Business Flow Design for AI Leasing Assistant

Background

In order to build an AI leasing assistant, it’s important to understand and abstract the best practice of property leasing workflows to accommodate various needs and scenarios, such as apartment, coliving and student housing, while envisioning the future product. As the first AI PM and with no standard SOP (standard operating procedure) in place company-wide, I needed to deep dive and start from scratch.

Pain Point

- Regional and Agent Variability: Leasing and sales processes vary significantly across regions and among leasing agents. For instance, some managers require specific chat formats and formalities, while others prioritize results and adopt a more flexible approach.
- Property Type Differences: The leasing process also differs based on property types. For example, the distinction between apartments and co-living spaces needs to be clearly outlined. Co-living, with shared spaces, is common on the East and West coasts but less familiar to customers in Chicago.
- Audience Variability: The leasing process varies depending on the audience. Foreign students, for instance, may be unfamiliar with credit scores and the need for a co-signer, whereas working professionals are typically well-versed in these aspects.

Goal & Achievement

- Customer Reply Rate: Improved reply rate by 10%
- Rental Preference Completion: Increased customer completion rate by 14%
- Tour Scheduling: Boosted scheduling rate by 16%
- Agent Efficiency: Helped leasing agents save 2 hours per day, improving response efficiency by 40%

Product Planning

- Portfolio Data Analysis to Identify Directions: Based on the data above, the Koreatown management team, with its focus on apartments, and the USC team, with its diverse co-living options, comprised the majority of the available portfolio. I decided to initially focus on apartment hunters due to the large total addressable market, as well as the mature and straightforward sales flow.

- User Interviews & Brainstorming: Engaged with team members from the leasing team across various roles and portfolios. Shadowed their daily workflows to gain a deep understanding of their pain points and operational processes, including preference collection, recommendation, tour, application.

- Competitive Analyses of EliseAI, BetterBot, and Respage: Through demo calls and system usage, I uncovered key insights about their SaaS services, including their reliance on integration with property management systems, built-in online chat functionality, and retrieving property information from existing systems. Our advantages lay in the granularity and richness of data, as well as in the expertise of domain experts with hands-on practice.

- Chat Flow Design: Mapped out and compared various distinct processes, identified common triggers, developed internal ML models, and crafted a unified action framework for the AI leasing assistant. This framework was tailored to be applicable across different property types, including apartments, co-living spaces, and student housing.

Solution & Release

- Major Leasing Flow Revamp
1. Standard main flow for all types of properties and different regions, which would include preference collection, recommendation, cross-selling, tour and application stages.
2. Personalized journey for various customers without following a rigid, step-by-step flow to accelerate conversion.
3. Added property type preferences, such as "apartment only" or "open to co-living." This better helps understand customer preferences and creates opportunities for cross-selling.

- Minimum Customization: Generally, it followed the flow outlined previously, but in specific cases such as the recommendation and tour stages, the AI leasing agent operated with a customized setup. For example, it would explain that leases for student housing ended on a specific date, while leases for regular apartments offered more flexibility. It also confirmed parking needs for apartment renters, especially given the limited availability in Los Angeles.

- Acceptance Testing & Go-To-Market: After development, I invited internal team members to test the system and gathered feedback. I refined the product based on their suggestions and then released it to customers searching for apartments. The reaction was positive, as timely, accurate, and detailed responses not only quickly resolved customer inquiries but also helped manage high traffic volumes during peak seasons and off-hours.

Learning

1. ‍Width VS Depth?
   When considering multiple property types and the entire flow from preference collection to application, should I focus on one step for all properties or the full flow for a single property type? After careful consideration, I opted for the latter for the following reasons:
   1. A comprehensive understanding of both the business and technical landscape, particularly in the context of AI leasing, provides a strategic overview that is crucial for making informed decisions and planning the next phase of development.
   2. Apartments, being one of the most common property types, present an ideal starting point. Successfully implementing a solution for this type would not only simplify technical reuse but also provide a solid foundation for rapid scaling to other property types.
   3. Focusing on width could introduce scalability issues, such as increasing traffic volume, and difficulties in collecting feedback across multiple regional teams and the risk of launching features before key components are fully ready, which could hinder overall system performance.
2. ‍Customization VS Standardization?
   Due to various factors, from team preferences to operational procedures, I received countless requests after each release, sometimes with differing opinions about the same feature. Through practice, I’ve developed a few guiding principles.‍
   1. Standardization through Best Practices: Standardization is guided by industry best practices, informed by domain knowledge and the judgment of a product manager. This ensures consistency and alignment with proven approaches.‍
   2. Prioritizing Key Elements: I identified issues and categorized them as deal-breakers, key steps, and nice-to-have improvements. Prioritizing those aligned with best practices ensures that critical needs are met while deprioritizing less impactful features. Additionally, well-documented action items and a clear, transparent roadmap with timelines help maintain focus and clarity across teams.‍
   3. Gradual Customization: For customization, I adopted a gradual approach. Taking small, incremental steps allows for testing and verifying customer needs, such as backend adjustments before investing in more complex, user-facing configurable settings.