In a keynote interview at ServiceNow Knowledge 2026, Vishal Talwar, executive vice-president and chief digital and information officer of FedEx, laid out the company's ambitious plans for agentic artificial intelligence. Agentic AI refers to systems that can act autonomously, making decisions and performing tasks without constant human oversight. For a global logistics company handling millions of packages daily, the potential is enormous — but so are the risks.
Why agentic AI matters for logistics
FedEx operates one of the most complex transportation networks in the world, coordinating planes, trucks, sortation centers, and delivery personnel across 220 countries. Every day, the company processes data from package tracking, route optimization, weather forecasts, and customer inquiries. Traditional AI has already been used to improve efficiency, such as predicting delivery times or rerouting shipments around delays. But agentic AI goes further: it can autonomously execute actions based on real-time data, such as deciding to redirect a shipment to a different hub if a facility is overwhelmed, or proactively contacting customers with alternative delivery options.
Talwar emphasized that agentic AI is not about replacing humans but augmenting their capabilities. “We see this as a way to free up our team members from routine decisions so they can focus on more complex, value-added interactions,” he said. For example, a customer service agent could be assisted by an AI that has already gathered context from previous interactions, suggested solutions, and even initiated a refund or reshipment authorization — subject to human approval for high-value cases.
Challenges in scaling agentic AI
Scaling agentic AI across an enterprise of FedEx’s size presents multiple hurdles. First, there is the question of trust. Autonomous systems must be reliable enough that employees and customers feel comfortable with their decisions. Talwar noted that FedEx is taking a phased approach: starting with low-risk, internal-facing tasks, then gradually expanding to customer-facing applications. “We want to build confidence incrementally,” he said. “Every action an agent takes must be auditable and explainable.”
Second, security and privacy are paramount. An agentic AI system that can access customer data, shipment details, and internal logistics databases must be protected against misuse or malicious attacks. FedEx is implementing strict access controls, continuous monitoring, and human-override mechanisms. Talwar stressed that “any agent that can execute an action must have a kill switch — a way for a human to step in if something goes wrong.”
Third, integration with existing IT infrastructure is complex. FedEx runs on a mix of legacy mainframes, cloud platforms, and modern microservices. Agentic AI needs to interface with all of these. Talwar described a strategy of building an “agentic layer” that sits above existing systems, using APIs and event-driven architectures to orchestrate actions without disrupting core operations. This layer also provides a unified governance framework to ensure that agents comply with regulatory requirements, such as data localization laws in different countries.
The role of partnerships and platforms
FedEx is not building its agentic AI capabilities entirely in-house. The company has partnered with ServiceNow, which provides a platform for workflow automation and AI agents. At Knowledge 2026, ServiceNow announced new features specifically for agentic AI, including a “reasoning engine” that allows agents to break down complex tasks into steps, and a sandbox environment for testing agents before deployment. Talwar said these tools help FedEx move faster while maintaining control.
He also highlighted the importance of data quality. Agentic AI is only as good as the data it learns from. FedEx has been investing in data cleaning, labeling, and governance for years, and that foundation is now paying off. “If your data is messy, your agents will make messy decisions,” Talwar warned. The company uses a combination of structured data from its operational databases and unstructured data from customer communications and sensor logs.
Industry context and comparisons
FedEx is not alone in exploring agentic AI. Competitors like UPS and DHL are also experimenting with autonomous logistics systems. However, Talwar believes FedEx’s early investments in digital transformation give it an edge. The company has been using AI for route optimization since the early 2000s, and its “FedEx Surround” system uses real-time data to predict disruptions. Agentic AI builds on these capabilities, enabling the system to not just predict but act.
Other industries are watching closely. In healthcare, agentic AI could schedule appointments, manage inventory, and even assist in diagnosis — but the stakes are higher. In finance, autonomous trading agents have been used for years, but with mixed results. The lessons from logistics, where the cost of a mistake might be a delayed package rather than a financial loss, make it a fertile testing ground. Talwar argued that “logistics is the perfect sandbox for agentic AI because the consequences are manageable, and the potential for efficiency gains is massive.”
Employee training and cultural change
Implementing agentic AI also requires a cultural shift. FedEx is investing in training programs to help employees understand how to work alongside AI agents. This includes upskilling in data literacy, understanding when to trust an agent’s recommendation, and knowing how to override it. Talwar said the company has created “AI champions” in each department who serve as liaisons between the technology team and frontline workers. “We need everyone to feel they have a stake in this transformation,” he explained.
Moreover, FedEx is mindful of ethical considerations. Agentic AI could potentially be used to make decisions that affect people’s jobs, such as reassigning drivers or adjusting routes that impact working hours. Talwar emphasized that the company consults with employee representatives and ethics boards before deploying any new autonomous capability. “We are not trying to automate people out of jobs; we are trying to automate the boring parts of their jobs,” he said.
Future outlook
Looking ahead, Talwar sees agentic AI becoming a core part of FedEx’s operations within the next three to five years. Near-term pilots will focus on warehouse automation, such as having AI agents direct autonomous forklifts and sort robots. Longer-term, the company envisions agents that can negotiate with suppliers, optimize global shipping routes in real time, and even predict customer needs before they articulate them.
But Talwar cautioned against hype. “The technology is exciting, but we must be disciplined about deployment. We will not release an agent that we cannot fully trust.” He advised other CIOs to start small, choose use cases where the impact is measurable, and invest heavily in governance from day one.
As FedEx continues its journey, the lessons it learns will likely influence how other enterprises approach agentic AI. The company’s pragmatic, safety-first approach may become a blueprint for responsible adoption of autonomous systems across industries.
Source: Computerweekly News