Marketing in the Pocket of Every AI & Employee: How SPS Commerce Built a Brand MCP

Julia Nimchinski:

Thank you. Thank you. And with that, we want to welcome Maria Bergolino, who leads Marketing and SPS Commerce. And she helped scale companies, including ActiveCampaign, Marketo, Conga, and Anaplan. Super excited to host you, Maria. She’s gonna be joined by Tuba Duraz, the founder and CEO of Amoeba AI. Huge pleasure. How are you doing?

Maria Pergolino:

If you meant that for me, I am doing great. I’m so excited to be here.

This is the first time I’m talking about something pretty special that we’ve built, and I think… unique from some of the path of others, so excited to share a little… I’m hoping for a little bit of grace as I, try to simplify maybe something that’s a little bit technical, and hopefully put some power, in everybody’s, products and messaging.

Excited for everybody that’s joining, thank you.

Julia Nimchinski:

Very excited. Tuba, take it away.

Tooba Durraze:

Yeah, thanks so much, Julia. I am, very excited to hopefully nerd out a little bit today, with Maria, so thanks so much for being here, Maria.

I think for folks in the room, kind of the context of the conversation today is, like, we’ve been talking about all sorts of Stacks in AI agents, so, like, systems of learning, systems of understanding, systems of intelligence, etc. This is, researching Maria.

This is my first example of tangibly something that cuts through the entire stack, essentially.

So I’m excited to dive into it, because I think that, practically how these stacks apply, like, make a lot of difference, and I think in this case, talking about how you have thought of brand, not just kind of, for obviously humans, but the world that we’re headed for agents, is exciting.

You picked a very, like, hairy problem to solve, a hard problem to solve, and based on what I read, have solved it well. So, maybe let me, pass it over to you to give a quick introduction to yourself, and then we can kind of kick off with some questions.

Maria Pergolino:

Yeah, and I appreciate the fast introduction that was already done. My name is Maria Perdolino, I am the Chief Marketing Officer for SPS Commerce. which, connects brands, growing brands, like all the ones that you’re surrounded by, with retailers. So when you walk down the aisles of Costco or Target or Walmart.

or maybe any grocery store, those shelves are filled because of the connections that SPS makes with those brands and those retailers. So, an exciting mission there, and so the topic we’re going to talk about today is not only helping SPS, but helping make sure all those growing brands are successful.

Tooba Durraze:

Yeah, it’s, like, the perfect example of, like, a one-to-many-to-many-to-many, essentially. Let me start off with, like, something easy. Obviously, you’ve had extensive experience in this field, but have seen, the field go through a lot of different types of waves, etc. I’m curious on… obviously, we know AI is, like, a different kind of shift.

But, in your framing, what does it feel like, based on, kind of, everything you’ve known and, like, where we’re headed now? for marketing in general, I guess, is, like, the question.

Maria Pergolino:

Yeah, I mean, that is, and maybe a good intro to our whole topic. I think right now, there is a lot of people in technical roles, engineering, just non-marketers. driving this new world of AI, and I think the fast reaction was, oh my gosh, maybe we don’t need marketing anymore. But I truly believe that marketing is the tip of the spear.

It is the strategy of a company, it is the connection of a product with that customer, and so… as I was thinking about AI, it was not just, what can you replace, because ultimately, AI is a equalizer. If you can create it, I can create it. where is the differentiation? How are we going to stand out as brands?

How are we going to let a customer or prospect know what we’re doing that’s different, and the value of them separate from anything else that AI can create? And so, in that framing, realize that marketing had so much to offer here, it just wasn’t framed in that way.

We were looking for all of this optimization instead of where could the growth come from. And so, I’m excited about AI because I think it’s forcing a lens of marketing that really says, what is the purpose of the function, and how are we going to create a difference? As everything around us is changing.

Tooba Durraze:

Yeah, that’s a great way to put it, because I think we’re also… seeing that, everyone talks about, kind of, like, brand, taste, etc. as being, like, the differentiator, essentially, but basically, in my head, it boils down to, like, the art of the craft. Like, there’s a real craft there, essentially, that gets, like, lost sometimes or abstracted.

You… there’s something that you talk about, which is essentially that, when you lay AI on top of AI on top of AI, so we’re talking about kind of, like, taking something basic and kind of layering things on top, sometimes what happens is it fractures the brand, if not done well, right?

To your point, if, like, you’re only building it from a technology perspective of solving for efficiency, you lose sight of, like, what you’re trying to do there. I’m curious in your world and kind of what you have seen, as you started to build towards this new future, where does that… where did you see that fracture kind of come up first?

And, How did you find it before folks were even trying to measure it?

Maria Pergolino:

This is such good build-up to how we’re going to show that we solved the problem, but essentially, as all of us rush to AI, maybe our, and it’s different for every company, but maybe somebody in your organization said, hey, wait a second, we can… do some automation of, sales development, or sales communications.

Maybe somebody said, hey, we can… create some new web pages, or we can create lots of blog posts.

Maybe somebody else in the company said, hey, we’re gonna, as a company, if you’re a software company, we’re gonna create our own AI solution, and it’s going to talk about us, or it’s gonna be able to reference us, maybe we’re building an agent, and so… there’s something that has to be said about the company in that, and so there’s all these places in the organization that people raced towards AI, and all of them needed, positioning and messaging.

They all needed context. And so, how did companies create that? They either, Dumped a bunch of files in.

and everybody kind of maybe went on their own path, and so now what you have is AI referencing maybe materials in the company, marketing may not know which materials, it may not know what is being, what’s being referenced, and so what that first AI SDR email goes out and says, maybe in conflict with how your product is talking about you, and then in conflict with maybe that new webpage that was created by AI.

And so, because there’s no way… we didn’t start with consistency in how we address these tools, what we started to see is fragmentation in message. Being, maybe dated products being referenced?

not the latest and greatest, and it’s a time when you need the latest and greatest, you need to be super clear on your message because the companies are changing because of AI. And so, it has created this, like, dilution because we’ve lost this crispness because AI is referencing lots of different places.

Tooba Durraze:

Yeah, and if the gravity of, like, what Maria is talking about is lost on people, like, generally when we talk about, kind of, context and brand awareness. folks are talking about, to your point, like, a lot of, kind of, documents or, like, flat artifacts, essentially.

You are talking about brand as its own, kind of, architecture, which is, like, a mind-blowing concept to me, because… folks don’t think about things in, like, terms of that layer. So, now, like, to tee up the thing, tell us what you have built, and I want to kind of pull on that a little bit through the rest of our questions.

Maria Pergolino:

Yeah, so, understanding that everybody was creating Markdown files or connecting to maybe a OneDrive or Confluence that had lots of information, because there wasn’t consistency in what was being referenced, or the way it was referenced. we saw a number of problems. One, that inconsistency in message.

Two, it could cause token leakage because people could be referencing an entire confluence drive instead of maybe the one article that needed to be referenced to write that message. We… we saw, no way for marketing to update anything, because we didn’t know it was referenced, and even if you updated the document that was the original source.

it may not… the person may not go back and grab that to update maybe their markdown file.

And so, what we did is we said, let’s stop thinking about this as marketing being the afterthought, and instead, how do we put marketing And when I say marketing, I specifically mean, the positioning, the messaging, and the brand in the pocket of every person in the company.

and maybe everybody outside of the company, depending on how you wanted to implement something like this. And so what we did, and I guess this is the drums roll, is we created, what we right now call our Brand and Messaging MCP.

And it is essentially a knowledge graph, well, it’s a knowledge repository that is referencing Markdown files, with a knowledge graph sitting above that.

And the idea is that it is a single source that we are updating that is then being leveraged inside our product, as we are, you know, using whatever LLM your company chooses, when we’re creating our TxDOT LLM, files so that we can rank properly. It is all coming from one source. And, while I said it starts with a set of Markdown files.

Instead of starting with this big dump of documents, we specifically built those files to be, referenceable by AI, so built in a smart way. And, with… Path to updating them. So, Essentially, it is… there’s now, for us. There’s 865 nodes in our MCP that cover 71 personas, 13 products. four, kind of, tone pillars.

I know we’re gonna talk about this, but a set of brand phrases, 55 different, like, term preferences, we have our writing standards, all of the typical brand assets, sitting in this MCP that can then be referenced, by whoever needs that marketing in their pocket.

So, a couple of examples of how this can be leveraged, in your product, if you have, an agent. it can speak in the language of your product, because it’s using the MCP. And when marketing maybe adds a new product, instead of having to go back and update the product itself, it’s just automatically updated through the MCP.

For, let’s say, a developer, let’s say a developer is using Cursor, or Codex, or whatever tool they’re using, they can reference the MCP, and maybe they type the previous name of the product in what they’re building. the MCP will know to update that with the proper naming.

For, the marketing team, as we’re building materials, it can… we, we’re leveraging Claude. We use this MCP both as an MCP that’s, brought into the product, but also, translated into a skill that can be referenced, so that it’s, using that same language. And then, again, sometimes I get the question, why not just build this as a skill in Claude?

Why have it as an MCP? But you wouldn’t get that benefit of all the places you can apply this outside of Claude. The MCP is something that can be plugged in lots of places, and so it’s become that single source that is allowing us to put that marketing in every kind of pocket.

Tooba Durraze:

Yeah, totally, and, like, that is also your IP. Like, owning, kind of, your IP of that, I think, makes a world of difference.

The one piece that also stood out to me is… how often on the product site, sometimes, like, in-product copy, it doesn’t necessarily match how you’re talking, so I can even imagine, for your engineers as well, like, that’s, like, a really good… reference point.

As you talk about this graph, and you talked about, kind of, all the things that are in this graph, including, kind of, personas, and, like, products, and phrases, and all of that. First question I have is, how do you pick what gets included in the graph versus not, and who is responsible for being the Sherpa who picks that?

Or, like, a team or whomever, yeah.

Maria Pergolino:

Yeah, so, this was definitely a decision and commitment by the marketing team, not something that kind of happens on the side, because, we had to… re-look at every… all of our documents, and really start talking in a language of AI versus writing for the human.

So, having concise documents, putting things in, clear tables, making sure that it was… there was no conflict. We weren’t using a lot of descriptive words to kind of bolster something. The LLM or AI could apply any of that, that was needed. It was really about getting clear on the descriptions of the personas, on the products, on the brand.

And putting that into Markdown files that made sense, and then building this knowledge graph above it that could cross-reference and put together something that was better than those individual pieces on their own. And so, one, all of product marketing had to re-look at their content and And start speaking in the language of AI.

As we think about maintaining it.

We, are expecting to have a full-time person on the marketing team that is ensuring that it is consistently updated, that it is maintained properly, and I know that that’s a lot of commitment, but as we think about how we continue to leverage AI, it becomes critical, as we think about, also the way that other systems will be able to plug into RII.

So right now, some of the things the marketing team’s doing is speaking with the vendors that we, we integrate with, and looking for them to accept the MCP. A very simple example of this is we use a software that does competitive monitoring. When they send us recommendations, they do, like, a weekly summary. It’s very generic.

We want them to plug into our MCP so that when they bring it back to us, those… that summary, it is contexted by our business and by our products.

And so the… you think about the value that that adds, the time that that saves, it’s worth the investment within the marketing team to have this MCP that can be referenced think about how that applies to every part of your marketing and every part of your stack. It is a worthy investment.

Tooba Durraze:

Absolutely, and if you see all the investment that’s going into, kind of, like, semantic layers and ontologies, like, I feel like this is what you’re talking about is such an important component that sometimes gets looked over, because… Yeah, this is a part of the ontology, the context of, like, what’s happening in an organization. Very cool.

One… of the things that happens a lot is, like, we always tell, kind of, when we’re creating these systems and neural nets, etc, sometimes, in my level, knowledge graphs, or even flat files sometimes, we have a tendency to, frame them as, like, what is X? not what is not X, right? So what we are not is, like, equally as important.

You… there’s a very interesting thing I found in my research. It’s this concept of, like, banned phrases. So, like, basically encoding something that’s negative is a better way to train the system. Can you talk about that a little bit?

Maria Pergolino:

Yeah, and super interesting. Maybe if you’re sitting as an engineer listening to this, this becomes more obvious, but if you’re on the marketing side, we’re used to having brand guidelines and really thinking about how to assert our brand. There’s not as much of a sense of how to suppress maybe other brands.

For example, some of the brand phrases or common words that our competitors might use, because AI may not have that sensitivity built in to how we would frame, or maybe we wouldn’t, you know, often, competitive tools might be named something that are common language terms, but we wouldn’t want them frequently referenced.

I think similarly, there’s ways that we talk about our products, where we might always say. I said here, I used the word knowledge graph, but maybe, maybe my company says, well, we always say data graph.

Well, I might want to suppress that word, knowledge, because AI might go to the more common usage versus the way that we’re bringing preference to it.

So, the way that I think about this is we have a brand guideline, but now we kind of have an unbrand guideline, guidelines that are trying to suppress the things that AI may drive towards, if we had just brought forward the general brand guidelines.

Tooba Durraze:

Yeah, that… to me, that’s actually such an… I’m not a marketer myself, so maybe my mind is a little bit blown on, like, how much the anti-space is, like, helpful to, like, compose the thing, because to your point, in math or engineering, these are definitely concepts, like, tell it what not to do, as well as, like, tell it what to do, essentially.

The other thing I’ll, again, like. flag to folks, kind of, who are listening in, is… The advantage of, like, how you’re describing, like, a system versus, like, an individual artifact, as a whole, and, like.

How you have designed and have treated it kind of like a living organism that is evolving as things are evolving, which is… kind of, like, you don’t want to end up in a world where it’s a couple of flat documents and someone’s, like, constantly, like, oh, I can’t remember the last time they were updated, versus, like, an entire system.

So, first, how… how long did the effort take you to get to even, like, an MVP of this? Because I think, folks in the room are probably like, okay, it takes… It’s a lot of heavy, like, lift up front.

Maria Pergolino:

Yeah, and just, even before that, all of the benefit you just said in getting to that outcome, this conciseness also is reducing hallucinations, which… And it is conserving tokens, because it’s so concise. And so there’s a win in every direction.

So you think about, oh my goodness, you’re gonna put a full-time resource to that, but I think we’ll save that resource in token, efficiency, because everybody’s not referencing these massive data sets. or trying to throw in, oh, read through, you know, to their LLM, read through these 10 PowerPoints to get to this information.

And so, there is this… there’s… it’s goodness in every direction. Like, there’s just not a negative to this approach, because it ends up paying for itself, too. as you… oh my gosh, I can’t think of, what your question was now.

Tooba Durraze:

My question was, like, how long did it take.

Maria Pergolino:

Oh, yeah, yeah.

Tooba Durraze:

I feel like an MVP of this.

Maria Pergolino:

So, just like anything else, so we have, I haven’t met somebody else who had another marketing team that had taken this approach. I just had the benefit of sharing this with some other CMOs, and I had, where CMOs had kind of tried to bridge this, it was with a lot of complexity.

The other thing I’ll say is this is… We spend a lot of time thinking about how do we make this the least complex? So that it was going to provide the value, but also be something that was not so complicated that, hey, if a team member changed, or something happened, that it was going to be impossible to keep forward.

It did… we started working on this in… the end of last year, and it is still a work in progress as we continue to update it. What I will say is if you’re listening to this, and everything makes sense, and this is going to be your path forward. it’s beyond just having somebody technical on the team build this MCP with these materials.

It does take, I think, a decent amount of leadership. When I, I just sent a note to the team today for an upcoming launch, I said, I don’t want to review the materials in a PowerPoint, I want to review what’s going to go into the MCP.

As I think about understanding both our messaging and our brand, I really want to, instead of saying, here’s our brand, and this is where AI got it wrong. my push on the team is that the AI… whatever AI puts out is what the brand actually is, and if somebody feels uncomfortable with that or disagrees with that, we have to get AI to better.

And so, what I would say is beyond just the, the… the work that you have to do here, it does take leadership not only in vision, but in really saying, we’re gonna think about this differently, we’re gonna do this differently, and… I, as a leader, am going to break my habits in how I’ve reviewed content, and how I got updated, and the ways that I asked for things from the team, so I really am experiencing it through the MCP.

Tooba Durraze:

Yeah, I think that’s a critical point that you’re bringing up, which is, like, the system around things needs to change as well, operationally, versus just, like, okay, I’ve now built this, like, behemoth, and… You can’t leverage it. And, like, the mindset as well.

So, very tactically, what changed for you operationally once the system started getting put in place? Like, how do you conduct your team now? How do you interact with other teams in the house?

Maria Pergolino:

Yeah, I mean some funny things.

I can now talk in nodes and edges and node types and, and all the things that the knowledge graph is built from, which is not what I was expecting as CMO, but has become important as I try to make sure that what is coming from AI, again, whether it’s from marketing, whether it’s from a third-party tool that’s serving us, or… Whether it’s our own product delivering for our customers, how do I make sure that it is, representing us properly?

And so, it has been worth the shift. Again, I do think it is about setting the vision, and not only… it is a team-wide effort. The product marketing team, the brand team. really has to be there with the content, but it has to be matched by the operational team that is going to piece this together.

I’m so proud of the team and its alignment with our engineering team, because we will treat this like a, like, a marketing-created product, and so it’s living in our AWS instance, and that needs government and security, and then, And so that relationship there is critical, and I couldn’t say enough about the SPS engineering team and making that possible.

And then that leadership layer, where you’re bringing clear vision to this, you’re changing your own path to make sure that you can, Keep this thing living, and that it’s… you’re making that change in behavior.

Tooba Durraze:

Yeah, I feel like the unintended, to your point, positive consequences of something like this is, like, there are more synergies between teams, where traditionally they might not have been, and it’s kind of… you become a little bit, I would say, a lot of it more central, essentially, to how everything, is happening. To that.

I think in your vision, and correct me if I’m wrong, it’s basically, marketing becomes kind of the fuel that’s powering AI-related things, like, around your organization, versus… Some of the perception has been that, you know. marketing… the death of marketing is being caused by AI. Like, I’m generalizing now.

Maria Pergolino:

Yeah, fair.

Tooba Durraze:

Yeah, so give me thoughts, because I think your perspective on, like, it is fueling the revolution is, like. Probably, I feel like, in my head, aligns more with where we see things going as well.

Maria Pergolino:

Yeah, I mean, again, AI is only going to be able to like, AI for me and AI for you is only going to produce the same thing without additional context, and what we’re doing is creating that. context.

We’re making marketing a differentiator for the organization, and I expect that SPS will continue to be the leader in space because we can bring that clarity And that this… this does allow marketing to continue to be that strategy. At the same time, we are looking for those efficiencies that we can drive.

Through marketing, but… I think that marketing has become more important to the company because of now enabling it through every employee versus less important. And so, I’ve really challenged the fact that marketing becomes less important.

I think that it is, the good thing is, I think that brands that kind of have that view are going to struggle with differentiation. It may not exist, if they take that, and so marketing will still continue to reign.

And I know that that’s, that’s a little bit of a provocative thought, but the, you know, we… I just constantly think about brands, and you look at the… their values, and it’s not in what they… they have, it is their IP and their brand on top of that, and… That, that brand really makes a big difference.

Tooba Durraze:

No, I agreed. I think… I don’t think, by the way, this is a controversial take, because, like, the math in my head says if there is no demand, then what are… what are you selling? And, like, all of that kind of centers around efforts that you would do in marketing. I feel like marketing does sit central to me.

we like doing things a little bit more practically here, so obviously these are, like, really big, behemoth things that you’re doing, you’re leading the kind of charge on. What is something practical, a practical piece of advice that you can give CMOs, on something they could do in the next 30 days to start to impact this piece?

Maria Pergolino:

Yeah, even if you weren’t going to build out the MCP the way that we had, I think just understanding where your brand is being leveraged, and getting a sense of how the… these other systems are talking about you in the absence to understand, hey, is this a change that you want to make?

I think beyond that, just continuing to be very crisp in all the things I talked about today, those personas, the segments, the products, the… all of that. positioning and messaging, continues to be important, and if you don’t have that strong, it’s really hard to put something like this in place.

So, continuing to iterate there, even if it’s… you’re not ready for it, in such a formal and composed way.

Tooba Durraze:

Yeah, it’s almost like going back to, like, first principles, like, the basics need to be good. Okay, and then personally for you, since you’re kind of, I would say, forward-looking in this direction, Let me ask it a different way.

Not some… I’m assuming there’s a lot you’re excited about, including nodes, which makes my heart very happy, because I love talking about neural nets, but what is something that you have maybe changed your mind on, related to marketing or related to technology or AI over the last couple of months?

Maria Pergolino:

Yeah, I’ve talked a lot here about the leadership that we have to… to bring, and this is doing something pretty different for a marketing team. I think. I’ve really tried to, while empower each of the individuals on the team with AI, I’m really putting my focus into where it breaks function.

And where we see gains that isn’t just in one area, but, you know, something that is gonna impact, the way that we currently do a process, and get more out of. You know, not just a siloed initiative, but across functions.

So I’ve really had to challenge myself in rethinking the standard ways that we organize, or the standard ways that marketing may work with sales, and say, how do we work across these? And that has been really exciting and really challenging, but also really hard.

Tooba Durraze:

Great. Well, thank you so much, and thanks for letting me geek out over all those questions. I’m sure people have a ton more questions, I have a ton more as well, but it was a pleasure chatting with you today.

Maria Pergolino:

Yeah, so fun. Thank you so much.

Tooba Durraze:

Yeah, of course.

Julia Nimchinski:

Maria Tuba, this is amazing. Maria, before we let you go, just one second, if you’re still here, the community is curious, what’s the best next step? How can they test this?

Maria Pergolino:

Yeah, I mean, I think the, Just, one of the things we did with the team to get buy-in that was so interesting is we had taken, we had… we had a big marketing team meeting, and we had some people… we asked, we had everybody bring up, kind of, their own, in this case, Claude instances, and show the output that they got with their own Markdown files.

And then we used, the mock-up of the MCP to do the same, and we were able to show, like, how much better, how crisp the response could be when it was so well organized, and we had the right information, and that really drove buy-in that this was a worthwhile effort.

So, I think just getting creative with the teams, letting them see the impact that can be created, and then if you’re getting friction outside.

of the marketing team, I think doing the same, you know, challenging that the engineering team, maybe say if there’s an agent, let’s ask it a question about the company, and now let’s ask it through the MCP, and just showing that you can get something that’s much more powerful, and show how it’s going to yield for the company.

Julia Nimchinski:

Thank you so much.

Maria Pergolino:

Thank you.