Descartes and agentic AI: from orchestrated demos to accountable logistics execution

## From logistics applications to coordinated execution **Author: Adrian Gonzalez**, founder and president of Adelante SCM and founder of Talking Logistics. In his October 26, 2026 article, "Descartes: From the Dot-Com Era to the AI Era," Gonzalez reflects on the evolution of Descartes after attending its Innovation Forum. His perspective draws on 27 years of following the company as a supply chain industry analyst. According to Gonzalez, this was Descartes' largest Innovation Forum, attracting more than 1,000 attendees. The opening keynote featured CEO Ed Ryan, President and COO Scott Pagan, and EVP of Product Management Ken Wood. The central technology question is how Descartes' increasingly broad portfolio can work together across applications, workflows and trading partners rather than remaining a collection of individual software capabilities. ### Agent Control Plane: coordinating agents across business boundaries The Innovation Forum showcased the Descartes Agent Control Plane, intended to coordinate AI agents working across different Descartes applications and trading partners. Gonzalez explains that the underlying Global Logistics Network supplies connectivity and information, while the Agent Control Plane is designed to coordinate tasks under customer-defined rules, policies and approval processes. During the keynote, Descartes demonstrated a scenario spanning 28 steps, 15 Descartes applications and four companies, following an online furniture order through final-mile delivery. The demonstration illustrated a possible architecture for distributed agentic execution. It also showed why autonomy is not necessarily an all-or-nothing choice: organizations can define which actions agents may execute and which require human approval, according to operational risk and business requirements. However, a controlled demonstration is not equivalent to reliable daily operation. Gonzalez emphasizes that agentic AI adoption remains early and that the decisive test will be performance under routine operational conditions, particularly when multiple applications and companies are involved and events do not follow the expected plan. He anticipates that actual customer case studies could replace demonstrations at the following year's forum; this is an expectation, not an established outcome. ### Routing at scale becomes Fleet Performance Management Gonzalez moderated the panel "Routing at Scale: Turning Complex Distribution into a Competitive Advantage," featuring executives from Williams Sonoma, HD Supply, White Cap and SRS Distribution. His central takeaway is the evolution from routing optimization toward Fleet Performance Management. Routing is not a one-time calculation producing an optimal plan. It is an iterative process in which organizations construct a plan based on operating realities, observe actual execution and use the differences between planned and realized performance to improve future decisions. This changes the unit of value from an individual optimized route to the continuing ability to learn from execution. The feedback loop must incorporate practical constraints, driver and fleet performance, service commitments and the causes of deviations from the original plan. ### Logistics service providers at a crossroads Gonzalez also moderated a CEO panel examining changes in logistics service-provider business models, the expanding role of AI and automation, and the need to differentiate through technology, expertise and customer service. He additionally participated in a panel on the tenth edition of the benchmark survey for brokers and forwarders. One observation was that trade-policy volatility makes international supply chains more difficult to manage, but added complexity can also create demand for brokerage and forwarding expertise. These observations concern business-model evolution and competitive positioning. They do not establish that a particular provider has already generated measurable benefits from AI. ### What the event reveals about decision architecture The Agent Control Plane and routing discussions point to a common issue: execution intelligence only creates value when recommendations can be translated into coordinated actions and subsequent outcomes can be observed. Cross-company agent workflows require clear rules for information exchange, authorization and escalation when expected events fail to occur. A carrier substitution, delivery exception or changed fulfillment commitment may be technically executable while still requiring agreement among parties with different service, cost and commercial interests. Likewise, fleet optimization is valuable when execution evidence is captured and used to revise the next plan, not merely when a routing engine calculates an attractive initial solution. ### Evidence and limitations This article is Adrian Gonzalez's analysis of a vendor event and panel discussions. It includes a Descartes product demonstration and practitioner observations, not independent verification of production outcomes. The operational resilience of the Agent Control Plane, its performance under exceptions and the scale of customer adoption remain to be demonstrated through real deployments. Original article date: October 26, 2026. Dataleo Radar publication date: October 27, 2026. Source: https://talkinglogistics.com/2026/10/26/descartes-from-the-dot-com-era-to-the-ai-era/