According to Gartner, half of all supply chain software could be run by AI agents by the end of the decade. At this week’s Gartner Supply Chain Symposium/Xpo in Barcelona, analysts said AI is quickly shifting from a tool that supports decision-making to one that makes decisions on its own.

What’s Related

One of the key takeaways was the rise of what Gartner calls agentic AI—a network of smart AI assistants that can handle supply chain tasks without constantly needing someone to guide them. These agents could buy materials, adjust inventory, respond to demand changes, and even suggest new business models.

“Supply chain technology investments need to improve efficiency or contribute to profit growth in today’s climate of uncertainty,” said Kaitlynn Sommers, Senior Director Analyst at Gartner. “Agentic AI has the potential to do both by providing a new means to enhance resource efficiency, automate complex tasks, and introduce new business models across supply chains.”

The first stage of agentic AI includes specific AI helpers, like a digital buyer that automatically places orders based on stock levels, demand forecasts, and market prices. Over time, these AI agents will get smarter, handle more complex jobs, and take more independent action based on the company’s needs.

Sommers explained how this is different from older automation systems. “Agentic AI represents a revolution from robotic process automation as the AI agents will autonomously complete tasks without relying on explicit inputs or predefined outcomes. Agents will continuously learn from real-time data and adapt to evolving conditions and complex demands.”

To help companies get started, Gartner shared three pieces of advice for Chief Supply Chain Officers:

1. Start with the right use case — Choose small, focused areas where AI can scale, work with clean data, and show strong ROI.
2. Work AI into your processes — Treat AI agents like team members and assign them specific roles. Companies may need to rethink how teams and workflows operate.
3. Set clear boundaries — Make sure AI agents know the limits of what they’re allowed to do, to avoid costly mistakes.