The difference between a chatbot and an agent is the difference between someone who answers and someone who does. An AI agent in procurement is autonomous software that takes data from procurement systems, makes decisions within defined policy guidelines and authorizations, and executes multi-step actions — opening requisitions, routing approvals, and activating supplier processes — without continuous human guidance. Unlike RPA, which works according to a fixed script and breaks when input deviates from expected, the agent solves the problems it encounters.

Why This Is the Big Story of 2026

Supply chain management software with agentic capabilities will grow from less than $2 billion in 2025 to $53 billion by 2030, an annual growth rate of 93.5%. The most advanced autonomous agents alone are expected to reach approximately $16 billion, about 20% of the total software market. By 2030, 60% of organizations using supply chain software will adopt agentic capabilities, compared to 5% in 2025; by 2028, 40% of procurement teams will implement at least one agent; and by 2027, 70% of software vendors will include agentic capabilities, compared to only 1% in 2024.

What Autonomous Agents Actually Do

An agent can continuously monitor supplier performance data and market price fluctuations, and renegotiate spot rates within approved limits, or reroute shipments when service failure is expected. Startups like Lio, which raised $30 million from a16z in March 2026, are building agents that analyze documents, evaluate supplier capabilities, run compliance checks, negotiate, and complete transactions. Oracle launched AI agents within Fusion in February 2026, alongside Ivalua and SAP Ariba. At the frontier is bot-to-bot commerce: a buyer agent negotiates with a supplier agent. Gartner forecasts that 90% of B2B procurement will be managed by AI agents within three years, at a volume of over $15 trillion.

Working by Procedures and Regulation Is Only Half the Picture

Human oversight shifts from executing every step to oversight and auditing. This gives rise to a new category of “guardian agents” that monitor other agents for compliance violations, illogical decisions, and boundary breaches, and Gartner estimates they will capture 10% to 15% of the market by 2030. A telling warning: Gartner forecasts over 2,000 lawsuits related to autonomous system safety failures by end of 2026.

The Gap Between Expectations and Reality

The phenomenon known as “agentwashing” manifests in a sharp gap: about 79% declare adoption, but only about 11% actually operate agents in real activity. More than 40% of agentic projects are expected to fail by 2027, because the legacy systems simply don’t support execution requirements. Most implementations are still very narrow in scope. The lesson: data, integration, and working by procedures are no less important than the agent itself.

What This Means for Israeli Organizations

The practical path is to start from the ‘tail’ (tail spend) and routine processes, with clean data and clear guardrails, and only then expand to strategic categories. In Mashik’s procurement and technology divisions, with experience in implementing RPA and automation and with the Comidor system, we are building the infrastructure that enables agents to work, rather than purchasing a promise that doesn’t materialize.

Read also: Procurement Experts – Mashik’s Procurement Division