From manual processes to scalable innovations
As a coastal Vancouver Island community with a population of just over 9,000, the Town of Qualicum Beach may be small in size but its innovations are not. Within their existing tech stack, the team leverages no-code tools like FME to deliver innovative solutions comparable to larger jurisdictions.
Leading IT and GIS operations at the Town is Reno Sun, whose team oversees everything from capital and operational projects, cybersecurity posture, data and systems integration, and day-to-day IT & GIS operations.
Reno’s journey with FME began at BCIT and evolved through his early roles as a GIS Tech at the Town of Qualicum Beach (2013–2019) and as a GIS Analyst (2019–2020) at the Capital Regional District. Upon returning to the Town in 2020 as the Manager of IT and GIS, he established FME as the backbone of the organization’s strategy. The platform now supports repeatable data transformation and automations, provides a single source of truth, and allows the Town to reimagine previously inconceivable projects using AI.
In leveraging FME alongside Microsoft and Esri technologies, Reno earned the MISA BC Spirit of Innovation Award for two consecutive years in 2024 and 2025. These awards recognize local government members whose presented solutions deliver the most beneficial, innovative, and valuable solutions to challenges faced by local government. Reno’s efforts were also recognized with an Award of Excellence at the 2025 GeoConnect Conference (formerly URISA BC) in May 2025.
FME’s no-code workspaces and geospatial strengths have empowered the team to do more with their existing resources. This foundation set the stage for a series of initiatives that transformed how the Town delivers services, improves transparency, and applies scalable uses of AI.
Boosting transparency with a self-serve development tracker
Previously, development applications were managed through email-based processes and required manual data entry and notifications. With no centralized system, fragmented data, and paper-based processes, there was risk for lost or missed information.
To improve transparency and reduce the administrative burden on planning staff, the Town of Qualicum Beach set out to modernize how development applications were tracked and communicated.
The Town’s objective was to provide staff and the public with a centralized way to manage and track applications. Inspired by land management solutions from the City of Maple Ridge, a public-facing and internal application to visualize status, location, and related documents for current and historical applications was pursued.
Creating a real-time, single source of truth
Residents first submit information through a Survey123 web form, where information is populated to a managed SharePoint list. FME Flow then syncs the data in real time from SharePoint into the Town’s GIS interface. Changes are automatically detected and reflected, allowing team members to focus on attribute data and access a single source of truth.
“FME is such a great tool to detect changes, updates, or to identify if something has been removed. It allows users to build on their existing tech stack, which results in huge training cost savings and is a successful factor of change management.” –Reno Sun
The result was an intuitive Development Tracker that improved transparency for both staff and residents. Planning staff reported a noticeable drop in calls about application statuses, as residents could now easily find relevant information.
Eliminating communication silos and streamlining public inquiries
Resident inquiries used to be submitted through a patchwork of emails, phone calls, and letters, none of which were triaged in a centralized way. This lack of visibility led to manual work, communication silos, and posed a risk for missed or forgotten requests.
Using AI to triage and route public inquiries
To change this, the Town implemented an AI-driven approach to centralize, categorize, and track incoming requests. Rather than forcing all communication through a web form, the Town pursued a more user-friendly email-based process. Using GovAI, incoming emails are read and automatically categorized based on a SharePoint lookup table and then assigned to the appropriate team member.
FME and Microsoft Power Platform tie everything together behind the scenes. FME integrates to the internal Work Request System and automatically creates work orders when required based on the inquiry type. Bi-directional updates via a FME Flow webhook keep records current, eliminating the need for manual updates.
What was once a siloed system is now automated and transparent. Externally, residents receive a confirmation with an expected resolution date. Internally, teams can easily monitor timelines and workload, while management gains full visibility into inquiry volumes and resolution performance.
This automation is projected to save an estimated 50 workdays annually, delivering nearly 94% in cost savings.
Testing semantic search for stakeholder buy-in
The Town’s SharePoint and public websites house thousands of documents, but keyword-based searches struggled to deliver relevant results. In some cases, it failed to produce the correct document altogether.
Within the internal SharePoint document library, a test search for “mobile device policy” ranked the correct file 13th. The intended results for a search of “sidewalk policy” placed 18th. Keyword searches didn’t understand the intent behind queries, forcing staff to manually sift through large volumes of information. In examples of natural language searches* (e.g., When do I need a dog license?), no relevant documents were returned.
*Traditional keyword searches using natural language can fail to produce the correct document. In this example, a question such as “When do I need a dog license?” did not return any relevant “Animal Control Bylaw” document.
Building a functional prototype with FME Flow Apps
In June 2025, Microsoft released the Copilot Retrieval API Preview, introducing semantic search for document libraries and facilitating access to relevant content. To support team adoption and gain stakeholder buy-in, Reno first had to test if the solution could query over 6,000 documents effectively across their SharePoint library.
This opportunity was explored by building a functional prototype with FME Flow Apps. Built in just 5 days, the app used FME’s HTTPCaller to retrieve names, relevant scores, and content of each document in the SharePoint library. With modern ‘vibe coding’ and GenAI, it’s possible to script equivalents, but the maintainability risk is high. The team estimates that development would have taken 2-3 times longer without FME and would have resulted in more fragile, harder-to-maintain solutions.
“FME Flow helped us build this app so easily. We didn’t need to write any code… We just built the app for stakeholders to try. FME Flow lets a small team deliver faster, standardize intuitive patterns, and reduce long-term maintenance burden versus hundreds of custom code lines.” –Reno Sun
A strong case for semantic search
When the same test queries were repeated, the difference was immediate: “mobile device policy” and “sidewalk policy” appeared as top results, helping users find the exact document needed and reducing search noise. The previously unsuccessful search using natural language, When do I need a dog license?, now found the correct “Animal Control Bylaw” as the third search result.
After, with Semantic Search: Now, a natural language search of “When do I need a dog license?” finds the correct “Animal Control Bylaw” as the third search result.
The rapid prototype allowed the team to quickly prove user value, as the semantic model retrieved results based on relevancy, not on matching keywords. Quickly built without code, the functional prototype demonstrated the benefits of pursuing AI-powered semantic search over keyword-based searches, providing a strong proof of concept for stakeholder buy-in. As shared by a member of the team, “I have done a number of searches through this test and have been quite happy with the results. It is definitely an improvement on the current website search. I would say I have been able to find whatever I searched within the first 5 results in most cases.”
Digitizing 1000s of legacy service cards with FME and AI
Service cards record essential information about utility connections and are vital for BC 1 Call’s “Call Before You Dig” service. With 4,400 legacy service cards at the Town of Qualicum Beach—each estimated to take 6 minutes to digitize manually—digitizing them all would be an incredible task.
The team’s approach to their service cards was re-examined while watching a Safe Software webinar hosted in collaboration with partner Avineon Tensing. The webinar highlighted how UK Power Networks used FME with Google Gemini to extract handwritten dates from over a million historical electrical service records. Nearly 1.1 million cards were processed in just 26 hours, averaging approximately 700 cards per minute. Otherwise estimated to have taken 19 years if done manually, the project inspired Reno to build an in-house approach for the Town of Qualicum Beach’s legacy service cards.
With the goal of creating dynamic service records, the Town engaged Avineon Tensing for hands-on guidance. Avineon Tensing built a prototype FME workspace with Google Gemini and provided 3 hours of handover training, equipping the team with a strong foundation for designing AI prompts and defining output schemas.
From inspiration to in-house application
Taking the guidance from Avineon Tensing, the Town pivoted to its existing AI platform, GovAI, to meet operational preferences and compliance requirements. Collaborating closely with GovAI, they co-developed a GovAI API to extract structured data from legacy service cards containing both handwritten and digital text. The FME’s HTTPCaller Transformer invoked the GenAPI with a clear prompt and schema to extract mixed handwritten and digital text.
Initially, the extraction was attempted within a single prompt, which overloaded the model and produced hallucinations. The process was then restructured into 2 stages to improve quality:
- Stage 1 (Recall-oriented): Extract raw text and key fields from each service record to JSON.
- Stage 2 (Precision-oriented): Re-prompt using the Stage 1 output and apply 20+ QA/QC rules to refine, validate, and standardize results.
Turning unattainable tasks into scalable solutions
This 2-stage GenAI extraction significantly improved accuracy and consistency as processing time dropped to just 15 seconds per card. With an intuitive ArcGIS Experience Builder used to verify and correct AI-extracted data, each service card connection record can be corrected in about 1 minute total by a team member.
Representing an 83% reduction in manual effort, the FME and AI workflow took a task once considered unfeasible into a scalable solution that can be reused for other legacy documents across departments.
Looking ahead, the Town plans to continue expanding its FME ecosystem, automating more multi-click workflows, deepening integration with Microsoft and Esri platforms, and delivering faster, more proactive public services. By automating repetitive work, improving transparency, and enabling practical AI adoption, the Town is giving staff more time to focus on supporting residents and delivering better outcomes with their existing tech stack.
