The Morning Download: IBM CEO Says AI Demands New Operating Models

Good morning. The evolution of AI models and infrastructure is under relentless scrutiny. Another AI evolution is taking place with a much lower level of visibility and even less understanding. That’s the steady march toward broad and deep AI adoption within companies.

I spoke last week with IBM CEO Arvind Krishna about the massive organizational changes that occur as companies scale AI from individuals to small teams, groups of teams, and ultimately, the entire enterprise. Broadly speaking, he told me that an entirely new operating model is required in order for companies to fully realize AI’s potential at scale. You can read the full story here.

“In the next year or two, the enterprise world will sort into two camps: companies where AI runs their business, and companies where AI is still a project,” Krishna said.

The line between those companies that make broader use of AI and those that don’t won’t simply come down to technology. “It will be their operating model,” Krishna said.

While model developers are competing to stay one step ahead of each other, IBM is approaching AI from a different angle, helping its clients scale their AI efforts. Today, at its Think conference in Boston, the company is announcing a slew of products and capabilities to accelerate that effort, including a new version of watsonx Orchestrate, a secure multiagent control plane, and IBM Bob, for securely building and deploying agents.

To enable AI to scale across the enterprise, processes need to be redesigned end to end and those processes need to be integrated with one another, according to Krishna. IBM’s Project Bob, as the initiative was known internally before it became a product, was designed to manage the entire software development life cycle, from writing new code to patching old code, generating documentation, creating test cases and ensuring security compliance.

“I don’t begin with eliminating steps. I begin with how many touch points can I take out? And how can I make it much more nimble and faster and end-to-end? That’s the goal. Out of that comes the fact that you should eliminate steps,” Krishna said.

That’s the end-to-end connectivity approach driving IBM’s work. And while IBM has made progress rethinking its operating models, more work lies ahead.

“I think it’s early days. We’re only a third of the way through what can be done,” Krishna said.

Trump administration recalibrates attitude towards AI. The White House is considering an executive order that could include formalizing a government oversight group to create standards for the most powerful AI models, such as Anthropic's Mythos, the WSJ reports.

A White House official tells the WSJ that any policy announcement will come directly from President Trump, and discussion about potential executive orders is speculation,

Whatever happens, the internal conversations show just how much Mythos has shifted the administration's attitude towards AI oversight. Previously, the administration was busy unwinding efforts to implement AI safety standards. It has also attacked states trying to impose regulations.

An AI bot may police bank accounts for financial crimes. Anthropic and Fidelity National Information Services, the financial software provider that underpins a swath of the financial system, on Monday announced a partnership to build AI agents for use by financial institutions, the Journal reports.

The first tool: An AI agent to investigate drug traffickers, terrorists and other criminals that use the financial system.

OpenAI signs deal with PE firms. The AI startup raised over $4 billion for The Deployment Company, a new joint venture focused on helping businesses implement its AI software, Bloomberg reports. The venture attracted 19 investors, including TPG, Brookfield Asset Management and Bain Capital. Rival Anthropic  announced its own joint venture with Wall Street firms aimed at selling its AI tools to companies.

OpenAI discussed spinning out robotics, hardware divisions. The proposal, which was ultimately rejected, offers a window into the difficult trade-offs the company has had to make as it races toward an IPO and clamps down on so-called “side quests” that aren’t helping it meaningfully boost revenue.

Data centers in space. Earlier this year, SpaceX filed an application with the FCC petitioning to launch up to a million satellites as part of an “orbital data center.” The WSJ asked an engineer to break down the biggest technical hurdles and costly barriers to putting data centers in space. Watch the video to see what it would actually take to make it work.

Data centers bobbing in the ocean. Peter Thiel is leading a $140 million investment in Oregon-based Panthalassa, which develops floating data centers powered by wave energy. Co-investors include Salesforce CEO Marc Benioff and venture investor John Doerr, who called the effort a "game changer in addressing global energy needs and clean power generation," according to the FT.

Within this shift, certain types of experiences can shine, Hill told me during a recent visit to New York.

Edited highlights below.

What we found is that – and this is not going to hold true with the test of time. What we found today is that they are often managers. And the benefit wasn't that they are managers. The benefit was that they met with ambiguity regularly, that they had to have a part of their job that was dealing with product ambiguity, part of their job that was focused on execution, and a ton of clarity around being able to marry those two into clear deliverables.

Because the tools are now filling in the gaps, which previously were met by decades of understanding how to code efficiently. And so it's this balance of being able to... between product design and engineering, everyone's sort of shifting more into the middle.

— Tom Loftus

Quantum moves closer to drug discovery. IBM researchers partnered with Riken and Cleveland Clinic to demonstrate quantum computing's real-world potential, linking IBM quantum systems to two powerful supercomputers to simulate a protein complex exceeding 12,000 atoms, the largest molecules ever modeled using quantum computing, Barron's reports. 

Not out of the race yet. The Journal profiles a group of Silicon Valley techies and industry misfits brought together by Ford Motor Co. to make electric vehicles that could compete globally with China. Their first model, a new truck Ford says will be nearly as fast as a Mustang and travel around 300 miles on a single charge, is aiming for a 2027 launch.

AI startup Sierra raised $950 million in funding round. Tiger Global and Google's GV led the round, valuing the company at $15.8 billion, joined by existing investors Benchmark, Sequoia and Greenoaks, CNBC reports. Founded three years ago by OpenAI chairman and former Salesforce co-CEO Bret Taylor alongside ex-Google executive Clay Bavor, Sierra sells customer-service agents.