What OpenAI’s Agent Builder Means for Marketers

OpenAI’s Agent Builder: A Quick Primer

At a high level, Agent Builder provides a visual canvas where users can create workflows with drag-and-drop nodes. You can compose logic, set guardrails, access a Connector Registry to see what connectors are active in your workflow, and even use RAG (Retrieval-Augmented Generation) capabilities for leveraging files and knowledge bases. 

All of that is a great step forward, but it isn’t perfect yet. Agent Builder only supports a limited number of native connectors and relies heavily on MCP (Model Context Protocol), which is a standard for linking AI systems with other apps. 

As it stands, HubSpot maintains an official MCP server, Salesforce is piloting one, and several third-party options are available for Marketo. But overall, the MCP ecosystem is quite young. While MCP in itself is a good solution, the fact that Agent Builder relies exclusively on such an underdeveloped protocol will pose some challenges.

With that aside, what makes the Agent Builder particularly interesting – beyond the welcomed visual approach to building workflows – is its adaptability. Agents built with OpenAI’s platform can reason dynamically. Meaning, they can decide which step to take based on context. 

While this opens up a world of possibilities when it comes to autonomous decision-making, dynamic reasoning isn’t always preferred, which is where the current iPaaS solutions come in.

Current iPaaS Landscape

For anyone unfamiliar with the concept, iPaaS stands for Integration Platform as a Service. These platforms act as “universal adapters” for software, connecting different apps together like Salesforce, Marketo, Slack, etc., so they can talk to each other and share data. 

Today’s iPaaS solutions have matured significantly. Here’s a brief overview of some of the major ones and what they’re typically used for:

• n8n: Open-source platform supporting AI agents that make autonomous decisions, multi-agent systems, and integration with various LLMs beyond OpenAI

• Zapier: 8,000+ integrations with AI Agents using natural language, though following predetermined workflows rather than adaptive autonomy

• Workato: Enterprise-grade with advanced error handling, retry logic, and monitoring capabilities for mission-critical operations
  
  
• Power Automate: Deep Microsoft 365 integration with AI Builder, supporting both RPA and digital process automation
  
  

When to Us What: Agent Builder vs. iPaaS

The fundamental difference between Agent Builder and other iPaaS solutions lies in their approach to automation. 

Agent Builder is a shift towards AI-native automation, where AI is determining the best path forward through fluid, conversational workflows. They’re ideal for tasks like intelligent customer support, content generation with context, exploratory data analysis, and other tasks where flexible reasoning is built in. 

Some AI-focused iPaaS solutions can handle conversational workflows too, but unlike Agent Builder, they excel at providing granular control over predetermined processes. They are great when you need practical workhorses for structured, repeatable tasks such as syncing records, triggering campaigns, or managing approvals. 

This is especially important when you want to automate processes that require a specific, enforced order of predetermined tests. 

For example: 

When importing a set of leads to Marketo, there’s a very specific order of API calls we must do. First, we have to create the job, then begin the job, then query the job status until it’s completed, then get the finalized import. The API calls must be in this order, and an iPaaS solution like n8n or Zapier allows us to enforce this order while OpenAI’s Agent Builder does not. 

The Path Forward

Over time, we think it’s likely that these categories will merge. OpenAI will expand its ecosystem, and iPaaS vendors will deepen their AI features. Marketers will find that the future of automation isn’t about choosing one over the other; it’s about combining autonomous intelligence with good infrastructure to get the best of both.

The next generation of marketing automation won’t just run scripts. It will understand goals, adapt to given context, and collaborate with teams. And that’s exactly what inspired Otto, our own AI teammate for Marketing Ops.

We designed Otto to integrate with the apps you already use, carefully assembling according to each client’s existing tech stack and governance model. It feels like working with another teammate in Slack, and it is ultimately designed to help more marketers do their best work.

If you want to learn more about Otto, go here.

We’re optimistic about AI Agents and where all this is headed. We can’t wait to see what major breakthroughs come next!