Did an AI Really Speak at an Earnings Call?

Nearly 30 minutes into Customers Bank's Q1 2026 earnings call on April 25, 2026, CEO Sam Sidhu told the analysts on the line that the prepared remarks they had just heard, delivered in his voice, covering the bank's quarterly results, had not been spoken by him.

"The prepared remarks you heard on my behalf today were delivered by my AI clone, not read by me," Sidhu said, describing it as a potential first in the history of public company earnings calls.

AI avatars on earnings calls are not new. Klarna CEO Sebastian Siemiatkowski used one in May 2025, and Zoom CEO Eric Yuan followed the same week, with both disclosing the substitution upfront. What Sidhu did was different.

His clone spoke for approximately 30 minutes before the disclosure came, making it the first known instance of an AI clone delivering unannounced prepared remarks to investors on a public company earnings call

Two days later, on April 27, CNBC reported exclusively that Customers Bank had signed a multiyear partnership with OpenAI, an arrangement in which OpenAI engineers will be embedded onsite at the $25.9 billion asset lender to rebuild core banking workflows around AI agents.

What the Deal Actually Involves

The partnership covers three areas of Customers Bank's commercial banking operations: lending, deposits, and payments.

As Piper Sandler analyst Manuel Navas noted in a research note published the same day, what differentiates this arrangement is that OpenAI's personnel will be embedded in Customers' operations to improve its AI tools.

This will make customers the first to implement improved AI tools it is already fully committed to using, while OpenAI gains the ability to improve its models in a live banking environment.

The financial targets Sidhu has attached to the partnership are specific. Commercial loan closing times, currently running 30 to 45 days including underwriting, document collection, and legal negotiations, are targeted to fall to approximately seven days.

Opening accounts for complex commercial clients, a process that can take more than a day, is targeted to collapse to under 20 minutes using conversational AI and automated document gathering. The efficiency ratio is targeted to improve from approximately 49% to the low 40s, with the impact on returns expected to begin in 2027.

"When you have an autonomous agent, you're essentially creating a digital worker… and they can work around the clock," Sidhu told CNBC.

The Co-Creation Angle

The partnership is structured as a two-way relationship rather than a standard technology procurement.

Customers Bank brings deep operational knowledge of a regulated financial institution including the use cases, the compliance constraints, and the friction points that generic AI products have not yet solved. OpenAI brings the frontier models and engineering capability to act on that knowledge at scale.

Sidhu said the output is intended to extend beyond Customers Bank itself. "We're going to be co-creating enterprise solutions they could potentially sell to other banks in the future," he said. "The goal here is end-to-end, automated agentic-led workflow" for lending, deposits, and payments.

OpenAI Chief Revenue Officer Denise Dresser confirmed the direction in a statement provided to CNBC: the company is 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."

Sidhu described the bank's intent in a single phrase: becoming "an AI native bank."

Why a Regional Bank Has an Advantage

​​Customers Bank is not a large institution by the standards of US banking. At $25.9 billion in assets it sits far below JPMorgan Chase's $4.9 trillion. But Sidhu argues that scale is an advantage in the current AI environment, not a liability.

Megabanks have sprawling global operations and significantly higher regulatory complexity governing AI implementation. Customers Bank, which targets the startup and venture capital community and reportedly bid for Silicon Valley Bank in 2023, operates in a more concentrated environment where AI can be deployed faster.

"Smaller banks are not going to be expected to have the same level of frameworks as many of the larger banks," Sidhu told CNBC, adding that regulators want community and regional banks to be able to compete with larger institutions.

The bank has already been building toward this moment. It first engaged OpenAI in 2023, deploying ChatGPT Enterprise. Over the past 12 months it built the operational and governance infrastructure to deploy AI at scale.

It currently uses AI to write half the firm's software code and has saved 28,000 hours of work, equivalent, Sidhu said, to not hiring approximately 15 full-time employees.

The OpenAI partnership extends that trajectory into core revenue-generating functions. Sidhu was direct about what that means for hiring. "This is an opportunity for us to potentially slow that hiring… and do more revenue per employee," he said.

The bank is also exploring AI-native business lines, new operations built around small teams overseeing automated systems that handle work previously requiring large numbers of people.

The Governance Question Nobody Has Asked

Sidhu described the AI clone moment as a potential first in the history of public company earnings calls. So then what are the disclosure obligations when a CEO's AI clone delivers prepared remarks to investors on a regulated earnings call?

Earnings calls are investor-facing communications governed by SEC rules on fair disclosure. The prepared remarks delivered by Sidhu's AI clone covered Q1 2026 financial results, material information in a regulated context.

Whether analysts were informed before the call, whether the substitution required any regulatory disclosure, and whether the SEC has a position on AI-generated executive communications in investor-facing contexts are questions that have not been publicly answered.

The Trump Administration's National AI Legislative Framework, released March 20, 2026, explicitly declines to require transparency about how AI is used in corporate communications.

It proposes no mandatory disclosure standard for AI-generated content in regulated contexts. That absence of guidance is precisely what makes the Customers Bank earnings call a meaningful data point as a test case for what happens when AI enters a disclosure environment that has not yet defined the rules.

For banks and public companies watching how this plays out, the governance implication is direct. If AI-generated CEO communications become common in investor-facing contexts, the question of what requires disclosure will need an answer that neither the SEC nor the current federal AI framework has yet provided.