Customers Bank, a $25.9 billion asset lender serving startups and small businesses, has signed a multiyear partnership with OpenAI that will embed AI engineers directly inside the bank to build autonomous agents for lending, deposits, and payments. CEO Sam Sidhu disclosed the deal on April 27, shortly after revealing that an AI clone had delivered his prepared remarks during the bank’s first-quarter earnings call.
The Deal
OpenAI will station engineers at Customers Bank to co-develop what Sidhu described as “end-to-end, automated agentic led workflows” across three core banking functions: lending, deposits, and payments. The bank expects to roll out agents across all three within six to 12 months.
The arrangement goes beyond typical software licensing. Both sides contribute resources, and Sidhu told CNBC the companies will “co-create enterprise solutions they could potentially sell to other banks in the future.” OpenAI gains a real-world testing environment inside a regulated financial institution. Customers Bank gets engineering capacity it could not build internally at the same speed.
OpenAI Chief Revenue Officer Denise Dresser said in a statement that the company was “proud to help Customers Bank as they build a more intelligent operating model that empowers employees, strengthens client service, and sets a new standard for regional banking,” according to CNBC.
Operational Targets
The numbers Sidhu is committing to are specific. Closing a commercial loan currently takes 30 to 45 days across underwriting, document collection, and legal negotiation. With agents, Sidhu told CNBC that timeline drops to approximately seven days. Opening accounts for complex commercial clients, currently a process lasting more than a day, is targeted at under 20 minutes using conversational AI and automated document gathering.
At the financial level, Sidhu expects the bank’s efficiency ratio to improve from roughly 49% to the low 40s, with higher returns beginning in 2027. The bank already uses AI to write half its software code and has logged 28,000 hours saved, equivalent to roughly 15 full-time employees not hired.
Why a Regional Bank Moves First
Sidhu’s argument for why Customers Bank can move faster than JPMorgan or Goldman Sachs comes down to regulatory complexity. Megabanks with trillions in assets face far more stringent AI implementation frameworks. “Smaller banks are not going to be expected to have the same level of frameworks as many of the larger banks,” Sidhu told CNBC. He added that regulators want community and regional banks “to be able to compete with larger banks.”
The bank is also exploring new business lines that would have been too expensive to staff before agents. These AI-native operations would run with smaller teams overseeing automated systems handling work that previously required large headcounts.
OpenAI’s Financial Services Push
The partnership fits OpenAI’s broader strategy of targeting finance as a core vertical. The company has previously hired former bankers to train its models on financial workflows, according to Bloomberg. Embedding engineers directly at a banking client represents an escalation of that strategy, moving from selling tools to co-building the operational infrastructure.
Customers Bank first engaged OpenAI in 2023 through Sidhu’s venture capital connections. The deal signed last week broadens the relationship from an arm’s-length software arrangement to an integrated engineering partnership.
The Regional Banking Bet
The AI clone stunt on the earnings call was calculated theater, but the underlying bet is straightforward: a $25.9 billion regional lender is wagering that agents can compress its cost structure faster than larger competitors can navigate their own compliance requirements. If the efficiency ratio targets hold, the bank becomes a case study for every mid-market lender evaluating the same trade-off. If they don’t, the multiyear commitment still gives OpenAI something potentially more valuable: a production environment inside a regulated institution where it can build reusable financial services tooling.