Agentic Orchestration: Are Marketers Having this AI Conversation?

I just returned from a great few days in San Francisco, attending a conference where I met people on the cutting edge of AI technologies, including people from Anthropic, OpenAI, etc. I met with people in all industries, especially those in the healthcare industry (with health systems, payers and health tech organizations).

Across all of my conversations, I encountered one dominant theme: Agentic Orchestration.

Last year, it felt like some people were slightly ahead of this conversation and that it was still a bit forward-looking. This year, it feels like it’s the only conversation people are having. Those who aren’t actively working toward it are now suddenly behind.

Now that I’ve expanded my aperture beyond marketing into data circles, it makes sense that people are talking more about agentic orchestration because data is at the foundation of what agents need to access to be successful.

But when I talk to marketers, I haven’t heard much about this particular theme. It’s not entirely surprising given marketing’s traditional role of interfacing with the consumer rather than dealing with internal processes. But right now, it’s as though AI in marketing is viewed strictly through the lens of execution and interface: smarter copy generation, faster asset creation, more intuitive chat interfaces, and “how to show up in an AI Overview.” The marketing conversation continues to be heavily focused on the front-end user experience.

Now that I have a much deeper background in data (and a clear view of how that data feeds business functions) it’s obvious there is a lot more going on behind the scenes internally, as well as in how consumers find information externally.

This week, none of the conversation centered around individual generative AI tools. Instead, it focused almost entirely on the orchestration of agents to surface the data that feeds those tools. Data is the foundation, because without it, the agent has nothing to reference.

While I was thinking about this theme, I realized I’ve actually been speaking and writing about the topic of data foundations for almost a decade. For over eight years during my time at Yext, we talked constantly about building knowledge graphs that connected entities to one another, and then connected those connections to the places where people were seeking information, like Google.

That need for a data foundation hasn’t changed. Over the last few years, it’s actually gotten even more important. Conversational platforms like Google’s AI Mode, AI Overviews, Gemini, and ChatGPT need to be able to access information quickly and easily in order to summarize a response to a user’s question.

“Digital marketing” isn’t about the user experience or how individual pages look on a newly launched website anymore. It’s about the ability of a machine to access data living in different platforms that serves to surface that data in entirely new ways.

But data doesn’t live in one place, and it never will. That makes it more important than ever for a machine to be able to access and compile that data to deliver the right answer. This is exactly where the agentic orchestration concept comes into play, and why it is so important right now. Consumer tools are taking advantage of this just as much as internal enterprise tools are. It’s critical to know how it works, but also how to ensure agents can actually access your data.

To help describe this, I asked Gemini for an analogy to explain agentic orchestration, and it came up with a great one:

Imagine someone is cooking a meal for herself. She knows what she wants to make, has only the ingredients to make that specific dish, and is only preparing it for one person. She assembles the ingredients, cooks the meal, and it serves one single purpose with no leftovers.

This person cooking for herself is like the traditional chatbot. In the past, a single chatbot was only capable of doing one thing: whatever it was specifically set out to do. Remember how we used to put chatbots on our websites and they could only successfully do one thing (like find a specific doctor, but only if asked in a very specific way)? People would ask them all sorts of questions, only to be disappointed when the chatbot returned a poor response.

Agentic orchestration is completely different.

Instead of a single chatbot only being capable of delivering one thing, we can now take advantage of a master chef coordinating a complex, multi-course meal for a single customer. Instead of cooking every dish herself, the chef takes a patron’s order, breaks it down into individual components, and delegates the tasks to specialized line cooks—one for pastry, one for the grill, and one for sauces. The master chef then coordinates the timing and manages the flow between each station, assembling their individual outputs into a single, perfectly timed feast.

Now, we have the capability to leverage agents to help us tap into different data sources that live in entirely different places, surfacing an outcome that far surpasses what that single chatbot used to deliver.

On the consumer side in healthcare, if someone is looking for health information, Google (or Gemini or ChatGPT) acts like that master chef. Instead of a patient conducting research and reviewing five different websites, blogs, provider profiles, and directories, they simply ask the platform for what they need. Google then breaks that request down through what they call “query fan out” and information agents. It delegates specific tasks to specialized AI line cooks: one to analyze clinical research, one to review condition descriptions, one to check local provider availability, and another to verify insurance coverage. Then, it seamlessly assembles the final answers into a single, trusted health recommendation, potentially books an appointment, or schedules a follow-up visit.

On the enterprise side, organizations can now surface information to data analysts, business users and others who may need access to data and insights. Imagine you needed to run a colonoscopy campaign: what if you could take data inputs from different sources to create better patient acquisition campaigns because you can get data from website visits, public health data about at-risk areas in your market, EMR data, provider availability and the like to surface the right list of people you could target to receive colonoscopy reminders?

Traditionally, a marketing team might pull a static list of patient emails, write a generic reminder, and blast it out. With agentic orchestration, a central marketing orchestrator coordinates a squad of specialized agents to access all of the data points I referenced above to run an intelligent, ongoing campaign. It’s just a smarter, more targeted and ultimately more efficient way to market.

It’s very nerdy. It’s very technical. And it is a massive departure from the old ways of marketing. It requires being up to speed on a lot of elements that are going to make a lot of marketers uncomfortable.

Marketing is getting really technical. Really fast.

Have you had a chance to step away and take a deep breath to think about how you could leverage agents to level up your marketing strategy?

Technology is moving faster than I could have ever imagined. But I think it’s important to come back to some core questions. While everyone is talking about agentic orchestration (and you should be thinking about it, too) it all comes down to data. If you can answer these questions, you will have a leg up on everyone:

• What are your marketing goals?

• What kind of data do you need to execute a particular goal?

• Where does your data live? (And I mean… all data you might need for that goal)

• How can you access that data?

• Where are the bottlenecks to accessing this data?

• Is this data accessible by agents (machines) today or is it stuck in a silo?

If you don’t have answers to these questions, it’s time to start finding them. Talk to your data and IT teams. Establish relationships with them so that you can foster an ongoing, trusting partnership with the team that ultimately controls access to this data. (I talk to them a lot and I can confidently tell you that they don’t consider marketing when they are making data decisions. You have to change this.)

The marketing function is quickly becoming a technologist function. Without understanding the value of data and being able to speak the language of technology (APIs, etc.), this world is going to move far past you.

This isn’t even about winning in search anymore. It’s about winning in a world where a person might not be doing the searching, but a machine will. It means ensuring your data is cleanly and clearly accessible to autonomous back-end orchestrators.

While this sounds like a death knell to marketing, it’s not. It just represents the foundational blueprint for how consumers will interact with brands and how complex marketing ecosystems will scale over the next decade. For marketing leaders, understanding this shift is essential. It moves us past basic automation and into an era where data infrastructure and market strategy completely merge.