Why Enterprises Shouldn’t Follow Meta’s AI Example

As companies move beyond the pilot stage to scaling and commercialize artificial intelligence, a technology giant is changing the way AI operations are organized in-house. Meta (Facebook’s parent) announced in early June that it would distribute AI to the company and distribute its ownership to Meta’s product group. According to CTO Andrew Bosworth..

“We believe this will accelerate the adoption of critical new technologies across the company and push the boundaries,” Bosworth wrote in a post announcing the changes.

This announcement shakes how AI is organized in Meta, and shows other changes such as the VP of AI Jerome Pesenti leaving the company and the integration of several separate AI teams.

The changes in Meta are questioning other leading companies across the industry. ‘Does Meta’s AI reorganize the examples to follow? What should you think about building your own artificial intelligence research and operations? “

How Enterprises Build Early AI Practices

Often, corporate organizations launch AI as an initiative driven by a single business unit. For example, marketing organizations within a company have long used AI technology, says Erick Brethenoux, lead AI analyst at Gartner. Organizations can then distribute AI practices to business units or product groups, as Meta just said, with the goal of accelerating adoption across the business.

“That’s not new, isn’t it? We’ve seen it many times,” says Bretenu. “By the way, people are moving from centralized to decentralized and from centralized to decentralized. By the way, it’s not just AI. They’ve done it with all kinds of other types of capabilities and capabilities within the enterprise. HR is one example, he says.

Better approach: hybrid

But Brethenoux was surprised to hear that Facebook is moving to a decentralized AI model in the future.

“They have to be one of the most advanced and mature companies,” he says. “I was surprised to see that they were doing what my client had done before, but they left.”

Instead, these companies that tried and abandoned the approach adopted by Meta (Brethenoux calls them his most mature clients) are under a model that is a hybrid of centralized and decentralized AI. It is operated.

How hybrid AI works

Here’s how he explains how they organize their practice. These companies typically start practicing AI under a particular business unit and then evolve to find a way to syndicate their knowledge of AI into a centralized location (physical or virtual). Science lab. However, these mature companies do not leave the AI ​​lab to operate independently, but also establish an executive committee (steering committee) with the actual authority to decide the project of this AI lab.

Next, this AI lab reports to corporate functions, not business units. why? Brethenoux states that this reporting structure establishes two important things. The first is neutrality between the various business units. The second is to make sure that the selected project is in line with the company’s overall strategy.

It may sound like a centralized approach. But these companies aren’t the only ones, says Brethenoux. Next, select an AI expert from the AI ​​lab and rotate it to various business units. These professionals spend 6-12 months in Business Unit 1, then move to Business Unit 2 and spend the same time there. After the full tour, they will return to the AI ​​lab for 3-6 months before returning to rotation again.

“As AI professionals face the reality of each business unit and understand what’s really happening in the field, they learn from the field,” he says. In addition, “they disseminate knowledge.” Rotation AI experts bring the resolved issues of one business unit to another business unit that may have similar issues.

“when [organizations] If you focus the model somewhere and let people rotate the entire business function, they find it to increase retention, “says Bretenu. “As AI professionals are exposed to and solve a variety of problems, knowledge sharing is focused. This helps to retain the normally curious people, and AI professionals They are usually curious people. “

This is Brethenoux’s current recommended approach, big or small, looking for the best AI setup within your organization. It may look a little different depending on the industry. Telecommunications is different from automobiles, and automobiles are different from pharmaceuticals. But the skeleton of the setup is the same in all industries, he says.

Due to multiple pandemic crises and all the sequelae of the pandemic (supply chain disruption, remote work, etc.), organizations have accelerated the transition to this type of setting for the practice of artificial intelligence, Brethenoux says. The timeline for technology initiatives has been accelerated.

A hybrid approach may be the answer for IT organizations looking to maximize the value of AI programs across the organization.

“People are starting to focus on the results of what AI can produce, not the technology itself,” Brethenoux says.

What to read next:

Very Successful Chief Data Officer Priorities

AI set to disrupt traditional data management practices

Why your company should evolve from data-driven to decision-driven Why Enterprises Shouldn’t Follow Meta’s AI Example

Back to top button