BNY builds “AI for everyone, everywhere” with OpenAI
OpenAI ● Covered by 4 sources
BNY (the bank) is letting over 20,000 employees build their own AI agents using OpenAI tech through an internal platform called Eliza. The bank says it's not just IT running the show anymore — regular staff are building tools that touch real client 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
Banks talk a big game about AI, but most of it stays locked in a data science team's sandbox. BNY is trying something different. Through a partnership with OpenAI, the 241-year-old financial giant has opened up agent-building to more than 20,000 employees across the company, using an internal platform called Eliza as the on-ramp.
The pitch here isn't a flashy chatbot demo. It's about distributing the work of building AI tools to the people who actually understand the problems — operations staff, client service teams, people far from any engineering title. Eliza gives them a way to spin up agents that handle repetitive tasks, surface information faster, or streamline workflows that used to require a ticket to IT and a multi-week wait.
What makes this notable is the scale. Twenty thousand employees is not a pilot group; it's a meaningful chunk of BNY's workforce. Big banks tend to move cautiously with anything touching client data or regulated processes, so opening agent-building this widely signals real institutional buy-in, not just a press-release experiment.
OpenAI, for its part, gets another marquee enterprise case study, this one in a sector famous for airtight compliance requirements and legacy systems that resist change. If BNY can show measurable gains in efficiency and client outcomes from a workforce-wide AI rollout, it becomes a template other financial institutions will study closely, especially ones still deciding whether generative AI belongs anywhere near core operations or just in the marketing deck.
The real test won't be how many agents get built in the first few months. It'll be whether they keep running six months from now, whether compliance teams stay comfortable, and whether the promised efficiency actually shows up in the numbers instead of just the case study.
My take — AI-written commentary, not fact-checked reporting
I like the instinct here more than the execution details we've actually been given. Handing agent-building to 20,000 people is the right move if you want AI to matter beyond a slide deck, but 'improves efficiency and client outcomes' is the kind of soft language that tells me nobody's published hard numbers yet. Ask me again in a year whether Eliza's agents outlived the hype cycle or quietly got shelved like half the chatbot projects from 2023.”
Read more about this at: OpenAI