← All discussions

Z
Other Discussion

HSBC's New AI‑Driven Credit Scoring Rollout

Zoey Patel · 19 Aug 2026

Since the AI credit scoring platform went live last month, our underwriting team’s dashboards have been spitting out error flags for 30 % of mid‑market applications, even when the data looks clean. We've found the model’s explainability reports are buried in a separate portal, so analysts spend an extra hour per file trying to justify decisions to senior managers. Management says the tool will cut processing time by 40 %, but we’re seeing longer turnaround and rising complaint numbers. Has anyone managed to integrate a third‑party explainability layer without breaking the vendor SLA, or found a pragmatic way to keep the team productive while the model is being fine‑tuned?
👍 497 💬 2

2 replies

Evelyn Phillips · 19 Aug 2026

Running the budget we see the extra hour per file is chewing up about five percent of our quarterly spend. When we rolled out a similar AI scoring tool at a previous bank we built a simple PowerBI‑Excel link that auto‑loads the model’s explainability PDF and shaved roughly fifteen minutes off each review, so a lightweight integration could turn the current slowdown into a net gain.
R

Rita Brown · 19 Aug 2026

Last month we tried pulling the model's output directly into our PowerBI reports via the Azure ML API, hoping to kill the extra hour, but the API throttles after 200 calls and the dashboard just stalls. We even set up a simple Python script using SHAP to surface the explainability metrics inline, but the script crashes on any loan over £500k and we lose the whole batch. So far the only thing that worked was falling back to the old spreadsheet, which is slower but at least the team can actually see the numbers.

Log in or create a free account to join the conversation.