Smart Maintenance Conversational AI
/If you’re a property manager or a landlord responsible for a portfolio of properties, you know the importance of providing high-quality, prompt maintenance service to your tenants. Maintenance issues - if not dealt with in a timely manner - can affect tenant satisfaction, tenant retention, and ultimately your company’s revenue.
Here we will show how AppFolio’s Smart Maintenance simplifies the maintenance process, helping tenants resolve their maintenance requests conveniently and efficiently. To demonstrate how tenants report maintenance issues using Smart Maintenance, let’s take a look at a hypothetical scenario involving Sandy, a tenant who discovers a clogged drain when doing the dishes after dinner.
Named entity Recognition (NER) used By Smart Maintenance to identify and categorize key information gathered from the conversation
Step 1: Confirming the Tenant's Identity and Address
The first step of Smart Maintenance’s process is to confirm the tenant's identity and address. This is a critical first step because the tenant who is reporting the issue might not be the person who signed the lease, and the property manager needs to know who to contact and where to send the maintenance technician.
To initiate a conversation with Smart Maintenance, a tenant can send a message to our AI-assisted texting interface, or text their property manager who starts the process. In this scenario, Sandy texts Smart Maintenance about her clogged drain directly and does not provide her full name. Because Sandy did not provide her name and address in the initial message, Smart Maintenance will conduct additional information-gathering and respond with follow-up questions to extract that information.
Sandy then replies with her full name and address, and Smart Maintenance uses an identity parsing model to extract and record her contact information and cross-checks it against the value in our database. Throughout this process, there is a human operator overseeing the conversation who can interject at any time - but our AI is automating the operator’s job to a large extent.
Step 2: Identify the Issue
Once Sandy’s contact information is confirmed, Smart Maintenance’s next task is to identify the issue that she is reporting. There are several types of maintenance issues, and each one requires a different course of action to resolve, and with varying urgency. For example, a leaking faucet might need a simple fix, while a broken heater during winter might need an emergency replacement.
Instead of simply providing Sandy with a long list of issues to choose from, or leaving the property manager to guess the issue from Sandy's description, Smart Maintenance uses a machine learning-based model to predict the issue based on her description. If Sandy described the clogged drain succinctly by stating "the kitchen drain is clogged" or in a more roundabout way e.g. "the water is not draining" – Smart Maintenance knows both of her descriptions are referring to the same issue.
We also provide an intuitive user interface for the human operator to review and confirm the issue for an additional layer of confidence. In the rare scenario where the operator finds that the model did not predict the clogged drain accurately, they are provided with the top five predictions that the model believes the issue could be and adjust the prediction accordingly.
Step 3: Gather the Details
After confirming Sandy’s maintenance issue is a clogged drain, Smart Maintenance needs to gather additional details about the drain. These details are necessary to understand the scope, severity, and urgency of the maintenance issue, and further, are used to provide the maintenance technician with the relevant context.
For each maintenance issue submitted to the system, the tenant will be asked a series of troubleshooting questions. These questions are designed to assist the tenant in resolving their issue without sending a maintenance technician, so that simple issues can be resolved promptly.
If a maintenance technician is needed, Smart maintenance will ask triage questions. Triage questions are meant to determine the urgency of the issue. For Sandy’s clogged drain, the system might ask whether only one drain is clogged, or what the severity is of the drain’s blockage. Additionally, Smart Maintenance inquires as to whether there are any fire, flood, or other hazards associated with the issue.
Smart maintenance uses another machine learning-based model to turn the tenant’s responses into standardized and structured data that can be easily stored and processed. Smart Maintenance also validates the answers and asks for clarification if the answers are unclear or incomplete.
Steps 4 and 5: Summarize and Close the Issue
After completing all the aforementioned steps, Smart Maintenance uses a summarization model to present the information to a human operator who creates a work order for Sandy’s clogged drain. The summary provided by the issue details model is included in the work order’s description.
If Sandy is happy with the work order created our closing model will end the conversation based on Sandy’s response. Sandy can also text in later if there is new information, or if she simply wants an update. Our closing model will then re-open the conversation and have operators address Sandy’s questions.
Step 6: Submit the Work Order to the Property Manager and Vendor
With the work order creation finalized by the tenant, Smart Maintenance submits the work order to the property manager. The property manager can now make an informed decision on which vendor is best suited to address Sandy’s problem, in this instance a plumber, and dispatches them to fix the clogged drain.
Once Sandy’s drain is fixed by the plumber, Smart Maintenance continues to optimize the process - handling billing, gathering customer feedback, and even allowing property managers to define rules that will automatically contact vendors based on an issue, the issue’s urgency, and the answers provided by the tenant to the triage questions.
Overview: How Smart Maintenance Streamlines Operations
Smart Maintenance provides a seamless, user-friendly experience for tenants to report maintenance issues and help manage their maintenance requests more efficiently.
Streamlined Issue Reporting: Tenants can simply text or chat to report their maintenance issues, instead of filling out complicated forms or calling an office. They describe their issues in their own words instead of choosing from predefined options while being guided to provide relevant information. This reduces frustration for tenants and encourages them to report issues promptly.
Less Time Investment for Property Management Staff: Smart Maintenance automates repetitive conversations, reduces errors related to data entry, and creates an accurate system of record
More Accurate and Complete Issue Identification: Smart Maintenance understands the tenant's description of the issue and is able to discern the issue's severity, frequency, and location of the problem. This helps to avoid miscommunications and ambiguity
Quicker and Easier Work Order Creation: Smart Maintenance summarizes the collected information to generate a concise and clear issue description that can be used to create a work order. Property managers can rely on the work orders being accurately submitted and can set rules on how they’ll be notified and how the work is assigned.
More Efficient Communication: Smart Maintenance facilitates communication with tenants and vendors, provides updates, and handles notification and assignment of issues. Smart Maintenance also handles common tenant queries, such as ETA requests, feedback, or follow-ups. This can help you improve your transparency and accountability, and keep tenants and vendors satisfied.
Authors and Contributors: Christfried Focke, Ken Graham, Joo Heon Yoon, Shyr-Shea Chang, Tony Froccaro