During the month of May 2024, I had the opportunity to deliver a speech at the Public Technology Institute (PTI) AI and Cyber Summit in Washington, D.C., where I was able to attend numerous impressive presentations from local authorities across the nation.
The standout practice that caught my attention at the PTI event — and the most remarkable local government technology endeavor I have witnessed in the previous year — originates from the “Monty” initiative located in Montgomery County, Md.
Shayna Taqi led the presentation team from Montgomery County. Taqi holds the role of the chief change officer for Montgomery County within the Department of Technology and Enterprise Business Solutions (TEBS). With over two decades of experience in guiding strategic change management initiatives, she is in charge of deploying strategies to enhance understanding and adoption of large-scale innovations and technology introductions. Taqi presently spearheads the initial generative AI (GenAI) assignment in the county. This innovation has extended Montgomery County’s assistance to residents and visitors, enabling them to effortlessly access information regarding county services from a virtual chat agent on the county’s 311 webpage.
Having received her B.A. from American University and obtaining a Certified Public Manager designation from the Metropolitan Washington Council of Governments, Regional Executive Development Program, Taqi’s credentials are well-established.
Following the PTI event, I reached out to Shayna regarding an interview for this blog. The aim was to introduce the Monty project to a wider audience.
Dan Lohrmann (DL): Enlighten us on how the Monty project commenced. What did Monty 1.0 encompass when it first launched?
Shayna Taqi (ST): Our initial website chatbot, Monty 1.0, made its debut in January 2021 to aid the county’s 311 team in handling the surge in resident service demands during the COVID-19 pandemic. The Monty 1.0 chatbot utilized Zammo.ai artificial intelligence technology and multi-turn conversation mapping in both English and Spanish to address resident queries in relation to around two dozen of the most popular 311 topics.
DL: With the pandemic receding, what evolution occurred in the project’s vision?
ST: During the COVID-19 pandemic, Montgomery County residents faced extended wait times to connect with a MC311 customer service representative. Recognizing the urgent requirement for an alternate communication medium due to time-sensitive requests such as rental and financial aid, the Monty 1.0 chatbot was swiftly configured and launched on the MC311 site in just a few weeks. Developed by a compact project team within TEBS, in collaboration with Microsoft and Zammo.ai, Monty 1.0 focused on the primary services residents frequently sought. Alongside, modifications were made to the telephony system to expedite the handling of urgent calls. Within two months post-launch, hold times diminished from 5 minutes to 2 minutes, while call abandonment decreased from 25% to 11%.
Despite effectively managing the burgeoning service requests directed towards the county’s 311, feedback from residents concerning the Monty 1.0 chatbot was occasionally negative as it failed to address every query that a customer service representative could handle. The process to expand Monty’s knowledge base was manual and necessitated advanced technical expertise, constraining 311’s ability to promptly enhance and sustain an informed chatbot. Resultantly, residents expressed that the chatbot’s scope was too limited and struggled with basic inquiries and requests. In response to this feedback, the county’s Department of Technology and Enterprise Business Solutions launched a proof of concept in May 2023 to explore the utilization of generative artificial intelligence technologies as a remedial measure.
DL: What features does Monty 2.0 presently incorporate? How many languages can the service cater to?
ST: The Monty 2.0 chatbot leverages generative artificial intelligence technologies to facilitate natural conversations with residents, eliminating the need for multi-turn mapping. Capable of instantly addressing queries across over 3,000 311 subject areas — a hundredfold surge from its prior version — the chatbot supports conversations in over 140 languages, encompassing the seven most commonly spoken languages in Montgomery County (English, Spanish, French, Mandarin Chinese, Korean, Vietnamese, and Amharic).
DL: Which GenAI and other technologies contribute to building Monty?
ST: The Monty 2.0 chatbot relies on a synapse pipeline (a cluster of data-driven workflow tasks) to transmit 311 knowledge base data to Microsoft Azure, the county’s chosen platform. Here, the 311 data undergoes transformation and indexing for integration with Microsoft’s GenAI Cognitive Search service. The chatbot’s application, workflows, and user interfaces are managed using tools provided by Zammo.ai.
In summary, a chat request is initially received and dispatched to the Microsoft Cognitive Search service via Zammo.ai workflow. The service processes and selects relevant 311 knowledge base data, aligning with the received query. This data is then forwarded to OpenAI through Zammo.ai workflow, where it is synthesized into a coherent response using ChatGPT GenAI technology. The response is subsequently relayed back to Zammo.ai and delivered to the resident, fulfilling the request promptly. The end-to-end request/response operation transpires within seconds.
DL: How is Monty accessible? What are the key advantages for Montgomery County residents utilizing this service?
ST: The Monty 2.0 chatbot can be accessed via the county’s 311 website. This chatbot empowers Montgomery County residents to swiftly access essential information on numerous county services across various languages without sifting through extensive content or waiting to reach a customer service representative. Users can simply type or voice their queries within the chat interface, with Monty 2.0 seamlessly handling the rest. The chatbot excels in responding to direct and concise requests, enabling the county’s 311 customer service team to allocate more time to resolving intricate resident issues.
DL: How has user response/feedback been thus far?
ST: Since its deployment in March 2024, the reception towards the Monty 2.0 chatbot has been generally positive, as indicated by the analytics collected by the county using Zammo.ai feedback mechanisms. Resident feedback is automatically solicited via thumbs up/down options and comment forms as part of each chatbot interaction. The usage of Monty 2.0 has more than doubled compared to Monty 1.0, when comparing corresponding dates year after year. Overall, residents have welcomed the enhanced scope and efficiency of the chatbot in comparison to its predecessor, with feedback aiding in uncovering previously unnoticed issues and other feature requests. TEBS is continually working…Absorb insights from resident input and evolve advancements to the Monty 2.0 chatbot by refining prompts, workflows, knowledge base information, and other backend elements to diminish the occurrence of inaccurate and unresolved resident inquiries.
DL: Can you furnish additional particulars regarding resident feedback?
ST: Feedback from residents has chiefly revolved around three focal points: enhancements to our knowledge base, betterment of the chatbot user experience, and requests for fresh content and functionalities. Resident input has been instrumental in aiding our team to identify flaws and glitches present within the knowledge base, like broken links, outdated content, and other related issues. We are also utilizing resident feedback to steer our prompt engineering backend endeavors and enhance the overall efficiency of the chatbot’s vast language model infrastructure. Furthermore, we are maintaining a backlog of resident enhancement requests to shape forthcoming iterations of the Monty chatbot.
The most frequent enhancement requests we receive pertain to incorporating functionality directly within the chat window itself — for instance, submitting county service request forms via the chatbot, transitioning conversations from the chatbot to a 311 customer service agent, and obtaining location-specific service details.
Exploring integration with other service tiers, such as 911, state and federal aid, might be considered down the line as Montgomery County consistently seeks ways to enhance service provision for our residents. Additionally, we are in the initial stages of revamping the county’s web pages.
DL: I was impressed by your team’s presentation at the recent PTI AI and Cyber Summit in Washington, D.C., especially how you involved your team members and allowed them to showcase a segment of the project. They exhibited immense passion for their roles and were evidently proud of their contributions. How did you cultivate such a positive environment? Do you have a combination of public- and private-sector personnel?
The Monty 2.0 Project Team comprises the following experts, with the image courtesy of Montgomery County:
Gail Roper, Chief Information Officer, Program Sponsor
Shayna Taqi, Project Lead
Manu Daniel, Technical Project Manager
Michael Zanfardino, Functional Project Manager
Tushar Parekh, Developer
Sri Harsha Kotagiri, Developer
Colin Cox, Product Owner
Skyler Grubbs, Change Management Lead
Lauren Der, Communication and Marketing Lead
ST: The Monty 2.0 chatbot was internally developed by TEBS, with close collaboration from our Microsoft and Zammo.ai vendors. Our project team established a unique bond over a span of 10 months, from conceiving the initial proof of concept to rolling out the chatbot. The Monty 2.0 chatbot project was one of several concurrent initiatives for team members, who were occasionally borrowed from different divisions within TEBS to lend expertise in areas like project management, business analysis, and software development. Consequently, we adopted an agile, sprint-based approach to optimize our team’s effectiveness.
Our sprints were collaborative and involved extensive cooperation among multiple team members to design, develop, and validate chatbot components. This validation process encompassed more than a dozen rounds of functional testing and the facilitation of a resident focus group to shape our final product. Throughout this journey, team members frequently juggled various responsibilities and interchanged roles according to project needs and phases. We embraced a culture of shared learning from failures and remained unafraid to experiment in pursuit of innovations. This amalgamation of diverse perspectives, skills, and receptiveness to constructive input fueled the project’s success — a camaraderie that continues to thrive within our team.
DL: What is the next chapter for Monty? What are your aspirations for the upcoming year or two?
ST: The project team is presently focusing on several significant upgrades to the Monty 2.0 chatbot, such as incorporating county GIS capabilities to better aid residents with personalized info requests, like polling station locations, snow clearance status, and related details. We are collaborating closely with our Microsoft and Zammo.ai vendors to refine the chatbot’s overall performance through backend component enhancements. These enhancements involve deploying Microsoft Azure Prompt Flow, a development tool used to evaluate and enhance the precision of the prompts guiding the Monty 2.0 chatbot’s GenAI components. Our budding product management team is also evaluating UI/UX enhancements and other knowledge base improvements to comprehensively address flaws identified by residents and boost chatbot adoption rates. Lastly, we aspire to apply our team’s accumulated expertise to tackle additional GenAI projects within the county domain.
From the project’s inception to this juncture, we have recognized the value of continuous enhancement and leveraging the operational efficiencies we gained during the COVID-19 pandemic. We take immense pride in the team’s agility in swiftly supporting our residents while modernizing our 311 technology and enhancing Montgomery County’s customer service delivery framework and capabilities in a more holistic manner.
About Author
