Project44 unveils fleet of AI agents at customer event Decision44

Project44’s 2014 founding was an attempt to fix that. The company spent a decade and $1.5 billion in R&D to build what McCandless called the connective tissue the global supply chain lacked: “We turned the lights on.” The result was visibility at unprecedented scale and a logistics data graph that captured more than 1 billion customer-created events per day and 4 petabytes of data monthly, the equivalent of watching YouTube 24/7 for 9,000 years.

But visibility brought chaos and analysis paralysis. The signal-to-action ratio was just 2 percent. The industry became informed, yet frozen. “What if this is as far as we go?” McCandless recalled asking himself. “What if we gave the world a clear view of its problems, but we can’t solve them fast enough?”

The answer, unveiled during Decision44, is a new class of AI agents that collapse the three-step lead time—truth, decision, action—into a single motion. A process that once took days now takes seconds. “Agency is a dial you control,” McCandless said. The technology is no longer the bottleneck. The last remaining constraint is the processes built around it: every approval chain, sign-off and “let me see.” The hardest part of what comes next “isn’t technology… it’s new processes.”

Building trusted AI for the supply chain

Chief Strategy Officer and Chief Operating Officer Jonathan Scherr picked up the thread, declaring the company’s mission unchanged—“to eliminate friction from global trade”—but now executed through AI that enterprises can actually trust. McKinsey estimates AI’s enterprise impact at $8 trillion annually, yet only 1 percent of global enterprises consider their AI mature. In supply chains, where the cost of failure is measured in lost cargo, stockouts or detained trucks, trust is the gap.

Scherr broke trust into three pillars: Data, Intelligence (logic and reasoning), and Action. Large language models like Claude, GPT or Gemini are powerful at collating public data, but insufficient alone for business use cases. “The data that sits on top of the LLM… is what makes it valuable,” he said. Project44 calls this “context”: the situational knowledge of relationships between pieces of information that lives mostly outside any single organization.

More than 80 percent of AI implementations fail because of insufficient context, Scherr noted. Project44’s decade-long investment in its data graph, interoperability with WMS, TMS, YMS, ERP and OMS systems, and AI agents that proactively reach out to carriers (via the recent LunaPath acquisition) have already delivered measurable gains: more than 25 percent improvement in ETA accuracy.

The company is now embedding “context-fueled intelligence” directly into the platform. Supply chain has its own semantics—“on-time” means something different to every organization—and the system must understand pattern recognition and reasoning. Scherr introduced “Mo,” a context-aware chatbot coming in early July that lets customers upload their own shipment history, business rules and semantic layer for true analysis grounded in their data.

From Autopilot to specific agents: p44’s product roadmap

Nick Ruggiero, director of product management for Autopilot, detailed how Project44 is turning that intelligence into automated, trusted workflows. The agents are configurable, explainable and human-controlled, with granular controls, transparent logic, audit history and intervention points at every critical step. Stored workflows function like a prompt library.

Early results: 17 percent reduction in manual exception handling. Configurable workflows are available today; agent transcripts arrive in May and multi-agent workflows are set to be released in July. Project44 itself is accelerating its pace of production, from shipping one workflow per week to one per day over the summer.

Ruggiero distinguished deterministic workflows (“if X then Y”) from agentic ones that decide and adapt. The architecture decomposes work into tasks, combines semantics and triggers, builds focused micro-agents, tests them in human-controlled Autopilot flows, and introduces supervisory agents capable of orchestrating complex sequences.

At this point in the presentation, numerous product leaders demoed specific agents in rapid succession:

• Ilias Pagonis, senior staff product manager for Transportation Planning, attacked the legacy of periodic bid cycles, manual spreadsheet negotiations and slow sequential workflows. Project44’s Freight Procurement Agent replaces them with continuous AI-powered sourcing and simultaneous carrier negotiations in seconds under governed autonomy. Significant results have already demonstrated: 4.1 percent freight cost mitigation, 75 percent reduction in sourcing cycle time, and 70 percent reduction in manual coordination. The Freight Procurement Agent is available now.

• Ellie Crist, VP of product management, and Lauren Fitzpatrick, director of data science, tackled logistics operations. Cargo theft losses are up 60 percent year-over-year. New agents automatically detect risk by collating signals like door-open events and GPS spoofing patterns, and cut intervention times from 40 minutes to 12 minutes. Ceva Logistics, a project44 customer, specifically cited p44’s AI capabilities as enabling them to intervene in a shipment “before an incident becomes a loss.”

Inventory management faces late stockout detection and transit variability that inflates safety stock. The Stockout Risk Monitoring and Inventory Risk Agent, also July, closes the gap so the world’s largest automakers can see risk while there is still time to act, which is critical when downtime costs $27,000 per minute in automotive plants.

Last-mile operations can see disruptions, like large weather events, that cascade across thousands of shipments. Project44 already offers predictive ETAs and consumer visibility. New capabilities move from reactive exception management to agentic disruption response: Last Mile Disruptions in June and Agentic Case Resolution in July.

• Aron Kestenbaum, VP of product for Yard Management, highlighted the hidden costs after a truck arrives on time. Yards remain manual, siloed and disconnected, with congestion, detention and labor waste in the gap between arrival, unloading and “in stock.” Integrated scheduling, gate optimization, visibility and task orchestration are the solution. The Dynamic Slot Booking Agent arrives in June and is expected to reduce detention time by 17 percent, improve yard optimization by 8 percent and save 10 hours per week on manual coordination.

Abercrombie & Fitch’s Kristen Kravitz on process and people

The most grounded moment came during a conversation between Kristen Kravitz, Group Vice President of Supply Chain at Abercrombie & Fitch, and Project44 Chief Revenue Officer Rick Turco. Abercrombie has transformed from a standard mall retailer into a lifestyle brand, forcing its supply chain to evolve from cost center to value creator and brand enabler.

Kravitz, now in her tenth year, described rebuilding the team to include operations, transportation technology and supply chain digital solutions. “We’ve tried to embrace a digital-first mindset,” she said. The biggest obstacles were manual workflows, siloed data, third-party systems and predictions built for a stable world that no longer exists. “We’re in the never-normal supply chain now.”

At Abercrombie, inventory discipline is considered paramount; Kravitz says that she considers the trendy, seasonal apparel the company makes to be ‘perishable’ for all intents and purposes. Technology is no longer the bottleneck, Kravitz emphasized: “It’s often now the process and people.” The company shifted teams from transportation mode-focused work to origin-and-destination decisioning and built hybrid data structures with central and embedded supply chain analytics. Cross-functional translation, including translating the language of supply chain operations into financial and planning language, has been key.

Project44 data, combined with Power BI rollout, has been instrumental. “Visibility is really an enabler of more efficient execution,” Turco noted. Kravitz stressed balancing movement data with other systems and the need for end-to-end predictions that incorporate factory production milestones. “By the time a vendor hands over the product… a lot of what is going to happen has already been predetermined,” she said. 

Ilias Pagonis on the Intelligent TMS

Pagonis returned to close the tech demo segment by explaining why Project44 is entering the TMS space. Traditional TMS platforms were designed for stable carrier relationships, predictable freight cycles and annual procurement—conditions he believes no longer exist. Many solutions remain on-premise and unchanged for 20 years.

Project44’s Intelligent TMS is “intelligent and unbiased,” acting on vetted carrier performance and market data rather than self-reported snapshots. It is AI-native rather than AI-bolted-on, modular rather than monolithic, and powers the entire shipment journey: procurement, planning, execution, visibility, settlement and payment.

Pagonis highlighted the AI value gap: 96 percent of transportation leaders say continued AI investment is a top priority, yet only 13 percent of logistics professionals report measurable value today. “Vanity AI projects won’t deliver value nor survive the next budget cycle.”

He demoed the Execution Recovery Agent, which instantly identifies the next carrier and contacts them. Additional agents launching later this year include Freight Procurement, Load Consolidation, Appointment Scheduling and Freight Audit.

“The only modular TMS on the market,” Pagonis declared. One control layer provides infinite flexibility. Customers can start where value is highest and scale on their terms—no multi-year migration required.

The bottom line

Decision44 made clear that Project44 is no longer content with visibility. It is handing execution to AI agents while keeping humans in the governance loop: “Let the agents handle the grind; your job is to govern the dials.” From freight procurement to cargo theft prevention, yard slotting to last-mile disruption response, the company is releasing purpose-built agents on an aggressive timeline: some live now, most by July.

For an industry still recovering from pandemic shocks, facing 60 percent higher theft losses, volatile freight markets and never-normal conditions, the promise is systems that finally move at the speed of thought. Whether the processes and people can keep up (as Kristen Kravitz candidly noted) will determine how much of that $8 trillion AI opportunity actually lands in supply chain P&Ls.

Jett McCandless is betting the next 12,000 years of logistics history begins today, with agents that don’t just see problems but act on them quickly, transparently, and under human governance. The lights are on. Now the machines are moving.