Will AI agents conduct the martech orchestra in 2025?

“If the technical group was an orchestra, which product would be the conductor?” asks Martec Chief Scott Brinker in his first column of the year. To cut to the chase, Brinker believes that artificial intelligence agents have an important role to play here.

This will certainly be the case for some companies – and almost by default for companies that rely on suppliers who have taken the middleman route themselves. Not only Salesforce (Agentforce), but also HubSpot (Breeze AI), where Brinker serves as VP Platform Ecosystem).

Before we look forward, it is interesting to look back.

How the martech stack became what it is today

Almost six years ago (March 2019) I spoke with Brinker about the transition from a single vendor marketing cloud versus a “Frankenstack” to a platform model. Let me break it down.

There was a period in the 1920s when vendors like Oracle and Salesforce essentially told companies that could afford it to leave everything to us. We will do our best to provide all the services you need. Indeed, the late Mark Hurd, then CEO of Oracle, predicted in 2015 that by 2027, two vendors would control 80% of the martech market.

On the other hand, “Frankenstack” was about subscribing to individual point solutions and merging them (sorry, integration) within the house. It was an often necessity-driven approach for companies that couldn’t afford to become an Oracle or Salesforce shop.

The transition taking place at HubSpot (and indeed Salesforce) that Brinker and I discussed was the emergence of the platform model. A single solution would be at the center of the stack, but connect via APIs to single point solutions that did things the main platform couldn’t (or did better). Hence the development of Salesforce AppExchange and HubSpot App Marketplace.

In essence, that’s where we are today, although orchestrating and integrating large collections of applications and making sense of the data flows between them remains a challenge. But let’s not ignore iPaaS.

iPaaS as a solution

The integration platform-as-a-service sought to ease these challenges by providing cloud-based tools designed to automate the integration and orchestration of martech applications. Take a look at Gartner’s definition:

iPaas is a suite of cloud services that enables the development, execution and management of integration flows that connect any combination of on-premises and cloud-based processes, services, applications and data within single or multiple organizations.

Salesforce acquired one of the leading iPaaS vendors, Mulesoft, in 2018.

In practice, however, iPaaS solutions promised more than they could always deliver. Despite the promise of automation and even the ability for non-technical teams to use them. However, the unprecedented complexity of integration and the frequent need to adapt out-of-the-box APIs meant developer involvement; and developers, it is saidit often prefers to work without a chosen iPaaS.

Dig deeper: AI Transformation: 2025 Predictions

iPaaS, meet AI

But don’t we now have an obvious candidate to run iPaaS, or even handle stack orchestration without iPaaS? Meet your friendly AI agent.

That’s Brinker’s thesis anyway. AI agents are already beginning to orchestrate applications across the board. As mentioned earlier, big players like Salesforce and HubSpot now have AI agents built into their platforms. As Brinker notes, mid-sized vendors are following a similar path (Braze and Zeta, for example).

Sounds simple enough. But there’s also a wild west of agent developer solutions that give companies the ability to create and customize their own agents for a wide range of purposes. Google and OpenAI support agent development, of course, but so do dozens of other much lesser-known vendors (including low-code and no-code solutions).

Echoes of Frankenstack? Agentstein?

Dig deeper: Customer Experience Management in the Age of Agent Artificial Intelligence

Find simplicity out of chaos — but not yet

The potential for AI agents to finally solve the challenges of stack integration and orchestration is undeniably exciting. As well as their potential for designing and running marketing campaigns and taking over 90% (or more) customer support and service activities. (Automating campaign personalization “is becoming known as ‘AI Decisioning,'” says Brinker; this will come as a surprise to Pega, who has been calling it that for years.)

I applaud Brinker’s enthusiasm. But he and I have been around the neighborhood often enough to know that many developments that promised to fix everything—from marketing automation to CDPs—proved to be only partial solutions and often actually added complexity to the pile.

Agenting the Salesforce of this world could be a huge development for Salesforce customers. But for many companies, implementing AI agents, whether for integration or other purposes, will be a long and complex task.

Read Brinker’s predictions here.