TRANSCRIPT

Claire Ebben – Qualified
Hi, everyone. Welcome to Agentic Marketing Summit. I’m Claire Ebben. I’m the Senior Director of Product Marketing here today.

Very excited for this session, Build or Buy: Choosing the Right Path for AI Agents in Marketing. The goal of this session is really for you to learn from some top marketing leaders about how to customize AI agents, when to leverage proven out-of-the-box solutions versus building your own agents.

So my speakers today, very excited, we have Angela Ferrante. She is the Head of Enterprise Marketing at Zapier. And Alexi Hatch. She is the Vice President of Global Growth Marketing at Quantum Metric.

So let’s kick things off. I’d love to start just with this first question around what key considerations really drive the decision for you both to build a custom agent versus buying an existing solution. And Angela, I’ll kick things off with you.

Angela Ferrante – Zapier
Yeah. So first, just some context on why this is so important. We are in an era where building almost anything becomes possible. You know, we’re all seeing the iVibe-coded Calendly-in-an-afternoon posts on LinkedIn and Twitter X every single day.

So you can replace a lot of SaaS tools and point solutions. It’s much more important to have a thesis on when to buy versus build. So my North Star on this is build for differentiation and buy for velocity.

So if you have a competitive advantage of this core IP or workflows where the outcome itself is a unique advantage, great time to build, versus buying when you just need predictable ROI, speed, and scale.

An example for us is we had our annual user conference, Zap Connect, back in September, and we had very specific lead flows for AI boot camps that we were running after that event. We built the solutions that would follow up with all of those different participants because there was a lot of core product data that we wanted to incorporate.

And I think there are three pitfalls I see when people think about do we build, do we buy. One is cost savings. Now, obviously, we care about cost, but when you hear folks say things like, “We can save thirty thousand dollars on this SaaS tool. I just have to spend the next two weeks building this thing.”

What’s the cost of that, right? Is that actually costing you three million dollars in pipeline? Guess what? That’s probably not worth it.

So cost savings is a big one. Starting too complex is another. People try to build these really mega systems out of the box. And then the third one, and I haven’t heard anyone talk about this before, is not using the agents that you build.

Maybe this is obvious, but I think the same way that sometimes you buy a new software tool, you need onboarding, enablement, and change management. I see tons of agents that get built that are really cool and really high potential, and they basically just get put on the shelf because no one knows they exist. They don’t get built into existing workflows.

Claire Ebben – Qualified
Hundred percent. Alexi, how about you?

Alexi Hatch – Quantum Metric
Yeah. You know, it’s funny. I have almost the exact same kind of thinking and litmus test here. I say, if it’s a differentiator, build. If it’s a multiplier, buy. The same thing that Angela’s talking about.

So for me, the decision always comes down to speed, integration, and ROI. Marketing just moves way too fast for really long development cycles. So if you need to show value in a quarter or two, an out-of-the-box solution or agent is usually the smarter move.

We looked at building our own SDR-style agent internally, but the reality was we just didn’t have the technical bench to maintain it, train it, ensure compliance. Buying ultimately gave us reliability, a faster path to market, and the ability to layer on our own unique data rather than reinventing the entire core system.

So I agree with everything that Angela’s talking about.

Claire Ebben – Qualified
Yeah. I mean, it’s such an exciting time to be a marketer in this agentic era because obviously you can become engineers and architect your own workflows and things you want to achieve. But there are also all of these pitfalls that are almost lying underneath the surface that marketing has never really had to think about before.

Almost like this gigantic iceberg of stuff. And Angela, you touched on some of these, like cost savings, but also the potential that you might not even use what you’ve built. So it’s great to hear you both are aligned in that thinking.

Can you both share with us any agents you’ve built for your marketing team? I know you touched on a few, but I’d love to double click on any agents you purchased out of the box. Alexi, maybe start with you.

Alexi Hatch – Quantum Metric
Yeah, sure.

Right now, most of where we’ve chosen to keep our time is really in the buy category. It’s about multiplication. Where do we multiply what already works? Our AI SDR is an excellent example of that.

Where the build conversation is getting more interesting, and likely where we’ll invest next year, is what I’d call the new marketing ops layer.

Anything that’s been traditionally automated with simple if-then logic, like account assignment, lead scoring, campaign tagging, QA, those are ripe for an agent. They don’t just follow a rule. They decide based on context.

They’re more reliable, and they raise the bar. Agents will actually make perfection the expectation because they can catch so many errors.

If you build the right agents and you’re using them, they raise the bar even higher, where standardization and overall flow work better because so many things are happening in the background.

Claire Ebben – Qualified
Yeah. Angela, how about you?

Angela Ferrante – Zapier
You may or may not be surprised to hear that we build a lot of things in-house at Zapier.

We do AI orchestration. In our product, we have workflows and agents, and you can mix the two. We also have a copilot that lets you describe what you want. You can say, “Build me an agent that reviews my calendar weekly,” and watch it build in real time.

This is a very frequent prototyping pattern for us. I’ll experiment quickly in five or ten minutes between meetings to build something off the shelf for a specific pain point.

I think about AI work in three categories: faster, better, and work that was never previously possible.

Across those, we have three main categories of agents. One is executive leverage agents. We built an executive coach agent that scans weekly calls and notes and makes suggestions for leadership improvement.

The second is team productivity agents, like an Inbox Zero agent or calendar management. These are fleet agents that give productivity to everyone.

The third is functional expert agents. For product marketing, we have an agent pod that reviews competitors and writes launch copy. They’re always on, connected to messaging guides, competitive briefs, and internal docs.

On the buy side, one example is our support bot using Ada. We could build it, but it wasn’t a differentiator, and speed mattered.

Another example is Profound, which we use for GEO and AEO to see how we’re appearing and optimizing our presence in LLMs.

Claire Ebben – Qualified
I love that. Well, first, those product marketing use cases, very exciting for me as a product marketer. We’re starting to do a little bit of that at Qualified as well, scaling internal knowledge that product marketers have and extending it to the entire team.

I really liked how you both talked about building faster internal use cases for delighting teams, while also vying for more external use cases.

I want to talk a little bit about cost now. How do cost and speed really influence your build versus buy framework in marketing? I know we touched on this at the top, but Angela, I’d love for you to double click into how you think about this.

Angela Ferrante – Zapier
Yeah. I know I mentioned cost savings being a pitfall. That’s not to say you should never prioritize cost savings. Obviously, that’s real and important.

But I think we can get trapped in, “Oh, we’re saving so much money on SaaS spend.” Instead, pay attention to intrinsic cost, total cost of ownership, and include onboarding and enablement.

There’s also hidden cost and opportunity cost. If you’re building internally, is this really a core competency? What else could you be doing with that time? What level of pipeline could you be delivering?

Another thing I think about is what unique advantages a solution provider has that you might be missing out on. That can be real and compounding.

On speed, it’s about velocity of product improvements. If we do this internally, are we going to upgrade it fast enough, or will a tool like Profound be adding functionality so quickly that we can’t compete?

Those compounding improvements and costs are very real.

Lastly, prototyping quickly internally lets you evaluate whether something is worth maintaining long term. We use an effort and impact matrix. Effort is on one axis, impact on the other.

The low-effort, low-impact quadrant is actually a great place to start. Lower the stakes, learn, build skills, get something out quickly, and then move into low-effort, high-impact refinements.

Claire Ebben – Qualified
Yeah. Smaller wins first, then gaining confidence as you go. Marketing doesn’t have the luxury of waiting long to see ROI.

Alexi, can you touch on cost savings and how you consider build versus buy?

Alexi Hatch – Quantum Metric
Time is almost always more expensive than software. People’s time is very expensive, so how do you maximize it?

I look at time to value. How fast can an agent start driving outcomes or solving problems? If it can be built fast and deliver quickly, great. If it requires heavy technical implementation, less so.

I also look at integration debt. How many systems need to plug into this, and who owns maintenance? Is this replacing something or adding more work?

Lastly, ongoing optimization. Who owns optimization? Who manages it long term?

Roles are changing. Marketing roles look wildly different than they did six months ago. It comes down to how people are spending their time, and that’s a huge cost factor.

Claire Ebben – Qualified
Ongoing optimization is part of that iceberg marketers don’t always think about. And if the agent is owned by IT, you can lose control over its lifecycle.

Time to value matters. Six months to ROI is over. It’s ninety days or less now.

Let’s shift gears to personalization. How do you balance personalization with scalability when choosing an agentic approach? Alexi, I’ll start with you.

Alexi Hatch – Quantum Metric
Personalization and scalability are everything for us because Quantum has an incredibly matrixed audience. Many buying units, industries, and use cases.

It’s overwhelming for sales teams to stay relevant. Agents, especially our AI SDR, allow personalization at scale because they understand the matrix.

It’s a force multiplier. It makes personalization possible and doable.

Claire Ebben – Qualified
Scaling personalization with humans means smaller output. Agents make it happen simultaneously.

Angela, how about you?

Angela Ferrante – Zapier
When I think about personalization, I think about vanity personalization. Yesterday’s AI SDR work.

Real personalization is about relevance. Not “I see you’re in San Francisco,” but what actually drives impact.

Zapier has deep product usage data. That lets us personalize meaningfully, like how many AI tasks an account has used, which teams are active, and how they benchmark against others.

Buying an AI SDR gives learnings on repeatable engagement, and then you layer in custom signals where it matters most.

That hybrid approach brings real value.

Claire Ebben – Qualified
It’s not “Hi, first name” anymore. It’s meaningful personalization humans can’t do in real time.

Let’s talk about how agents impact team roles and marketing transformation. Angela, how are you seeing roles change?

Angela Ferrante – Zapier
We’re early. Most teams are still measuring activity, which is fine at first.

Eventually, it moves to impact, and metrics like revenue per headcount come into play.

Teams need technical acumen, even if it’s not traditional technical skills. There’s a willingness to experiment and be bad at something first.

Human creativity and intuition become more important as AI scales execution.

Claire Ebben – Qualified
Alexi, how about you?

Alexi Hatch – Quantum Metric
Agents don’t just automate work. They rewire workflows.

Everything changes, from content production to QA to research.

We invested early in AI enablement. Try it everywhere. If it works, great. If not, move on.

We saw workloads drop by eighty percent, freeing teams to try new things and manage agents.

It’s fun and incredibly challenging.

Angela Ferrante – Zapier
One thing this brings up is AI fluency in hiring and performance management. We now evaluate AI usage from unacceptable to transformative, broken down by department.

Claire Ebben – Qualified
It takes leadership to make that happen. We’re also seeing new roles emerge, like agent managers in marketing ops.

Final question. What best practices would you share when approaching build versus buy? Alexi, I’ll start with you.

Alexi Hatch – Quantum Metric
Buying introduces risk. Data privacy, model transparency, where data goes.

We have a robust vendor review process now. Vendors know brands care about this.

There’s also risk in setting it and forgetting it. Agents are living systems. They need tuning, guardrails, and validation.

Accountability doesn’t go away. It just changes.

Claire Ebben – Qualified
Angela, any final lessons learned?

Angela Ferrante – Zapier
Three best practices.

First, have a framework for build versus buy, tied to impact metrics.

Second, have a plan to ensure what you build gets used and maintained. Give people time to build and improve, not just “do this too.”

Third, governance. Verified sources of truth, shared context, and human-in-the-loop judgment.

Agents scale thinking, but humans own judgment.

Claire Ebben – Qualified
It has to be maintained and never set and forget.

Thank you both so much for sharing your perspectives and insights today. It was incredibly valuable, and we really appreciate it.