Export Development Canada

Export Development Canada finances and insures Canadian companies selling abroad. Its relationship managers hold exporters whose exposure sits across several products at once (guarantees, receivables insurance, working capital, political risk), and whose circumstances change with a shipping lane or an election.

The organisation was moving that relationship management onto Salesforce Financial Services Cloud. The platform was settled; what the people inside it would do all day was not. Five processes had to survive the move intact, each carrying obligations, CASL consent among them, that could not be designed around.

And a Salesforce rollout has a well-documented failure mode. Adoption and data migration sink more of these programmes than technology does: every team configures its own corner, people quietly keep the spreadsheet, and eighteen months later nothing resembles anything else.

Relationship managers, underwriters, sector leads, compliance, and two exporters on the other side of the glass. Transcribed, tagged, clustered, and then re-clustered until the themes stopped moving.

Three of the four findings were the ones you would predict. The fourth was not: consent was being managed by memory. People had been trained on the rules and were trusted to follow them, with nothing in the system to make the safe path the easy one. That reframed the brief from a migration into a redesign.

Impact scored from what the research actually said; effort from build estimates with the Salesforce team. Plotted together, the argument mostly makes itself, and the conversation moves from what people want to what the first release has to prove.

The cut line was a phase boundary rather than a wish list. Everything to the right of it needed platform changes the programme had not yet earned the right to ask for.

The failure people described was silence. A referral left one desk and nobody could say where it had gone, so the exporter was chased by whoever remembered.

Every transition now writes owner, state and consent to the record, and the consent check sits before any outbound contact rather than in a policy document. The compliance obligation became a step in the flow, which is the only place it was ever going to hold.

Same content, same data, three arrangements, in front of eight people across the five roles. Deciding structure while it is still cheap is most of the job; everything after it is refinement.

The list-first version was the fastest to build and the closest to the legacy tool, which is exactly why people kept leaving it to find context somewhere else. Record-first won because it matched how every single role described arriving: with an exporter's name already in mind.

I built and governed the SLDS V2 design system and the Figma library under it: tokens, variants, guidelines, audits. Governance is the part that gets skipped and the part that decides whether a design system is still true a year later.

The screen below is composed live from that system's own tokens rather than screenshotted, which is the point of building it this way: every surface in the programme reaches for the same values, and this page cannot drift from them.

Working with Salesforce admins and developers on component specs and handoffs, with the backlog run in Azure DevOps, the MVP shipped phased rather than promised whole, and WCAG 2.2 was held as a design constraint from the first grey frame.

The whole programme was assembled from the SLDS V2 design system, and WCAG 2.2 was held as a design constraint from the first grey frame.