you probably know right now, the market is saturated with AI imagery. So you've lost that first movers advantage because everyone's doing it now.

Arjita: Yeah, and it moved so quickly. Like we had the first versions of the software coming in last year, November, and we are only like eight to 10 months from there. And then so many new models have developed. Like me, yesterday I was reading about this model that is very good at just doing landscape photography or landscape imagery generation. And there are several others that do like just the product.

image regeneration. So like even within the GNI, you know, these niches that are developing so quickly that it is almost amusing to test.

Yeah, so we've talked about generative AI, and you also mentioned that you also worked a lot in the data and analytics space. So within those processes, any other applications of AI that you've bought on within your teams and that has really helped you anything, like any smaller applications or anything.

Mark Brady (14:19.391)

Yeah, I think some of the things that we're looking at right now, although we haven't quite yet implemented it, but it kind of goes to the question around like what is potentially the future of creative operations. One thing that AI is absolutely amazing at is its ability to understand user behaviors, understand unstructured data and classify it. So when you think about the things that typically happen within creative operations, like one of our largest operations that we have within my group, is our dam operation, like all of our digital assets. So we have about 10,000 digital assets that we provide to vendors and our creative team. But the challenge with that is you have what I would call the Netflix experience where a marketer wants to create content really quickly, they'll go in and it'll be presented with 10,000 assets to pick. And that's paralyzing because you have so many options to pick. 

And I think where AI can kind of really, the future of AI, I believe is figuring out a user's intent. So what type of content are you creating? Is it a landing page? What is that landing page about? And based on that information, to streamline those 10,000 assets down to about 30. Based on what you're looking to create, and based on our taxonomy, we think this is the best content for you today. I think that's where the advantage is going to be in the future of really helping the likes of Dam operations to apply rich taxonomy.

and tagging onto all of our assets and streamlining that entire process. And that's also multifaceted in the sense that if I'm a creative vendor, what discipline am I? Am I video? Am I visual? Or am I copywriting? Because based on the answer to that question, you can probably remove 50% of the assets in a given DAAM and only give them what matters. And I think that's kind of where the future is, I think where the real benefit could come from.

of the classification and structuring of unstructured data.

Arjita, Artwork Flow (16:18.57)

That's really interesting. So, you know, some of these features, we've also built in our workflow, like say if you have a dam, you know, the semantic search, it can do an auto detection in terms of what are the elements of the images. But this is a very interesting problem statement where you can actually say you have a dam which has.

20, 30,000 resources and you're creating a blog for example. So based on what intent of users or ICPs that you want to target, you should be able to search in them and then find 30 assets, as you said. So that's a very interesting use case.

Mark Brady (16:57.819)

It is an ongoing challenge, particularly when it comes to taxonomy, because let's say, and this is completely hypothetical, but let's say if I wanted to create a blog article and it happened to involve some sort of image of a garbage can. So you might search for a garbage can, but the person next to you might search for a trash can. Both of those things are the same thing, but if your DAAM has a very strict taxonomy, it won't show you the results for trash cans. It will only show you the results for the garbage can. And I think...

Arjita, Artwork Flow (17:11.438)


Mark Brady (17:25.671)

It's that intent, like someone is clearly looking for a visual representation of a can that stores rubbish. So it could be either those terms and it's really understanding that intent and kind of bringing them the right assets at the right time instead of having a very strict taxonomy. I think the other thing to bear in mind as well for some larger companies, when you have your brand values, who you are as a company, like what's your voice of incorporating that into your taxonomy as well. And that kind of speaks to that intent.

piece as well. So for example, one of HubSpot's values is joyful, vibrant. There's lots of assets in our DAM that could match that. And I think of leveraging AI to kind of auto tag some of your assets in a meaningful way. And the key there is the word meaningful, because some DAM providers will give you AI tagging. But it might not be that useful. So for example, we no longer use this asset, but we used to have a little plant pot.

that we grow better together. It was part of our branding, to grow better together. And when we used AI tagging, it would tag that as organic matter. Nobody's gonna search for organic matter. They might search for plant pots or flowers, but they're not gonna search for organic matter. And I think that's the trick. It's like tagging your taxonomy in a meaningful way that makes sense to a user's intent.

Arjita: I'm going to go to bed.

Arjita: Yeah.

Arjita, Artwork Flow (18:48.002)

Hmm, right. That's really interesting. Yeah, and that's a wonderful example as well. I'm gonna take that back to my product team as well. Great. So yeah, I mean, that's also a very, you know, interesting use case of AI, like not just JNI, but actually, you know, in larger companies where...

The assets are numerous and it's humanly impossible to search for the right things that you need. You know, just having those classifiers handy within your DAM could really make a difference between publishing your blog in five days versus publishing the next day.

Mark Brady (19:29.183)

Yes, absolutely.

Arjita, Artwork Flow (19:31.358)

Yeah. So you mentioned briefly about ethics and the biases that some of these models that are out there have. So how do you approach these issues in a company like HubSpot with intellectual property rights and ethics and biases when using tools like AI?

Mark Brady (19:57.075)

Well, this is a hot topic, I think, generally across the board. But I think when I think about ethical considerations and using AI ethically, fundamentally, we should always use AI in an ethical way. And I think not unlike using any software package, you should have good judgment, acceptable use policy. In HubSpot, whenever we ran our pilots, we instructed all of our users.

don't create anything that could be regarded as a likeness, either in voice or visual representation of somebody without their expressed permission. And like just being careful about that. It's worth also mentioning that both in North America and Europe, there are several legislative sessions happening right now as the world is trying to comprehend what is the future around.

licensing of AI based content. Like can you truly copyright AI based content? Because it may have been trained on images that couldn't, that might not have been licensed to that particular model. But I think one thing I would say to kind of push this back in you a little bit is originality. Like how common is original thought? Like show me a designer who doesn't have a collection of

Arjita, Artwork Flow (21:11.142)


Mark Brady (21:17.963)

show me a designer who doesn't go onto dribble.com and have a look at other designers' portfolios for inspiration. So in that sense, is AI really any different? I think if you have a designer who is expressly looking to replicate and copy like another artist's work, that's a different scenario. But if you're using AI in a way for inspiration and ideation, I think that's probably where it's most powerful. And I think, I think companies

Arjita, Artwork Flow (21:28.312)


Arjita, Artwork Flow (21:46.926)

I'm too fired.

Mark Brady (21:48.356)

As they adopt AI, I think they should really see AI as a co-pilot, not as a replacement.

Arjita, Artwork Flow (21:53.494)

Right. That makes a lot of sense. And as long as it can augment some of the work that is done by creative teams, content teams, and really just be a tool for them to brainstorm, do some of these mode-boarding exercises, there is definitely an implementation scope for these new technologies that are emerging.

Mark Brady (22:22.867)

Yeah, I think AI is really good at enhancing efficiency, but the core of what we do in creative, the storytelling, the emotion, the connection, that comes from a human touch, that comes from your team. So I think if you're using AI to kind of help you think and get you to the end result quicker, fantastic. But if you're a company and you're planning to use AI to replace your designers, I don't think that's a good idea and I wouldn't recommend it.

Arjita, Artwork Flow (22:32.728)


Arjita, Artwork Flow (22:48.742)

Yeah. In fact, you know, like my personal hypothesis is that AI will

push for more creativity because everything that can be automated, everything that is mundane, everything that has already been done, there is some way you can get it from some of these models. But when it comes to storytelling, when it comes to creativity, when it comes to actually pushing the bar on originality, it will probably further propel that fire in the industry.

Great. So, you know, some in terms of, you know, your current work, if there was any one repetitive task that you could automate, what would that be? That takes up a lot of your time, you don't want to do it, anything like that.

Mark Brady (23:37.387)

Just one? I could think of many, many tasks. I think, yeah. When I think about the world of creative operations, and particularly dealing with internal stakeholders and potentially with external vendors, one of the biggest time sucks that I see is time tracking and resource allocation. It's like...

Arjita, Artwork Flow (23:42.undefined)

No other talk for me on your list.

Arjita, Artwork Flow (23:58.205)


Mark Brady (23:59.039)

That's a very laborious repeating task. You have a vendor, they send in an invoice, that invoice has to be paid, it has to be attributed to a certain stuff, like expense code. That takes up a lot of time, but it's also very parameterized. So I think AI in the future could just take that off your hands. And I think the more of that administration and busy work that AI does, speaking to your last point, will free up our creative teams and our creative operations professionals to focus on what matters.

Arjita, Artwork Flow (24:28.015)

Right, interesting. Great. So we have just a few rapid fire kind of questions to just end this very interesting session with. So what's your favorite project in your entire career? Just like a two, three line description of that.

Mark Brady (24:47.723)

Probably the most recent one that we ran and that was leveraging mid-journey to kind of really help us augment how we think about creativity. I think that was fascinating and to see the difference between how creatives used it and how marketers used it. And one of the really cool, just very quickly, one of the very kind of interesting things that came out of that is one of the larger challenges that we had with mid-journey was users were really struggling to like prompt in a very visually rich way.

to describe what was in their mind to get to the iPod that they wanted. That came really easy for our creatives because that's the world. They speak in visuals. It came much more harder for our marketers who don't work in the creative field. I think that was interesting. I think for me it's definitely anything where I can lean into innovative technology and see how that actually works. That always stands out for me.

Arjita, Artwork Flow (25:39.554)

great. So in that, you know, what's the one like funniest or the longest prompt that stood out like that? You were also like, okay, what is this?

Mark Brady (25:50.367)

Oh my gosh, I don't think there's anything that stood out for me in particular, but I think what was fascinating is we started really toying with our brand in very nuance and new ways. That could be something like, what would an office look like on the moon? That's not something that you'd spend time as a designer kind of trying to do it. But when you have this ability to like rapidly create that, it allows you to kind of think outside the box.

Arjita, Artwork Flow (26:06.497)

I'm going to go to bed.

Arjita, Artwork Flow (26:17.016)


Mark Brady (26:17.423)

So we saw some very interesting outputs that were very much out of this world.

Arjita, Artwork Flow (26:22.13)

Yeah, that's actually that is the most fun part about AI just to like make content or make images that are totally out of the world that you can't really even imagine or fathom in the first place.

Mark Brady (26:35.167)

It's also one of the downsides though, because you spend so much time in that fun ideation space, as opposed to like, I actually have a job and I need to create content.

Arjita, Artwork Flow (26:44.158)

Right. Great. What are some of the tools that as a creative ops professional that you can't work without?

Mark Brady (26:53.403)

Ooh, I personally, I use Lucidchart a lot. So in my job, I focus on the strategy and the overall vision for where my group's going. And I think getting the whole team into a collaborative whiteboarding session is great, especially in today's remote world. We don't have a whiteboard that we can draw on anymore. So Lucidchart for me kind of stands out. I probably wouldn't be able to do my job as effectively without it.

Arjita: So most of your team is still working remotely.

Mark Brady: Most of our team is still working remotely. Our team's also quite globally dispersed. We have East Coast America, West Coast America, UK, Ireland, Columbia. So we're all over the place, in different time zones. And the majority of our folks work hybrid or completely remotely.

Arjita: Okay, interesting. Got it. What's the most interesting book that you've recently read?

Mark Brady: Most interesting, what sorry?

Arjita, Artwork Flow (27:51.95)


Mark Brady: Book. Interesting. Good question. I'm reading a book at the moment called The Billionaire Who Wasn't. It's about Chuck Feeney who gave away all of his money to charity to help education initiatives across the world. But he did it in a very unique way that he didn't want any credit for. So it's kind of fascinating reading into that mindset.

Arjita: Interesting. What's your favorite city in the whole world?

Mark Brady: Ooh, I'd be killed if I didn't say my home city, which is Dublin. But I think I've been very fortunate to travel a lot around the world. And I think I really like Boston, actually, as it happens, where HubSpot is based, because it's a great mix of like, it's not too busy. There's a lot of Irishness there, and obviously I'm Irish. So there's a lot of like ex-patriots that live there. So it's a great city.

Arjita: Interesting. I also really like Boston like there's a river and it's so green and there's also that old-school charms. It's quite nice

Mark Brady: Yes, absolutely, but it's data. Yes, absolutely.

Arjita: Great. So I think that brings us to the end of this conversation, Mark. Thank you so much for taking out the time. I learned a lot from our conversation. And as always, I'm a big fan of HubSpot, so keep doing the great work that you all do out there.

Mark Brady: No problem. Thank you very much for having me. Thank you.

More episodes

Episode 8

Designing for Diversity : How Gojek reaches millions with it’s brand strategy

Watch now
Episode 7

Balancing Data and Intuition to Lead Large Design Teams

Watch now
Episode 6

How to scale a hyperlocal brand through 360 degree digital campaigns

Watch now
Episode 5

The secret to running 100+ successful luxury brand campaigns in a year

Watch now
Episode 4

How to build a high-performing creative team

Watch now
Episode 3

Chilipiper’s spicy approach to anti-boring content

Watch now
Episode 2

Crafting Solutions in Creative Operations

Watch now