Let AI Go 'A Little Bit Rogue,' SharkNinja CIO Says
SharkNinja's Velia Carboni wants AI pushed to its limits before rivals do. PepsiCo and Salesforce executives explain guardrails, costs and digitizing executive judgment.
By Daniel Okafor
4 min read
Updated

What's News
- SharkNinja CIO Velia Carboni said she wants AI 'a little bit rogue' so the U.S. finds its limits first.
- PepsiCo's Athina Kanioura said 99% of AI outputs are trusted without traceability in China, unlike the U.S.
- Salesforce is building a digital twin of president Robin Washington to surface business correlations.
- SharkNinja, a $27 billion company, is digitizing CEO Mark Barrocas's judgment for inventory decisions.
- Salesforce gives each employee an AI spending dashboard after opening broad access and seeing the bill.
SharkNinja's chief information officer wants AI to go "a little bit rogue" — because finding the technology's limits first is a strategic advantage. "I want them to be a little bit rogue because I'd rather we find [the limits] than someone else finds it," said Velia Carboni, CIO of the $27 billion appliance maker. "We wouldn't want to lose that as a country."
Carboni spoke at Fortune's Most Powerful Women Summit in Santa Barbara, Calif., alongside Athina Kanioura, PepsiCo's global chief strategy and transformation officer and CEO of its Latin America Foods business, and Shibani Ahuja, senior vice president of data and AI strategy at Salesforce. Fortune's Diane Brady moderated the panel and asked directly whether AI will kill us.
"For sure, it won't kill us," Kanioura said.
Why do executives think the AI panic misses the point?
The panelists argued the public discourse fixates on the wrong thing. Much of the recent alarm stems from AI labs' own stress tests — and their sometimes tardy or incomplete disclosures — in which frontier models are pushed to deceive, break rules, or resist shutdowns.
"The models themselves are not AI. AI is so much more," Ahuja said. The recent media coverage, she said, describes "testing that is by design and intentional around the models, and the models going rogue by design."
Ahuja said she wants more discussion on two points:
- The guardrails labs use while testing their models.
- The "harnesses" companies build to use the models safely day to day.
How do companies police AI without killing creativity?
At SharkNinja, guardrails are firmly in place, but Carboni does not lead with them when she talks to employees. "I don't want to limit creativity," she said. "As a CIO, I have to have the guardrails and security… but I don't lead with that story because I think that also takes away some of the creativity."
Kanioura framed the burden differently. PepsiCo has 320,000 employees worldwide, billions of consumers, and it feeds and hydrates the population. The company uses AI "everywhere," she said, and the onus is on businesses of PepsiCo's size being "extremely responsible" with the technology across every interaction with consumers, customers, employees and operations.
"If I felt that it would kill and destroy, then we would be the first ones not to use AI in every part of our operations," Kanioura said. She added that tech companies building labs are not the ones engaging with consumers. "They are engaging with consumers through us," she said.
Does trust in AI vary by market?
Sharply, according to Kanioura. "In China, 99% of the cases they trust the output," she said. Workers there do not ask for traceability. In the U.S., employees want proof of sources and evidence of bias, and they often signal that they "want to control everything," she said. As a result, Pepsi's guardrails differ by country.
Cost shapes deployment too. Ahuja said Salesforce initially opened access to AI tools broadly — and then saw the bill. Instead of cutting access, the company studied its power users and gave each employee a spending dashboard. "You are now a leader; you have a budget," she said.
Where is the line between AI and human judgment?
All the panelists agreed judgment divides what AI can be trusted to do from what still needs human guidance. Ahuja noted early AI systems were too unpredictable to leave alone. Pairing models with deterministic guardrails now lets agents do more work, with humans spot-checking instead of redoing it.
"Judgment is a combination of having the facts and the data, but also feeling," Ahuja said. Salesforce is building a digital twin of its president and chief operating and financial officer, Robin Washington. The goal is not to replicate her but to surface correlations across the business no person could compute in that amount of time.
SharkNinja, which launches dozens of new products per year in up to 15 countries, is digitizing some of the judgment of CEO Mark Barrocas and his top commercial leaders — for decisions such as where to move scarce inventory. "So you don't have to pick up the phone and call Mark and say, 'Hey, what should we do?'" Carboni said.
The gap between lab stress tests and boardroom practice is closing: the companies that harness judgment — and meter the bill — will set the pace.
Source: Fortune
More from Daniel Okafor
Show full bio
Correspondent covering money, funding and market shifts around entrepreneurship, startups and venture funding at Business Bearings.
1,038 articles