Meta creates ‘Business AI’ group led by ex-Salesforce AI CEO Clara Shih

Meta has hired the former CEO of Salesforce AI, Clara Shih, to lead a new “Business AI” group.

Salesforce has still not publicly acknowledged Shih’s departure, but it will be hard to ignore now that she’s announced her new role on her LinkedIn profile. Shih is now a vice president at Meta and the head of a new business AI group, she said in a post there.

“Meta’s global reach and leadership in AI represent a generational opportunity for businesses, and I couldn’t be more excited and grateful to help take this from zero to one to scale,” she said.

“Our vision for this new product group is to make cutting-edge AI accessible to every business, empowering all to find success and own their future in the AI era.”

Commercializing Llama

Shih may be building the business unit from scratch, but its technology core is already there, in the form of Meta’s Llama large language models.

“Meta’s Llama models have over 600M downloads to date, and Meta AI has more than 500M monthly actives,” Shih said.

From the start, Meta has made the Llama models available to other enterprises under a license it describes as “open source,” but the creation of the new business group makes clear that Meta’s interest is commercial, not philanthropic.

Meta’s licenses and its acceptable use policy contain numerous restrictions on how enterprises may use the models, flying in the face of traditional definitions of open source software and in particular of the new Open Source Initiative definition of open source AI.

OSI’s Open-Source AI Definition requires that systems “must be made available under terms and in a way that grants the freedoms to use the system for any purpose and without having to ask for permission,” and to “share the system for others to use with or without modifications, for any purpose.”

Keeping control

However, anyone wanting to use the latest Llama 3.2 model must agree to a 630-word acceptable use policy that, for example, excludes its use by governments in national security roles unless Meta grants permission, as it did recently to the US government.

Meta also has a veto on competitors using Llama for anything too big: The Llama 3.2 license also requires that enterprises offering services to more than 700 million monthly active users must request an additional license from Meta, “which Meta may grant … in its sole discretion” and may not use the software until Meta expressly allows it.

Since Meta licenses each of its AI models separately there’s nothing stopping it from lowering that threshold for future versions to bring more applications for the software under its control or demanding financial compensation for broader usage licenses.

Full circle

Finding ways to make money from business AI is clearly on Shih’s to-do list: In the post announcing her new role, she thanked four Meta execs for the new opportunity: CEO Mark Zuckerberg, COO Javier Olivan, Chief Strategy Officer David Wehner — and VP of Monetization John Hegeman.

Shih has clearly been considering the move to Meta for some time — and is no stranger to the company’s social media platform, Facebook.

During her first stint at Salesforce, from 2006 to 2009, she worked on a side project with a colleague to build Faceforce (later known as Faceconnector), an app for sales staff to connect their social graph with their CRM system using public APIs from Salesforce and Facebook.

That experience led her to create Hearsay Systems, a software developer combining social networks, CRM, and AI to help sales staff at financial services companies.

She returned to Salesforce in 2020, first leading its Service Cloud team and then Salesforce AI, culminating in the launch of the company’s Agentforce autonomous bot platform.

“The past four years at Salesforce have been some of the most rewarding in my career,” but “There was only one call I knew I would answer, and it was Meta’s,” she said in the post. “Today truly feels like a full-circle moment.”