How We Built our First Multifamily AI Agent

At Lineups, we’re always looking for ways to make life easier for property managers and marketers in the multifamily space. So in early 2023, when we set out to build our first GPT-powered application, we knew automation would be a game changer. But it wasn’t just about bringing “AI” into the toolkit—it was about building an AI agent that could handle real work.

In this post, I’ll walk through how we made our Multifamily AI Design Agent, using real workflows from inside our own product. While our tools are specific to Lineups, the same core concepts can help you connect the dots in any AI-driven workflow.

The Starting Point: Micro-AI Solutions

Our journey began with a bunch of micro-AI solutions. Think of these as tiny, specialized tools: a caption writer, a file namer, a text classification tool, an image tagger, an image generator, and so on. Each solved a single, tightly scoped problem for our users when they were designing.

But as our feature list grew, we hit a wall. Our users didn’t just need a bunch of one-off tools—they needed a way to combine these little AI's to complete actual, workflow-level tasks. This is where an AI agent comes into play: something that could orchestrate a series of micro-tasks to reach one big goal.

What Did We Start With?

We picked a high-impact, but tedious, part of the user experience: data entry, especially when customizing marketing designs.

Our design workflow is built around a flexible, text-based editor. It’s easy to use, but even easier if you barely have to touch your keyboard. Imagine if you could just describe what you want in plain English, and the AI fills out the entire form for you—headlines, subheadings, dates, times, and more.

Here’s How Our AI Agent Works:

1) User submits one “master prompt”.
“We want to post the local events for this weekend.”)

2) The AI writes and formats the content.
It generates the copy, breaks it into logical sections, and identifies which web form fields map to which pieces.

3) It fills out the form automatically.
No more copying and pasting into six different inputs. The AI does it instantly.

But what if you could go even further. Ask questions you have never been able to ask before.

MCP Server: Connecting to External Data

What’s really powerful is when your agent isn’t limited to what you type in. What if it can search the web, fetch property profiles, or even pull in up-to-the-minute inventory—all from your prompt?

For this, we built a custom MCP Server (Model Context Protocol). Think of it as a translator that lets our AI agent talk to other software (i.e. Entrata, RentCafe, etc).

An Example Prompt:

“I want to promote a floor plan on social media. Pick a floor plan that has the most units available right now.”

Here’s what happens behind the scenes:

• The AI rewrites the prompt and asks the MCP Server for inventory data.
• The MCP Server queries the Property Management System (PMS), pulls available floor plans, and returns whichever has the highest availability.
• The AI formats this info as structured data and passes to the design assistant.

All in one step, triggered by a single natural-language prompt.

Why It Matters (and What Comes Next)

By letting your AI agent connect to external data sources and tools, you multiply its power. Tedious, repetitive tasks disappear. Content gets more accurate and relevant. And your users save real time.

This is just the beginning. As we build deeper integrations and smarter agents, more and more of the heavy lifting will just… disappear. Imagine AI that’s not just helpful, but knows your business as well as you do.

We’re excited about what’s possible. If you have prompt ideas or challenges you want solved, drop them in the comments—I’d love to hear what you’re dreaming up.

Want to see a demo? Check out the video/GIF above for a behind-the-scenes look at our AI agent in action!

‍

‍