ChatGPT and Codex reshape a bank: Singular Bank's Singularity creates 90 minutes a day!

Hey everyone, it's Shii-chan! Today I found a really concrete story about how much time AI can create on the front lines of a bank. The numbers are crisp and it's a great read, so let me walk you through it!

OpenAI News openai.com

What was announced?

Singular Bank, a private bank based in Madrid, built an internal assistant called "Singularity" using ChatGPT and Codex. This is a customer story published on OpenAI's News.

Singularity helps bankers analyze portfolios in real time, prepare for meetings, and draft compliant follow-up communications. Across the team, each banker saves 60-90 minutes per day.

The story so far

Before Singularity, analyzing a client portfolio was a real chore. Bankers had to pull positions, reconcile the data by hand, and assemble a coherent view before every meeting - and they repeated it for every single client. That preparation ate up a lot of time.

With Singularity, bankers can immediately see what matters in a portfolio. It automatically flags things like concentration risk, elevated exposure, or portfolio imbalance. It even recommends actions - reducing concentration, locking in gains, or rebalancing toward a more stable portfolio.

What changes

The biggest shift is how meeting time gets spent. Instead of reviewing data, bankers can now focus on advising clients and talking about next steps.

Javier Marín, CEO of Singular Bank, puts it this way:

"Instead of reviewing data, the banker is leading the conversation."

Now a banker can say something like "you're up 38% overall, but you're heavily concentrated in a few positions - the next step is to protect what you've gained." The conversation shifts from operational busywork to what to do next. I love that.

After each meeting, Singularity also helps draft personalized follow-up messages and identify next actions. Because it's built into the bank's core systems, every output is captured and structured, which improves traceability and makes reporting more reliable.

Dive Deep

What I find fun is the concrete before-and-after numbers for each task:

• Meeting preparation: ~20 minutes → under 1 minute
• Call reports: 15-20 minutes → under 30 seconds
• Investment arguments: 10-15 minutes → ~20 seconds
• Client communications: 5-10 minutes → under 30 seconds

And it's genuinely used in daily practice. Over a 30-day period, bankers ran more than 3,500 operations, averaging around 120 per day across 19 different workflows.

The key point is that outputs are generated from approved data sources and structured workflows. That means more consistent communications and improved traceability for regulatory reporting - so bankers can move faster without compromising regulatory standards.

Wrap-up

• Singular Bank built an internal assistant, "Singularity," with ChatGPT and Codex
• Each banker saves 60-90 minutes per day
• Task by task, the work shrinks dramatically - meeting prep goes from ~20 minutes to under 1 minute
• With 3,500+ operations across 19 workflows in 30 days, it's truly used in production
• Approved data plus structured workflows aim to balance speed with regulatory compliance

This one will resonate with anyone who wants AI to fit in as a tool that strengthens human judgment rather than replacing it - especially teams in fields like finance, where accuracy and traceability are everything!