Singular Bank helps bankers move fast with ChatGPT and Codex
OpenAI
A Spanish bank built an in-house AI helper using ChatGPT and Codex. Bankers say it's saving them an hour or more a day on grunt work.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Singular Bank, a private wealth outfit based in Madrid, has spent the last stretch of this year quietly rolling out something called Singularity, an internal tool stitched together with ChatGPT and OpenAI's Codex. The pitch is simple enough: bankers spend a huge chunk of their day on prep work that has nothing to do with actually talking to clients, and Singularity is meant to claw that time back.
According to the bank, the tool is shaving 60 to 90 minutes off the daily workload of relationship managers. That time was going toward things like assembling notes before a client meeting, digging through portfolio data to spot what's changed, and writing up follow-ups afterward. None of that is glamorous work, but all of it eats hours, and in private banking those hours are supposed to be spent building relationships, not formatting spreadsheets.
What makes this rollout notable isn't the use case itself, plenty of banks want AI to handle admin, but the choice to build it in-house rather than bolt on a vendor product. Using Codex alongside ChatGPT suggests Singular's engineering team wrote custom code to wire the assistant into their own systems, rather than accepting a one-size-fits-all interface. That's a heavier lift upfront. It also means the bank controls how client data flows through the system, which matters a lot when regulators and compliance teams are watching everything.
There's a broader signal here about where enterprise AI adoption is heading. A lot of the loudest AI announcements over the past two years have been about flashy consumer features or splashy chatbots bolted onto existing apps. Singular's approach is quieter and, frankly, more typical of how this technology actually earns its keep: not by replacing bankers, but by stripping out the tedious 20% of their job that never should have taken a human that long anyway.
My take — AI-written commentary, not fact-checked reporting
This is the boring-but-real version of AI adoption that actually sticks, unlike the flashy demos that get a headline and then quietly die in a pilot program six months later. Financial services building custom tooling instead of renting a chatbot is the tell that they're serious, not just checking a box for the board. I'd rather see ten more stories like this than another benchmark chart claiming a model passed the bar exam.
Read more about this at: OpenAI