Everyone is chasing the next AI headline. CEOs posting

When a Bank CEO Lets His AI Clone Handle the Earnings Call

Everyone is chasing the next AI headline. CEOs posting screenshots of ChatGPT outputs. Banks putting “AI-powered” banners on loan applications. The surface-level view: AI is just the latest tool to polish the same processes we’ve run for decades. Tweak efficiency. Save a few hours. Maybe automate some back-office forms.

But then Sam Sidhu, CEO of Customers Bank, does something that breaks the script.

He lets his AI clone deliver the prepared remarks for an earnings call. Not as a gimmick, but as a signal of a deeper transformation. Within a year, his bank is signing a multiyear partnership with OpenAI to drive operational efficiency and product innovation.

Directly targeting a sub-40% efficiency ratio and cycle times that make legacy banks look like they’re running on dial-up.

Most people saw the AI clone and focused on the optics. The real move is systemic. Sidhu isn’t just playing at the edge of automation.

He’s rearchitecting how the bank works, who does the thinking, and how scale actually happens.

The biggest mistake?

Treating THIS as a surface-level skill. It’s not about putting a chatbot on your website or letting a synthetic voice read your script. It’s about moving from AI as a parlor trick to AI as a core operating system.

The Surface Game: Mistaking Output for Transformation

The average enterprise sees AI as a bolt-on.

Something to make an existing workflow faster, maybe a little smoother, but fundamentally unchanged. I see this mistake everywhere: a process is handed to an LLM, and people marvel when the cost per ticket drops or a form processes in seconds instead of minutes. The AI is measured by its ability to imitate a human.

Or more often, to substitute for one.

That approach never escapes the surface. You get speed gains, but the structure of the work doesn’t evolve. It’s the same conveyor belt, just with a faster motor. The bank still closes commercial loans in 30-45 days. It still staffs for process friction. It still assumes that “efficiency” means doing the same thing, just with fewer keystrokes.

This is why so much “AI innovation” stagnates after the headlines fade.

A chatbot answers FAQs. An LLM drafts a few emails. The actual bottleneck.

How decisions are made, how value flows through the system, how human and machine intelligence are orchestrated.

Remains untouched.

Sidhu’s AI clone made headlines for its novelty.

But if that’s all you see, you miss the point. The prepared remarks weren’t the product. They were the interface to a system that had already shifted beneath the surface.

Instead of just automating the words, Customers Bank started automating the work. Sidhu didn’t use AI to patch a process. He used it to dismantle and rebuild the model.

Pushing the efficiency ratio from 49% toward the low 40s, and shrinking commercial loan cycles to a fraction of their legacy duration.

That kind of delta doesn’t come from faster text generation. It comes from rethinking what workflows are for, what steps are necessary, and which thinking jobs can be productized in software. Not “let’s automate the call script”.

But “let’s automate the work that makes the call script necessary in the first place.”

When you see 28,000 hours saved (the equivalent of fifteen full-time staff), that didn’t happen because a bot answered a few more emails. It happened because the system itself was rearchitected to think and act differently.

I remember a project where a logistics company tried to implement AI for route optimization. They just slapped it on top of their existing system, expecting magic. What they got was faster routes but no real change in delivery times because the warehouse processes were still manual and cumbersome. It took reimagining the entire workflow, not just the routing, to see real transformation.

The Correction: AI as Operating System, Not Surface Patch

The real move

The one most practitioners miss

Is using AI as the backbone of a new operating model. Sidhu’s earnings call clone mattered not because of what it said, but because of what it represented: the organization’s willingness to let AI step into core, even symbolic, decision moments. If you’re comfortable letting an AI represent your leadership to investors, you’re also ready to let it drive the work that delivers the results.

This isn’t about synthetic voices or cloned personalities.

It’s about letting AI own the flow of information, the logic of decisions, and the shape of outcomes. That’s why the partnership with OpenAI matters. It’s not a vendor deal. It’s a handshake around the architecture of the business itself.

Look at the actual changes on the ground.

Commercial loan cycles drop from up to 45 days to just 7. Complex account openings, once a multi-hour manual effort, now take under 20 minutes. These are the kinds of numbers you only get when you use AI not as a patch, but as the new substrate for how work is done.

This is the core correction: Stop treating AI as an overlay. Treat it as the system.

You don’t bolt it on to existing roles. You redesign the roles themselves.

Moving the locus of decision, exception, and value creation from human operators to orchestrated intelligence.

The bank doesn’t just save labor. It creates new capacity. It can scale growth without growing headcount on a linear curve. The logic of the business is re-written.

There’s a subtle but crucial difference between “automation” and “intelligence.” Automation cuts costs by removing steps. Intelligence compounds value by inventing new ways of working. That’s the leap Sidhu is making. The OpenAI partnership is about co-creating enterprise tools, not just buying off-the-shelf scripts. That’s why he talks about selling solutions to other banks.

Because the product is the operating model itself.

This is the only way small and midsize banks can compete with giants. Not by matching their call centers or balancing their ad spend, but by fundamentally outpacing them at the level of cognition.

Making the bank itself smarter, faster, more adaptive.

If you’re still thinking of AI as a tool to help your human staff do the same work a bit faster, you’re missing the point. The winners are those who see AI as the new core worker, reshaping the value chain from the ground up.

Beyond Optics: Building the AI-First Bank

There’s a lesson here for every practitioner. Don’t let the surface headlines fool you into thinking the hard work is about cloning a voice or generating a script. The real transformation is invisible until it manifests as compounding operational leverage.

Sidhu’s playbook isn’t about publicity.

It’s about architecture. The AI clone on the earnings call is a signal that the bank is ready to trust AI with higher and higher levels of agency. Once you cross that threshold, every process is open for redesign.

Every bottleneck, every handoff, every decision.

That’s how you go from incremental efficiency to exponential capacity.

The fix isn’t complicated. It’s about letting AI move from the edge to the center of the business, and then building everything else around it. Not as a tool for surface-level wins, but as the operating system for how value is actually created and delivered.

If you want to see where this leads, watch the banks that aren’t just automating scripts, but rearchitecting the system itself. That’s the play.

The only one with the leverage to matter as AI becomes the new baseline.