If you run a brokerage, or run operations for one, you’ve probably experimented with AI by now. We have, too. We built custom GPTs for brand guidelines, compliance checks and internal policy questions, and they worked.
What I didn’t see at the time was the gap between our front-of-house AI and everything behind it. While we were playing with forward-facing tools, our management and back office systems were teetering on the edge of Word documents, multiple manuals and disorganized file systems that weren’t shared or consistent across the team. If you manage a brokerage, some version of that sentence probably describes your operation, too.
So over the past three months, we completely restructured the way our brokerage is organized, from the ground up, and the shift from AI that answers questions to AI that does the job has changed how our entire company runs. This is what that rebuild looked like from the management side and what I’d tell another broker or operations lead before starting.
The problem with the chatbot stage
Most brokerage leadership teams using AI right now are stuck at the same stage we were: You build a chatbot, it answers questions, and everyone feels ahead of the curve. That’s a fine start.
But a chatbot is only as organized as the documents behind it, and if your processes live in one manager’s head, three versions of a Word doc and a shared drive nobody can navigate, AI can’t fix that. It just gives you faster access to the mess.
We learned this the hard way when we lost a long-time staff member. Institutional knowledge walked out the door with her, and we struggled to maintain quality and voice consistency in our marketing. We couldn’t bridge the gap and didn’t want to publish content that wasn’t written to our standard. If you’ve ever lost a key admin or manager, you know exactly what that gap feels like.
So, we stopped adding AI tools and started rebuilding the system they sit on.
What our office AI system looks like
Every process at our brokerage now lives in Notion, and we’ve built Claude projects as the working layer for each department.
The payoff for our management team has been immediate: Instead of getting bogged down in training and task management, we point new hires to Claude, and they can ask how to stock the office supplies, what needs to be ordered at the beginning of each month or the steps for marketing a listing and get the answer exactly as we would give it. Onboarding stops depending on a manager’s calendar.
For our marketing team, the projects do something even more valuable: They hold the line on our brand. Every content piece gets checked against our writing standards and brand guidelines before it goes anywhere, which means our voice stays consistent no matter who is doing the work.
AI workflows: the blog pipeline
The best example of the shift from chatbot to workflow is our blog process.
Here’s how a post gets made now. Our admin team takes a topic and works up an outline. The draft runs through a fact-checker skill that verifies accuracy before anything else happens. Then a blog optimizer skill takes over, checking formatting and applying SEO and GEO best practices so the post is built to perform in both traditional and AI-powered search.
The before and after: A blog post used to take us five to six hours start to finish. Now it’s done in under an hour, and it comes out on-brand, fact-checked and optimized every single time. The blog that went quiet is back, and the quality bar we were afraid of lowering is now enforced by the workflow itself.
The part everyone dreads (that turned out to be easy)
When brokerage leaders hear “document all your processes,” they picture months of tedious writing. I expected that, too. It was the easiest part of the whole rebuild.
Our approach is simple: Record yourself doing the task. That’s it. We take the transcript into Claude, where a transcript-to-SOP skill turns the messy, rambling task description into detailed steps with checklists. Then an SOP-to-Notion skill publishes it straight into our Notion system with all of our predetermined required fields, formatted and ready to use.
When we started, I assumed getting an SOP into Notion would take hours per process. Now it’s as easy as telling Claude to do it, and it takes minutes.
The honest caveat: This is a rebuild, not a quick fix
I want to be clear about the commitment here, because this is not a “sign up and see results” situation, and it’s not something you can hand off to one tech-savvy agent.
Restructuring our operations forced us to zoom out and rethink everything, from how our files are shared (we moved our server to Dropbox) to how we handle task management and team communication. This is a leadership-level decision that touches every department.
If your file systems are inconsistent and your processes are scattered, AI workflows will expose that immediately, and you have to be willing to fix the foundation before you build on it. It takes buy-in from the whole team, but we’re already reaping the benefits, and every process we add to the system compounds the return.
Where to start Monday morning
Don’t buy another tool, and don’t delegate AI workflows before you understand them. Start by learning the basics of the AI you already have access to. For us, that’s Claude, and the two concepts that matter most are the difference between projects and skills and the importance of building the right documentation to train your projects.
AI is only as good as the information you give it. A brilliant model sitting on top of disorganized documents will give you disorganized answers. But once you know what to give it, how to build it and how to use it, it can change your business.
This October, Inman turns its focus to AI and its rapid rise in real estate. During Artificial Intelligence Month, we’re digging into the startups shaking things up and the established players folding AI into their offerings.
Kate Hulbert is the marketing director at Bozeman Real Estate Group in Bozeman, Montana. Follow her on Instagram or Facebook.
