Singular Bank speeds up bankers' workflows using ChatGPT and Codex

If you work in banking or financial services and wonder how AI actually fits into a regulated, client-facing environment, Singular Bank's Singularity project is worth a close look. The Spanish wealth management firm built an internal assistant on top of ChatGPT and Codex, and the results are tangible: bankers report saving between 60 and 90 minutes every day. This review-style walkthrough explains what Singularity does, how it was built, what it delivers, and whether a similar approach could work for your own team.

What Singularity actually is

Singularity is not a client-facing chatbot and not a trading algorithm. It is an internal productivity assistant designed for Singular Bank's own bankers and advisors. The tool combines OpenAI's ChatGPT for natural language tasks with Codex for anything involving code or structured data work.

The idea is simple but practical. Bankers at wealth management firms spend a huge portion of their day on preparation rather than advice: pulling together meeting notes, reviewing client portfolios, drafting follow-up emails, and summarizing documents. Singularity takes over much of that groundwork so advisors can spend more time with clients and less time shuffling paperwork.

The three tasks where it shines

Meeting preparation. Before a client meeting, an advisor typically needs to review the client's history, recent portfolio movements, and any relevant market developments. Singularity condenses this into a digestible briefing, cutting what used to take half an hour or more down to a few minutes of review.

Portfolio analysis. The assistant helps bankers interpret and summarize portfolio data. Rather than manually digging through reports, advisors can ask questions in plain language and get structured answers back. Codex plays an important role here, since parts of this work involve handling structured data rather than free text.

Follow-up work. After a meeting, there are notes to write and emails to send. Singularity drafts these based on what the advisor provides, turning a tedious end-of-day chore into a quick review-and-approve step.

Why the numbers matter

A saving of 60 to 90 minutes per person per day might not sound dramatic at first. But spread across a working week, that is five to seven and a half hours — essentially an extra working day. Multiply that across a team of advisors, and the capacity gains become substantial.

What makes this case interesting for the wider industry is that the gains came from everyday tasks, not from anything exotic. Singular Bank did not try to automate investment decisions or replace human judgment. They targeted the unglamorous administrative layer that surrounds every advisory conversation, and that is exactly where generative AI currently performs best.

The build approach: existing tools, not custom models

Another point worth noting: Singular Bank did not train a proprietary model from scratch. They built Singularity on top of ChatGPT and Codex, meaning the engineering effort went into integration, workflows, and guardrails rather than model development.

For most firms outside the tech sector, this is the realistic playbook. General-purpose models are already highly capable; the value comes from connecting them to internal data and processes in a controlled, secure way. In a regulated industry like banking, that also means attention to data privacy, compliance, and keeping sensitive client information properly protected — challenges any financial institution adopting a similar tool would need to address seriously.

What other firms can learn

Three takeaways stand out from Singular Bank's experience:

1. Start with internal use. An assistant for your own staff carries far less risk than a client-facing product, and it lets you build confidence and processes before expanding.
2. Target time sinks, not decisions. Preparation, summarization, and drafting are where AI delivers quick, measurable wins. Leave judgment calls to humans.
3. Measure in hours saved. The 60–90 minute figure is a clear, understandable metric. Vague claims about "innovation" are much less persuasive to leadership than a concrete daily time saving per employee.

Who should consider a similar approach / who should skip it

Consider it if: you run an advisory, consulting, or client-service team where professionals lose significant time on meeting prep, report writing, and follow-up correspondence. Singularity's model — a general-purpose AI assistant tuned for internal workflows — is repeatable, and the tools behind it (ChatGPT and Codex) are commercially available.

Skip it for now if: your workflows are already highly automated, or if you cannot yet satisfy your compliance and data protection requirements around external AI services. In those cases, the foundation needs to come first. AI assistants amplify good processes; they cannot substitute for them.

Singularity is a modest-sounding project with outsized implications: an hour a day, per banker, gained not through replacement but through removing drudgery. That is a realistic and encouraging template for how AI enters knowledge work.