The history of AI at American Honda Motor Co. contains two clearly defined eras: before the public release of ChatGPT and after.
“I think we’re all going to be telling time sort of like B.C. and A.D.,” Bob Brizendine, VP of IT at American Honda, told CIO Dive.
ChatGPT’s launch disrupted the company’s planned strategy for the year, causing leadership to shift priorities and goals to embrace generative AI.
American Honda didn’t jump into the generative AI craze immediately, but by January, the company's leadership knew it needed to rethink its approach to navigate the short term and plan for the future, Brizendine said.
The result was a five-prong generative AI strategy intended to:
- Decide what to do with public versions of tools.
- Bring the capability in-house.
- Identify departments that require more tailored capabilities.
- Retrain internal developers for effective generative AI use.
- Work with software suppliers to understand their roadmap for integrating generative AI.
The motor company wanted to cover its bases by addressing the immediate risks and benefits of generative AI while planning for the technology long term.
Prior to ChatGPT, AI was in the hands of the few while the company worked to improve data and analytics platforms and struggled to get good intel, according to Brizendine. Generative AI democratized access to data insights, and as the company dramatically shifted its approach to AI adoption, it needed to remain nimble.
“You have to seize the opportunity or it will seize you,” Brizendine said.
American Honda isn’t alone in pivoting its plans. Nearly one-third of C-suite technology leaders adjusted their strategy because of generative AI, according to SoftBank data in November.
PwC shifted focus to generative AI in April, laying out a $1 billion, 3-year roadmap to bolster offerings, strengthen tech partnerships and upskill 65,000 employees. Mercedes-Benz plans to invest more than $2.2 billion by 2030 to train employees. The company gave employees access to generative AI via ChatGPT and Microsoft Azure OpenAI service pilots in July ahead of a global rollout.
Solution selection
When surveying the vendor landscape, American Honda judged solutions based on three factors: capability of the tool, risk mitigation and speed of deployment.
The motor company ultimately went with Microsoft’s Copilot, but it is continuing to assess other services. Formerly known as Bing Chat Enterprise, the tool has a chat interface that lets users compare strategies or draft documents. Prompts and responses are not saved or used to train underlying models, according to Microsoft.
“When we think about [Copilot], it provides most of the capabilities of public versions but it adds that responsible AI layer,” Brizendine said. “It allows our teams to get the benefit of ChatGPT without the risks, such as loss of intellectual property, protection against bias and copyright.”
American Honda operates in a Microsoft Office 365 environment, which also drove the decision. “We’ll have other models for the other prongs of the strategy,” Brizendine said, emphasizing the company’s desire to diversify its portfolio as time progresses and the need arises.
Copilot and the public version of ChatGPT satisfy around 80% of the raw count of use cases for the company, Brizendine said. But in some departments, such as research and development, legal or compliance, employees will likely need more tailored solutions.
“We’re standing up individual working areas for those types of use cases,” Brizendine said.
The company rolled out Copilot broadly to workers in November after the IT department did a soft release at the beginning of October. Employees are using the tool to assist in daily tasks, such as research, summarization, ideation and translation.
“It’s really taken root and is being adopted,” Brizendine said. “There are some areas of the business [that] are more inclined to use this type of analysis, such as marketing people, product quality people and those in analytics, but we’re really seeing a pretty uniform adoption by everyone.”
The motor company is also looking to take advantage of embedded capabilities within other existing providers’ products.
“We’ve been actively working with all of our software suppliers, Salesforce, SAP, ServiceNow, and really getting to know what their action plans are in their roadmap for ingraining generative AI capabilities into their toolset, so that we can then begin to socialize that and take advantage of that as part of our platform play,” Brizendine said.
Executing the strategy
As American Honda began working through the details of its strategy, leadership wanted to quickly educate employees on how to leverage publicly available instances of generative AI tools.
For its first statement to employees about generative AI, the company decided to emphasize caution instead of outright banning tool access.
“We purposely chose for our first statement on generative AI to be issued by our risk management officer,” Brizendine said. “What we created was a very balanced statement, trying to educate the entire workforce on the benefits but also to explain the risks, with the loss of IP probably being first and foremost the most vital thing we had to get our arms around immediately if we weren’t going to shut down its use.”
American Honda wanted to make sure its leadership team was educated, too.
“We were trying to move our entire executive management team from being in the camp of curious or interested to educated,” Brizendine said. Each member was required to participate in four-hour informational sessions.
“These were not small commitments of time,” Brizendine said. “It was very interactive and hands-on. We had them develop their own use cases … really getting to know what the technology was about.”
American Honda has now organized content for education covering safe use guidelines, risks and prompt engineering, which include catalogs of effective prompts as examples for staff.
“We’re also doing virtual office hours with our top AI person,” Brizendine said. “It’s maybe an hour a week and it’s for whoever wants to join the Teams session and ask what they want to ask with him giving them advice.”
At the center of it all is the idea that generative AI augments human creativity and productivity but does not replace it.
“We’ve been very, very clear in our messaging that this technology should enable us to take on new challenges with the staff that we have today,” Brizendine said. “We’re going to keep the human in the loop on AI.”
