Reimagining operations with agentic AI at Toyota

How is agentic AI driving Toyota’s digital transformation?

Toyota has been using various forms of artificial intelligence for several years, beginning with machine learning optimization engines and, more recently, generative AI embedded into the company’s technology products. “Now we’re leaning into agentic AI,” Ballard says, “starting with single agents and working toward a system of agents to help our different operations.”

Take the company’s resource allocation process to meet customer demand. Historically, the operation had been “pretty messy,” Ballard explains, involving 75-odd spreadsheets, 50-plus team members, and hours and hours to come up with the right plans for suppliers and manufacturing teams.

The digital innovations team is building a global planning system that shrinks that team to 6 to 10 planners, allowing the remaining members to be redistributed to other, more pressing areas, and eliminating all those spreadsheets. The process starts with an AI agent that pulls in demand data, looks at supply, and walks the planning team through possible scenarios. Agentic AI “handles the routine, repetitive tasks and lets the team members make the advanced decisions around which scenarios they choose,” Ballard says. “It helps them with the constraints. It helps them optimize for revenue or other factors. It gives them models in minutes rather than all those hours of overtime.”

Another use case for AI agents is managing performance against a key metric: estimated time of arrival (ETA). Toyota has relied on decades-old mainframe systems to determine where a vehicle is in the pipeline and whether the company can meet its customer commitments. A new vehicle management tool retires 50 to 100 mainframe screens and provides team members with real-time data on each vehicle’s journey, from premanufacturing through delivery to the dealer.

Agentic AI has the potential to take things further. If a team member wants to know how many vehicles are delayed in the West and what’s causing the issues, they can use an agentic AI–powered automation prompt to get a status report on vehicles in that specific region. To achieve this, the digital innovations team is currently training agents to address delays. Say, a vehicle is in the yard ready to be loaded on a truck but, for some reason, is just sitting there. The agent can draft an email to Toyota’s logistics provider, asking them to place that vehicle in a particular bin on a specific truck and expedite the process to maintain the ETA. The agent can also communicate directly with the dealership to explain what’s being done to get the ETA back on track. “And [the agent] can do all these things before the team member even comes in in the morning,” Ballard says. Once fully implemented, these capabilities are expected to allow human employees to focus on higher-order work, like preventing such problems from recurring. 

Real value lies in process redesign and people

The real value of agentic AI is not in automating existing processes—something many companies did with their initial implementations for incremental gains—but in process redesign. “We’re reimagining our entire operations,” Ballard says. “We’ve made that critical decision to just go ahead and invest in this area a bit further. We feel like that’s where the differentiator is going to be going forward.”

The technology driving Toyota’s AI ambitions

The infrastructure underneath it all is a public cloud platform. A data hub layer provides real-time, connected data from across the value chain—from suppliers to customers. On top of that is a services layer, and above that is the intelligence layer, where agentic AI lives. Sitting atop those are Toyota’s technology products, all sharing a similar user interface or user experience, accessible through a common portal called “Cube.” “Everything we do, we lead with the platform,” Ballard says. “And that makes it a seamless experience for our team members, whether they’re on the supply chain, manufacturing, or sales teams.”

Ballard’s group is currently integrating agentic AI into what it calls the Cube Command Center to provide greater observability of systems and processes. Right now, those agents are monitoring and managing uptime. Down the line, agentic AI will monitor costs and agent interdependencies as usage expands. “The writing is on the wall for [agentic AI] to completely transform how work happens going forward,” Ballard says. “We’re not implementing AI agents because it’s exciting. We’re implementing them because we believe agentic AI needs to be integrated into our operations to better serve our customers, our team members, and our company overall.”