Proptech
LISTING LAUNCH AGENT
As a self-initiated concept for Crexi PRO, I designed an AI agent that keeps commercial listings moving after they go live. As the solo product designer, I researched broker pain points, defined the agent's rules and permissions, and built a working coded prototype with a guided tour. The agent checks facts, ranks buyers, catches stalls and drafts seller updates, and the broker approves every step that leaves the building.
My Role
My Role
Product Designer (Solo, Self-initiated Concept)
Industry
Industry
AI in Commercial Real Estate
Contributions
Contributions
UX Research / AI Agent Design / UI Product Design / Interaction Design / Coded Prototyping
Launch with Confidence: An AI Agent That Keeps Commercial Listings Moving
Launch with Confidence: An AI Agent That Keeps Commercial Listings Moving
Overview
Overview
CHALLENGE
A broker publishes a new listing and the clock starts. Who should see it? Which inquiry is a serious buyer, and which is just browsing? Why did views drop by half in week three? Crexi's AI already helps brokers create listings, offering memorandums and document summaries (PR Newswire). Once a listing is live, though, the broker is on their own. Reviewers describe the leads section as cumbersome (G2), and listings go stale when status isn't kept up to date (Pierce Commercial Real Estate). The result is slower replies, lost buyers and frustrated sellers
A broker publishes a new listing and the clock starts. Who should see it? Which inquiry is a serious buyer, and which is just browsing? Why did views drop by half in week three? Crexi's AI already helps brokers create listings, offering memorandums and document summaries (PR Newswire). Once a listing is live, though, the broker is on their own. Reviewers describe the leads section as cumbersome (G2), and listings go stale when status isn't kept up to date (Pierce Commercial Real Estate). The result is slower replies, lost buyers and frustrated sellers
SOLUTION
I designed an AI listing agent that picks up where Crexi's current AI tools stop and runs a listing from launch to close. Here's how a broker works with it:
Every action has its own permission setting (Auto, Ask first or Off), and every step can be undone from the activity log. The design goal: nothing reaches a buyer, a seller or the public unless the broker approved it or chose to put that task on Auto.
I designed an AI listing agent that picks up where Crexi's current AI tools stop and runs a listing from launch to close. Here's how a broker works with it:
Every action has its own permission setting (Auto, Ask first or Off), and every step can be undone from the activity log. The design goal: nothing reaches a buyer, a seller or the public unless the broker approved it or chose to put that task on Auto.
RESEARCH
RESEARCH
Before designing a single screen, I studied where brokers lose time after a listing goes live. I combined Crexi's product announcements and public user reviews with my own experience selling land through MITS Land. Four patterns stood out:
Lead follow-up is clunky: Users describe Crexi’s lead tools as awkward to manage, and the volume of leads coming in is uneven. (G2 user reviews)
Listings fall out of date: When brokers don’t keep status current, buyers waste time on properties that are no longer available. (Pierce Commercial Real Estate)
AI help ends at publishing: Crexi’s AI can draft offering memorandums, create listings and pull data from documents, but nothing manages the listing once it’s live. (PR Newswire, Crexi AI announcement)
Bad data sinks land sales: Wrong acreage and unconfirmed utilities often come to light only after a deal is weeks under contract. (Field experience, MITS Land)
My own land deals added a fourth pattern: wrong facts are expensive. An acreage mismatch or an unproven utility can derail a contract weeks later. These findings shaped the agent's core rule: check facts before launch, and flag conflicts instead of hiding them.
Testing the Agent's Judgment
With the problem defined, I built a coded prototype instead of static mockups, so the agent's behavior could be tested, not just viewed. I then wrote a 45-minute interview guide for commercial and land brokers. Every question maps to a screen in the prototype:
Before designing a single screen, I studied where brokers lose time after a listing goes live. I combined Crexi's product announcements and public user reviews with my own experience selling land through MITS Land. Four patterns stood out:
Lead follow-up is clunky: Users describe Crexi’s lead tools as awkward to manage, and the volume of leads coming in is uneven. (G2 user reviews)
Listings fall out of date: When brokers don’t keep status current, buyers waste time on properties that are no longer available. (Pierce Commercial Real Estate)
AI help ends at publishing: Crexi’s AI can draft offering memorandums, create listings and pull data from documents, but nothing manages the listing once it’s live. (PR Newswire, Crexi AI announcement)
Bad data sinks land sales: Wrong acreage and unconfirmed utilities often come to light only after a deal is weeks under contract. (Field experience, MITS Land)
My own land deals added a fourth pattern: wrong facts are expensive. An acreage mismatch or an unproven utility can derail a contract weeks later. These findings shaped the agent's core rule: check facts before launch, and flag conflicts instead of hiding them.
Testing the Agent's Judgment
With the problem defined, I built a coded prototype instead of static mockups, so the agent's behavior could be tested, not just viewed. I then wrote a 45-minute interview guide for commercial and land brokers. Every question maps to a screen in the prototype:
Design
Design
IMPACT
IMPACT
For Brokers
Less time chasing leads and more time closing. The agent drafts replies, ranks buyers and flags stalls early, and the broker keeps the final say on every message.
For Crexi
