More details and context can be provided in an interview.
CivicTech
AI-Powered Citizen Request Chatbot
AI chatbot to help citizens report problems to cities.
A new AI-powered chatbot product that can categorize reports by citizens of problems in their city, provides the citizen a tracking link for their service request, and automatically notifies the correct city employee about the problem.
TEAM UI/UX Designer (Leo), 1 PM, 2 Engineers, 2 Data Scientists
GOAL Create an easy way for citizens to report service requests to their city through the local government website.
TIMELINE 7 Months
OUTCOME 90% Cost Reduction
EMPLOYER CivicTech
TOOLS Figma
Citizens need an easy way to report problems to their local government
Current methods for citizens to report a problem to local governments or utility providers either involve long and cumbersome forms or talking on the phone to an untrained call center agent. Existing forms require complex and conflicting categorization be done by the citizen whereas reporting over the phone makes it difficult to give an accurate location on a map or provide a photo of the issue.
Competitor products are long forms that require citizens to make complex decisions
The current industry standard method for citizens to report service requests on the web are long, complex web forms with many fields and a lot of complex decision making is required on behalf of the citizen to decide which amongst similar sounding categories their problem fits into.
Reporting via Call Centers was expensive and inefficient, resulted in data errors, and couldn’t collect photos
At the time, citizens reported service requests through CivicTech to local governments and utility providers via call centers. This method was found to be inefficient, costly, and prone to errors, especially regarding accurate location data. Photos couldn’t be attached to a report over the phone.
Custom AI could collect and categorize data and identify missing information
Using proprietary machine learning models, CivicTech is able to take a written description of a problem and accurately categorize it with high precision. This reduces the time and effort needed for citizens to report a problem.
Our goal was to leverage CivicTech’s core AI capabilities to streamline the service request reporting process through a user-friendly interface, allowing citizens to report requests accurately and efficiently on their own devices. The goal was to eliminate human error, enhance scalability, and provide a seamless experience that guides users through the reporting process while collecting all necessary information for problem resolution.
High-level goals were:
- Make it fast and easy for citizens to create a service request on the web
- Collect photos and more accurate location data
- Build a solution that can handle increasing volumes of requests efficiently over time
User Flows
Prototypes of a traditional form showed data collection needed to be non-linear
Initially we tried a traditional form that helped citizens pick the correct category through a search field but this still put too much effort on the citizen and we quickly realized that the collection of information about a request didn’t follow a strict sequential order of linear steps. Different problem categories required different pieces of information to fully understand the issue.
Traditional Form Concepts
A chat interface provided a familiar interaction that could guide users through questions one-at-a-time and ask the next question based on all previously provided information.
Introduction
Description Entry
Followup Questions
Photo Upload
Verification Method Selection
Editing
83% task success rate in usability testing
Usability testing was performed on a beta version of the chatbot with 3 users which achieved an overall success rate of 83%. Additionally, three areas of improvement were identified for future revisions.
Submission process
Outcomes
90% cost reduction
The cost to CivicTech for a citizen to report a problem via the chatbot was 90% less, or more, than the cost of using a call center to collect data. The chatbot also enabled reporting locations without an address (such as a fallen tree in the middle of a public park) and enabled citizens to upload photos of their requests which, based on user interviews, were highly valued by city staff, crew managers, and responders.
Market disruption
The first version of the product was still in active late-stages of development when I left the project. When it is released, the chatbot will dramatically disrupt the existing market and bring forth a modern, innovative way for citizens to quickly and easily report a problem to their city.
Takeaways
The primary learnings from this project for me were the various established UI patterns of chat interfaces. I took various components and patterns from other chat products and implemented them in a way that best fit the needs of this project. Sometimes new patterns needed to be created due to the unique requirements of this experience; in particular, editing previous answers without restarting the conversation and helping users understand that their edited answer resulted in a branching event in the conversation flow to explain why some questions may be asked again.
Transfer of technology to call centers
The chatbot won’t entirely replace the need for reporting via a call center. Some citizens will prefer calling in their problem due to aversion to more modern technology or accessibility needs. The proprietary decision making technology CivicTech developed for the chatbot could be used to build better tools to aid untrained call center agents in collecting data.
Further Improvements
Future versions of the chat widget might include viewing all past reported problems, information on active outages, using image recognition on photos to inform the categorization of the problem, paying service bills, and possibly querying knowledge bases for answers to common problems. There is a huge amount of potential for innovation and modernization of communication between citizens and local governments.
Figma Prototype
Demonstrates animations and transition from initial greeting screen to conversation.
Verification Code
Name
Phone Entry
Drafts
Location Input
Manual Category Picker
Write
Publish
COMPANY NAME HERE
WRITE TITLE HERE
Widget Example
OUTCOME Custom white-label AI ChatBot widget envisioned including 9 input types, onboarding, editing, contextual followup questions, and drafts.
The chatbot was meant to be used by adult citizens on smartphones, tablets, and desktop devices; but primarily on mobile devices. User personas included homeowners, parents, and senior residents.
Local governments need better way to collect & categorize service requests from citizens.
Service requests examples: water shutoff request, pothole report, new stoplight request.
Competitors use traditional forms that require too much effort by citizens
The current industry standard method for citizens to report service requests on the web are long, complex web forms with many fields and a lot of complex decision making is required on behalf of the citizen to decide which amongst similar sounding categories their problem fits into.
CivicTech uses AI to solve the categorization problem
Using proprietary machine learning models, CivicTech is able to take a written description of a problem and accurately categorize it with high precision. This reduces the time and effort needed for citizens to report a problem.
Call centers are expensive and inefficient
Citizens reported service requests through CivicTech to local governments and utility providers via call centers. This method was found to be inefficient, costly, and prone to errors, especially regarding accurate location data. Photos couldn’t be attached to a report over the phone. Also, untrained call agents often didn’t record critical details, leading to ineffective issue resolution.
A chatbot provided a familiar, simple framework for the user to focus on answering questions one at a time and ask for whichever piece of data was needed next.
I made use of the company’s AI technology to create a new industry-first chatbot widget through which citizens can easily report in natural language a description of their service request.
The chatbot categorizes the reported problem; asks contextual followup questions to fill in any needed missing information; then collects contact details and, if relevant, location and photos of the problem.
More Projects
ProjectsFixed PDF Ticket Scanning Issue for University Events, 15% Accuracy ↑Fixed PDF Ticket Scanning Issue for University Events, 15% Accuracy ↑
