News

What it really takes to train an entire workforce on gen AI

Westend61 | Westend61 | Getty Images

What it really takes to train an entire workforce on gen AI

  • Training a large workforce on generative artificial intelligence is a big lift given the technology's novelty, but organizations are tackling it nonetheless.
  • Microsoft's 2024 Work Trend Index says 66% of leaders wouldn't hire someone new without AI skills.
  • For gaining hands-on experience with generative AI solutions, some companies see hackathons — events that allow all kinds of employees to present new opportunities for the technology's use — as an effective tool.

Companies with hundreds or thousands of employees across a range of functions have their work cut out for them in the upskilling race. Training a large workforce on generative artificial intelligence is an even bigger lift given the technology's novelty, but organizations are tackling it nonetheless — in part for the company's efficiency and innovation, but also for the long-term success of their employees (after all, Microsoft's 2024 Work Trend Index says 66% of leaders wouldn't hire someone new without AI skills).

These organizations pursuing widespread gen AI training are learning lessons along the way.

Take global IT services and consulting organization Synechron, for example. The majority of Synechron's approximately 13,500 team members are AI-enabled, a feat resulting from highly intentional training. Because of the regulated environment many of their clients operate in, Synechron created a suite of nine secure internal solutions, one of which is a ChatGPT-like application called Nexus Chat. Of Synechron's workforce who are not working at restricted client sites, 84% are using Nexus Chat.

"Once we could give people access to that, we could then start teaching them," said Synechron CTO David Sewell. They started with an online course, which taught beginner-level prompt engineering (how to talk to AI and get the most out of it). They also created videos showcasing potential use cases for different non-technical roles, such as human resources or legal, and included questionnaires to fast-track proficiency.

As a sort of trial period, Synechron gave technologists and a small cohort of the general workforce early access to some of the tools, including Unifai, an AI-powered human resources bot trained on sensitive HR policies and company data, which has now been rolled out to 74% of employees.

On the technology side, Sewell says productivity in the software development lifecycle is up 39%. Non-technical roles are harder to measure, but Synechron chief marketing officer Antonia Maneta said, "After only a few months, I don't know how I was operating without AI in my business. It's unbelievable. The change in productivity frees up our time to do the important things, the focus areas."

Amala Duggirala, chief information officer at financial services company USAA, is in the process of planning AI training for the company's 37,000 employees. Her strategy boils down to three key steps.

The first is governance and risk management as a central focus. Then, senior leaders receive training on any solution that passes the governance analysis (senior training takes place largely through in-person sessions with industry experts). Finally, different types of teams receive educational courses, which vary depending on whether their roles create technology, protect the organization from risk or simply use the tools provided to them.

Hack, train, repeat

For gaining hands-on experience with generative AI solutions, USAA and others see hackathons — events that allow all kinds of employees to present new opportunities for the technology's use — as an effective tool. The most recent USAA employee hackathon had record-level turnout from technical and non-technical teams alike, suggesting widespread enthusiasm from software to compliance. Duggirala said they were able to take about 55 use case ideas into a controlled environment for testing. "I was amazed at the amount of interest there was in the organization for this," she adds.

Synechron's Sewell said about their hackathon, "We got feedback from people in the hackathon that they didn't quite feel prepared." He added, "We took that feedback from that event and started to create more training material to educate them. We gave them a bit more space, a bit more time to get familiar with the technology before they would be expected to demonstrate any efficiency gains or benefits from it."

Not all workforces are created equal

Terry O'Daniel is head of security at Amplitude, a digital analytics platform with customers like Atlassian, Under Armour, Walmart and BeReal. O'Daniel, who formerly ran security at Instacart, Netflix and Salesforce, prioritizes clear guidelines and tinkering over proactive training. "We started out trying to boil the ocean and understand all the possible use cases," he said. "We quickly found out that's not what the business was asking for. The business was asking for very practical, tangible ways that they could solve their problems right in front of them."

Still, O'Daniel believes AI evangelization is helpful, and that his IT and security team is responsible for much of that. "We help people understand that we are not the department of 'no.' We're the department of 'try this instead.'" His guidelines, which focus on data privacy and security, intellectual property and output verification, enable employees to ask questions about any platforms they intend to use, often coming to his team to get approval before implementing a new AI solution into their workflow.

For example, Amplitude has a corporate subscription to the OpenAI API feed. O'Daniels makes sure his employees know to use that rather than public solutions that share data.

This is possible for Amplitude with its more than 700 employees, but companies reaching a workforce in the tens of thousands are opting for more structured solutions, as in the case of USAA and Synechron. Even Synechron's evangelization methods are more structured, with the company's head of AI Ryan Cox traveling to a number of global offices and identifying enthusiastic employees who could deliver the message of AI and their training solutions to the local audience.

This variation confirms that companies must find what works best for them, as long as they're not leaving responsible AI usage in the dust.

"We will be behind if we don't think about it," said USAA's Duggirala. "We will be further behind if we don't think about it in a responsible way."

Copyright CNBC
Exit mobile version