--- title: "Which AI Agent Should a CRE Firm Build First? | Clustr" description: "How a commercial real estate firm can pick its first AI agent, using the three agents Clustr built with Cresa in ChatGPT Enterprise as the worked example." url: https://getclustr.ai/blog/ai-agents-commercial-real-estate-first-build source: https://getclustr.ai/blog/ai-agents-commercial-real-estate-first-build --- # Which AI Agent Should a Commercial Real Estate Firm Build First? ## Which AI agent should a commercial real estate firm build first? Build first for a workflow your team repeats on every deal, that runs on documents, and that a person can check quickly. For Cresa, that meant three agents in ChatGPT Enterprise: RFP drafting, landlord proposal analysis, and prospect briefings. The RFP agent is now used in production. Pick the one where a checked draft saves the most repeated effort for the people who own the work. Most commercial real estate firms do not lack ideas for AI. They lack a way to choose. Brokers want help with prospecting, analysts want help with proposals, and marketing wants help with pitch materials. Trying to build all of it at once usually produces a demo that nobody uses. Starting with one well-chosen agent produces something people rely on, and a pattern for the next one. ## What makes a CRE workflow a good candidate for an AI agent? A good candidate has four traits: it happens on every deal or every week, it starts from documents or research rather than relationships, the output has a clear expected shape, and a person can check the result in minutes. If a workflow fails one of these tests, it is usually a better fit for training than for an agent. - **Repeated.** The work comes around on every deal or every week, so a small saving each time adds up and people get plenty of practice with the agent. - **Document-heavy.** The input is a proposal, an RFP, a set of notes, or public research. AI is strongest when it is reading and organizing text. - **Clear output.** The team can describe what good looks like: a house-format response, a table of lease terms, a one-page briefing. - **Easy to check.** Someone who knows the work can review the draft quickly and catch what is wrong. Work that depends on judgment, negotiation, or a long relationship with a client is a poor first candidate. An agent can prepare a broker for that work, but it should not try to do it. ## What three agents did Clustr build for Cresa? For [Cresa](/case-studies/cresa), the Chicago-based commercial real estate firm with 1,300+ employees and 55+ North American offices, we built three agents in ChatGPT Enterprise. One drafts RFP responses in Cresa's house format. One extracts lease economics from landlord proposals. One researches prospects and briefs brokers before meetings. Each was built inside tools Cresa already licenses. 1. **RFP drafting agent.** Tenant-rep teams answer RFPs constantly, and every response has to match the firm's format and voice. The agent drafts responses in Cresa's house format from the team's own material. It is now used in production. 2. **Landlord proposal analysis agent.** Comparing proposals means pulling rent, escalations, concessions, and term out of documents that are each formatted differently. The agent extracts lease economics in minutes. By hand, that work took analysts hours per deal. 3. **Prospect research and briefing agent.** Brokers walk into meetings better prepared when someone has done the research. We built this agent with the brokers who use it, so the briefing covers what they actually want to know. Context matters here. Cresa had already rolled out ChatGPT Enterprise across one of its highest performing offices before the agents were built. They sit on top of a tool people already had, inside a governed, IT-approved enterprise agent workspace. Nobody had to buy new software, and deal data stayed in tools the firm already governs. ## Which of the three should most CRE firms build first? For many tenant-rep and brokerage teams, an RFP drafting agent is a sensible first build. RFPs arrive often, the house format gives the agent a clear target, and the team can judge a draft quickly. At Cresa, it is the agent now used in production. Your best first agent may differ, so score your own workflows before choosing. That is not a rule. A team with a heavy analysis load might get more from proposal analysis first. A team focused on new business might start with prospect briefings. The point is to choose on evidence about your own workflows, not on which agent sounds most impressive in a leadership meeting. ## How do you score your own workflows before choosing? Sit with the people who do the work and list the workflows they repeat. For each one, ask how often it happens, how long it takes by hand, how clear a good output is, and how quickly someone can check it. The workflow with strong answers on all four, and a team willing to test it, goes first. 1. **List the repeated work.** Ask brokers, analysts, and coordinators what they do on every deal. Write down the inputs, the steps, and what a good output looks like. 2. **Check frequency and effort.** A workflow that takes a long time but happens twice a year is a weaker candidate than one that takes less time but happens every week. 3. **Check how easy it is to verify.** A draft RFP response is quick to review. A financial summary needs a careful figure-by-figure check. Both can work, but plan for the review. 4. **Check the data rules.** Confirm the workflow can run inside your licensed enterprise workspace, with connectors IT approves. If it needs data that cannot go there, it is not a first build. 5. **Find an owner.** An agent with no one on the team who wants it will not survive its first month. ## What should you avoid in a first CRE agent? Avoid three common mistakes: building an autonomous agent that acts without a person checking it, building outside the AI platform your firm already licenses, and treating the build as the finish line. A first agent should produce a draft, live in a governed workspace, and come with training on live deals so people learn to trust and correct it. In commercial real estate, a wrong rent figure or a misstated concession can cost a deal. That is why we design every agent to hand a draft back to a person. Building outside your licensed platform creates a new data path that IT has to approve, which slows everything down. And an agent with no training behind it tends to get tried once and forgotten. ## How long does a first CRE agent take to build? A configured agent for one workflow typically takes weeks rather than months, because it is built in a tool the firm already licenses. Most of the time goes into mapping the workflow with the team and training people to use it on real deals. The configuration itself is usually the smaller part of the work. ## Where should a CRE firm start? Start with a baseline of who will use the agent, then pick one workflow using the four tests above. At Cresa, we baselined the growth and finance teams with SINA, our AI literacy platform, before building anything. The engagement target is a 40% uplift in GenAI understanding against that baseline, which is a goal, not a reported result. From there, build the agent with the people who will run it, teach it in live sessions on their own deals, and hand it to a small group of champions who improve it and carry it to the next team or office. Our page on [AI agents for commercial real estate](/ai-agents-for-commercial-real-estate) sets out the five-step approach we use, and our [commercial real estate](/commercial-real-estate) page covers the wider training program. To see how champions keep agents alive after launch, read about our [AI champions program](/ai-champions-program). Ready to enable your teams with AI? Book a call to discuss your AI adoption challenges. We use cookies We use cookies and Google Analytics to understand how visitors use our site. [Privacy Policy](/privacy)