AI Supply Chain Leaders: What to Watch at ARC 2026

The busiest supply chain teams I talk to aren’t short on AI pilots. They’re short on results that survive contact with real operations: messy master data, uncertain demand, carrier volatility, labor constraints, and sustainability reporting that keeps getting stricter.

That’s why events like the 30th Annual ARC Industry Leadership Forum (Feb 9–12, 2026, Orlando) matter for anyone in transportation, logistics, manufacturing, or procurement. Not because you’ll see flashy demos—because you’ll hear what actually works when AI has to run inside plants, warehouses, fleets, and planning towers.

This post is part of our “AI in Supply Chain & Procurement” series, where we focus on practical ways AI improves forecasting, supplier management, risk reduction, and end-to-end optimization. Here’s what I’d pay attention to at ARC 2026—and how to turn the ideas into a plan your team can execute.

Why AI in supply chain is finally getting operational

AI adoption in supply chain used to stall at the same point: a promising model built in isolation that couldn’t integrate with planning, execution, or shop-floor constraints. That’s changing. The ARC forum’s emphasis on industrial operations + supply chain is the right framing because the best AI outcomes happen when planning and execution stop acting like separate worlds.

Three shifts are making AI more “operational” in 2026:

• Better connectivity across systems (ERP, WMS, TMS, MES, EAM) so models aren’t starving for context.
• Decision automation is getting narrower and safer: more “automate this specific dispatch decision” and less “let the model run the network.”
• Digital twins and simulation are being used as guardrails, not science projects.

A sentence worth stealing for your internal alignment: AI only creates value in supply chain when it changes a decision, and that decision survives the exception path.

The ARC themes that map directly to logistics and procurement value

The RSS announcement highlights AI, cognitive analytics, digital twins, and predictive technologies—plus energy transition and resilience. That can sound broad, but it maps cleanly to the decisions supply chain leaders are judged on.

AI + predictive analytics: from “forecast accuracy” to “service and cost control”

Predictive analytics is no longer just about demand forecasting. The highest-ROI applications I’ve seen focus on controlling service and cost under volatility:

• ETA prediction and exception triage that reduces expedite spend
• Predictive maintenance that prevents unplanned downtime in DC automation or yard equipment
• Order risk scoring (late risk, short risk, damage risk) to prioritize interventions

If you’re attending ARC, listen for specifics like:

• What horizon the model predicts (2 hours? 2 days? 2 weeks?)
• What actions happen when risk is detected (reroute, re-promise, split ship, reallocate)
• Whether humans can override—and whether overrides are measured

A practical benchmark: if an AI initiative can’t point to a measurable decision latency reduction (minutes/hours saved), it’s probably not ready for scale.

Digital twins: the safest way to test AI before it hits operations

Digital twins get marketed as “virtual replicas,” but their real supply chain job is simpler: simulate decisions before they create expensive surprises.

In transportation and warehousing, digital twins help you answer questions like:

• What happens to on-time delivery if we tighten appointment windows?
• If we change pick methodology, where does congestion move?
• If we re-source a component, what breaks in inbound lead times?

For procurement teams, a digital twin mindset also applies to supplier networks:

• If Supplier A fails, what’s the service impact by SKU and customer segment?
• What inventory buffers minimize risk without bloating working capital?

At the forum, I’d look for talks that connect twins to closed-loop improvement: simulate → deploy → measure → update assumptions. If it’s only a one-time model, it won’t keep up.

Energy transition: logistics can’t hit targets without better decisions

Sustainability is no longer a brand-only metric. In 2026, it’s increasingly tied to customer requirements, carrier agreements, and reporting.

AI can support decarbonization, but only in grounded ways:

• Load building and mode optimization (fewer partials, less air)
• Network design scenarios that trade off cost, service, and emissions
• Carrier selection that considers reliability and emissions factors

My take: the fastest sustainability wins in logistics come from better planning discipline, not new hardware. AI helps by finding the “quiet waste” humans miss—recurring expedites, chronic dwell, avoidable rework.

The use cases you should pressure-test (and how to know they’re real)

Conference conversations can blur into buzzwords. Here are five AI in supply chain use cases that are mature enough to demand specifics—and the questions that separate substance from slides.

1) Demand sensing and inventory optimization

What it does: Uses near-real-time signals (orders, POS, promotions, weather-like disruptors, lead time changes) to adjust forecasts and inventory targets.

Ask:

• Did you reduce stockouts and inventory, or just shift the problem?
• How do you handle sparse data and new-item launches?
• What’s the cadence—daily, weekly, intraday?

2) Transportation optimization with dynamic constraints

What it does: Improves routing, tendering, appointment planning, and exception management using AI predictions.

Ask:

• What percent of loads run through the optimized flow?
• How are constraints managed (hours-of-service, docks, driver preferences, detention risk)?
• How do you prevent “optimization whiplash” where plans change too often?

3) Warehouse labor planning and slotting

What it does: Predicts workload, assigns labor, recommends slotting changes, and reduces travel/congestion.

Ask:

• Did travel time drop? By how much?
• How do you measure adoption (are supervisors following recommendations)?
• Do you optimize for throughput, safety, or both?

4) Supplier risk and procurement intelligence

What it does: Flags supplier risk (financial, geopolitical, performance), recommends alternates, and supports negotiations with better should-cost inputs.

Ask:

• How do you avoid false positives that burn relationships?
• Are risk scores tied to concrete policies (dual-source triggers, safety stock rules)?
• Can you trace recommendations to data, or is it a black box?

5) Predictive maintenance in industrial operations

What it does: Predicts failures in assets that directly affect supply chain flow (conveyors, sorters, robotics, plant equipment) and schedules maintenance to minimize disruption.

Ask:

• What’s the reduction in unplanned downtime?
• How do you handle sensor gaps and inconsistent maintenance logs?
• Do planners get informed early enough to re-sequence production or reprioritize orders?

If you take one filter into ARC: only believe AI that ships with a workflow. Models don’t create value—workflows do.

A practical “AI readiness” checklist to bring back from Orlando

If your goal is lead generation or internal buy-in, the temptation is to come back with a stack of vendor decks. Don’t. Come back with a short list of decisions you want to improve.

Here’s a checklist I’ve found works well for supply chain and procurement leaders evaluating AI initiatives after an event.

Define the decision, not the technology

Write a one-liner your operations team agrees with:

• “We will reduce late deliveries by intervening on high-risk loads 12–24 hours earlier.”
• “We will cut expedite spend by preventing schedule instability in the last 72 hours.”
• “We will reduce line stoppages by predicting failures 7–14 days in advance.”

If you can’t write that sentence, you’re not ready to buy anything.

Demand measurable targets (with baselines)

Good AI programs start with a baseline and a target. Examples:

• Detention hours per load
• Expedite spend per unit shipped
• Forecast bias by product family
• OTIF by customer segment
• Mean time between failure (MTBF) for critical assets

Plan for data work (because it’s always more than you expect)

Most companies underestimate data cleaning and integration. Budget for:

• Master data fixes (locations, units of measure, SKU hierarchies)
• Event standardization (what counts as “arrived,” “loaded,” “departed”)
• Feedback loops (capture overrides, exceptions, outcomes)

A blunt truth: your AI model’s ceiling is your data quality floor.

Build governance for AI decisions

The forum’s focus on resilience and sustainability implicitly demands governance:

• Who owns the model’s outcomes (Ops? IT? CoE?)
• Who approves policy changes when the model shifts behavior?
• How do you monitor drift and retrain cadence?

If you’re deploying generative AI for procurement or operations support, add controls for:

• Approved data sources
• Audit logs
• Role-based access
• Human approval for high-impact actions

What to do before and after ARC 2026 (so it turns into pipeline)

Events create momentum. Momentum disappears fast when you return to a backlog of fires.

Before the forum

• Pick two processes you want to improve in 2026 (example: late-load prevention + inventory buffers).
• Agree on three KPIs that define success.
• List your top constraints (systems, data, change management, labor rules).

During the forum

• Prioritize sessions and conversations that include numbers, timelines, and adoption lessons.
• Ask peers what broke during rollout. That’s where the learning is.
• Collect “architecture stories” (how they connected planning to execution, how they handled exceptions).

After the forum

• Run a 30-day feasibility sprint: data audit, workflow mapping, KPI baselining.
• Select one use case for a 90-day pilot with a production workflow.
• Decide upfront what “scale” means (sites, lanes, suppliers) and what has to be true to get there.

If your organization is serious about AI in supply chain management, the ARC Industry Leadership Forum is a smart place to compare notes with leaders who are already doing the hard part: making AI behave in real operations.

The question to carry into 2026 is simple: Which supply chain decisions are you willing to standardize enough that AI can improve them—and which ones will stay artisanal and slow?