A prototype the frontline helped build.

The challenge

For organizations that serve people directly, time is the resource that never stretches far enough, and the systems meant to support the work quietly eat into it. At Multi-lingual Orientation Service Association for Immigrant Communities (MOSAIC), its clients are often newcomers to Canada working through employment, housing, language, and immigration paperwork. The settlement professionals who help them do that work across nine disconnected systems. After each client session, the details are manually re-entered into their case management system, a tool built more for reporting than for the work itself, often requiring hours of work.

MOSAIC serves newcomers across a wide range of services. Our work focused on two of its programs, the Newcomer Community Building Program (NCBP) and the MOSAIC Moving Ahead Program (MAP). Leadership knew the administrative layer on these programs was heavy but lacked a clear picture of where the time was actually going. MOSAIC needed an evidence base before investing in new systems, and it had to land with frontline staff, not just management, so that whatever changes followed could be implemented.

What changed

Over ten weeks, MOSAIC's leadership went from a general sense that admin was heavy to a clear, evidence-backed picture of it, along with a validated prototype to build on.

• An interactive blueprint. We mapped the end-to-end settlement worker journey across both programs, quantified the workload by task, and surfaced the biggest pain points for leadership to act on.
• A prototype that turns conversation into the record. In the prototype, workers have the client conversation and AI structures the notes into the required form fields, and a worker checks and confirms the AI's output before it becomes part of the record.
• Frontline workers pushed it further than we scoped. We piloted with the same frontline workers whose journey we had mapped, across four languages reflective of MOSAIC's client base, using real conversations rather than demos. When they saw it working, they began pointing to other steps AI could take on next, which is the kind of momentum a working prototype is meant to create.
• Not every bottleneck we found called for AI, and we built accordingly. The process mapping also exposed scheduling across MOSAIC's four offices as a source of missed appointments. We designed an MS Bookings process in a blueprint for implementation considerations. Seeing the process clearly is what lets us match each problem to the right fix, sometimes a prototype, sometimes a tool.

What we delivered

Why it worked

• We started with the work, not the system. Interviews showed the gap between how leadership understood the process and how workers actually experienced it. From there we designed the map and prototype based on how client conversations really happen.
• The people behind the work validated it. Joint validation sessions meant findings landed as accurate with frontline staff first.
• AI was designed into the workflow, with the worker in control. The AI runs inside the primary workflow and drafts the record, and every session note passes through worker review before it becomes official. The worker decides, which is what keeps the output trustworthy and the data clean.