There aren’t many tech companies that can look back at their history over multiple decades, but 41-year-old Qualcomm is one. Durga Malladi, executive vice president and general manager, gave me the abbreviated history.

The company, founded in 1985, started as a pioneer in wireless communications. It worked to develop the CDMA standards upon which cellular communications were built, and pivoted to manufacturing cell phones in the mid-1990s.

More recently, the company moved into computer hardware construction. Qualcomm now makes specialized CPUs that can run AI, as well as a variety of chips for AI data centers. Qualcomm data center racks include not only the workhorse chips built for training and memory, but also less intensive chips built for inference.

Looking at the processing needs of data centers, Malladi notes that CIOs should do the same thing: Look at the AI processing needs of your company, and run your AI through different sources. Some should go through data centers, but some can be handled on your device and some could be processed through an on-premises server. He told me why this is good for cost and security, and more is later in this newsletter.

This is the published version of Forbes’ CIO newsletter, which offers the latest news for chief innovation officers and other technology-focused leaders. Click here to get it delivered to your inbox every Thursday.

Artificial Intelligence

Many enterprises have given AI agents the keys to make big decisions—and are watching them get things wrong. A new report from risk mitigation platform Optro shared exclusively with Forbes shows many companies have had big problems.

A quarter have had a control failure with an agent, 30% have seen an agent take an unintended action, and about a third have seen agents make inaccurate decisions that were acted upon. And while more than two-thirds of organizations said they could stop agents or detect issues quickly, 28% said they only discovered an agent’s mistake after it caused downstream impact.

Guru Sethupathy, Optro’s general manager of AI governance, said these problems are concerning and should give company leaders pause—but aren’t unprecedented.

“When a technology comes in, a lot of the reason it takes a while for it to become really valuable is the change management around figuring out: What is this technology actually good at? What does the human need to do? Where does the human play a role? And how do the human and the technology interact?” he told Forbes. “And so part of what we’re experiencing now is actually the messiness of........

© Forbes