Restoring Partner Trust Through Compliance and Automation
Category
Compliance & Automation
/
/
At Rail, a compliance stack that was not calibrated tightly enough was letting some higher-risk transactions reach banking partners, eroding their trust at a critical moment. As trust dropped, partners scrutinized more transactions and raised more RFIs, creating friction and leaving transactions stuck for longer. The timing made it especially sensitive, this was unfolding during the acquisition process, where keeping partners confident mattered a great deal. I worked with the compliance team to address the cause, then rebuilt the RFI process on top.
The challenge
The volume of RFIs was high and growing, but the volume was a symptom. The real problem was upstream.
The compliance stack was not calibrated tightly enough, so some higher-risk transactions were being processed when they should have been caught.
Sending partners transactions that should not have moved eroded their trust, so they scrutinized a greater share of transactions and raised more RFIs, which created friction, held some transactions up, and pushed processing times higher.
On the response side, a full team manually reviewed every case, cross-checking the partner, the Rail ledger, and client documentation. Because the process was entirely manual, resolution was slow, with some RFIs taking five or more business days to close.
The approach
We tackled the cause first, then the symptom, and started implementing from day one rather than spending months analyzing.
Phase 1: fixing the root cause
Working alongside Rail's CCO and compliance team, we recalibrated the compliance stack, guided by partner feedback, past cases, and FINTRAC MSB regulations, so it captured what was needed, kept us fully compliant, and let through only transactions that could legitimately be processed. I contributed to the analysis and the changes as part of that team effort.
Phase 2, stabilizing the RFI process
Redesigned the ticketing workflow in Freshdesk with custom tags, statuses, and automation rules to track every RFI, and added real-time dashboards for open RFIs and turnaround time.
Phase 3, the AI agent
Built a Gemini Gem to cross-check the Rail ledger against the documentation provided by the client, validating sender, beneficiary, amount, and consistency, while also assessing transaction context, purpose, and country risk and flagging prohibited categories and nested transactions, drawing on previously cleared cases to resolve similar new alerts and RFIs, and refined through a weekly improvement loop.
Phase 4, up-front RFI
In collaboration with the engineering team, shipped a feature in the client-facing platform that let clients attach supporting documentation when submitting a transaction. I wrote the PRD based on client feedback and data, and the engineers implemented it. If a transaction was held by the compliance stack or queried by a partner, we could clear it straight from those documents, with no back-and-forth and far less delay.
The impact
Tightening the compliance stack cut off the source of the problem, and rebuilding the RFI process made what remained fast and reliable. Within 3 months the operation was markedly faster and more trusted. By the end, the company had tripled its RFI capacity while reducing team size, and its compliance function had become a genuine strength in the period leading up to its acquisition by Ripple.
With fewer transactions being queried and RFIs resolved quickly, transactions moved through far more smoothly.
Partner trust recovered, and escalations dropped sharply.
By the end of the project, around six months in, the operation had turned around on every front.
Monthly RFI volume fell by roughly four times, as the root cause was fixed and fewer transactions were queried.
The work moved from a full compliance team dedicated to answering partner RFIs to a single person handling it end to end.
Average handling time dropped by about five times.
Manual effort was reserved only for transactions flagged by the compliance stack or that received an RFI, and even those were reviewed through the AI agent with a human in the loop.
Partner confidence recovered and client satisfaction rose, as transactions previously held up by RFIs were processed around five times faster.
Fix the cause, not just the symptom
A spike in RFIs is rarely the real problem. Here it was a signal that the compliance stack needed tightening. By addressing that first and then rebuilding the RFI process with automation and AI, the operation stopped the issue at its source, restored partner trust, and turned a compliance bottleneck into a genuine strength.
