Deploying Enterprise AI to 1,200 Users in a Relationship Business

Deploying enterprise AI to 1,200 users across a 30-office real estate brokerage taught me that the technology is the easy part. The hard part is organizational: setting expectations, building trust, and creating conditions for people to experiment without fear.

In early 2025, we made the decision to deploy Google Gemini Workspace across our entire organization: 1,200 users spanning 30 offices in four states. At the same time, we developed a custom AI chatbot built on our institutional knowledge base with guardrails designed for our specific business context. Here is what that process actually looked like.

Why Did William Pitt Sotheby's Choose Google Gemini?

The conversation about AI in our organization did not start with AI. It started with a question we had been asking for years: how do we make it easier for 1,100+ advisors and support staff to find what they need, when they need it? Our intranet, Agent Connect, already centralized most of our tools and knowledge. AI was the next layer: making that information accessible through conversation rather than navigation.

We chose Gemini Workspace because it integrated directly into the Google ecosystem our teams already used daily. No new login. No new interface. AI showed up inside the tools people were already in. That mattered more than any feature comparison.

But the decision was also driven by something more urgent: governance and data security. Deploying an enterprise AI platform was so important to us because the alternative was not “no AI.” The alternative was unmanaged AI. Organizations that do not have a strategy or have not deployed a company-wide enterprise toolset will drive their users — whether real estate agents or any other professional service — out to the open market with consumer products. When that happens, client data and proprietary information can end up in models that use it for training purposes, or worse. An enterprise deployment was not just a productivity play. It was a security imperative.

How Did We Build a Custom AI Chatbot for Real Estate?

The Gemini deployment addressed general productivity: drafting emails, summarizing documents, analyzing data. But we also needed something purpose-built for our business. Our custom AI chatbot was trained on our institutional knowledge base: policies, procedures, market data, onboarding materials, compliance guidelines, and operational playbooks.

The guardrails were non-negotiable. In a regulated industry, an AI that confidently provides wrong answers about commission structures, compliance requirements, or legal procedures is worse than no AI at all. We built citation requirements into the system, so every response traces back to a source document. We restricted the knowledge domain to prevent the chatbot from improvising on topics outside its training data.

What Drove Adoption Across 1,200 Users?

Adoption was faster than expected, largely because we met people where they already were. The tools did not require behavior change. They augmented existing workflows. An advisor could ask the chatbot a policy question that would previously require a call to the office or a search through multiple documents. A manager could draft a market update in minutes instead of an hour.

The compounding effect was real. Once people saw AI handle a task successfully, they started looking for other applications. Usage grew organically. Agents ran more than 50,000 prompts through the Gemini application in the first three months alone. The best marketing for an AI tool is someone on your team saying, “Have you tried asking it?”

There has been a genuine openness across the organization to embracing these tools. That surprised me. I expected more resistance, more skepticism, more “this is a fad” dismissiveness. Instead, what I saw was curiosity — people experimenting, sharing prompts with each other, finding applications we had not anticipated. When adoption is organic like that, you know the tool is solving a real problem.

What Is an AI Hub and Why Does a Real Estate Firm Need One?

One of the lessons from the initial rollout was that giving people access to powerful tools is not the same as giving them the ability to use those tools effectively. Google Workspace, NotebookLM, Gemini — the ecosystem of AI products available to our agents is growing rapidly, and it can be overwhelming.

That is why we are building an AI Hub: a centralized resource within Agent Connect that helps advisors understand all the AI products available to them, what each one does best, and how to get started. It incorporates a prompt library so that people do not need to reinvent the wheel every time they sit down to use a tool. If an advisor in Greenwich figures out a great way to use Gemini for comparative market analysis, that prompt should be available to an advisor in Litchfield County the same day.

The AI Hub is designed to lower the barrier to effective use. It is not enough to deploy the technology. You have to build the scaffolding that helps people develop fluency with it over time.

What Would You Do Differently in an Enterprise AI Rollout?

I would invest more in structured onboarding for AI tools specifically. We treated the rollout like a software deployment: training sessions, documentation, and go-live support. But AI tools require a different kind of onboarding. People need to develop intuition for what to ask and how to ask it. That is a skill, and it takes practice. The AI Hub is partly a response to this realization.

I would also build measurement into the system earlier. We can track usage — and the 50,000 prompts in three months was a strong signal — but measuring productivity impact and time savings at scale requires instrumentation that is harder to add retroactively. If you are planning a deployment like this, define your success metrics before you turn the tools on, not after.

What Is the Biggest Lesson from Deploying Enterprise AI?

Deploying AI to 1,200 users taught me that the technology is the easy part. The hard part is organizational: setting expectations, building trust, maintaining guardrails, and creating the conditions for people to experiment without fear of getting it wrong.

Real estate is a relationship business. AI does not change that. But it changes the operational substrate underneath it. The firms that invest in that substrate now, thoughtfully and with discipline, will have a structural advantage that compounds for years. The ones that wait are not standing still. They are falling behind, one unmanaged consumer AI prompt at a time.