What Are Major Pitfalls to Using Artificial Intelligence in Supply Chains?
Too much noise. In global, multi-tier supply chains, critical supplier information is often unavailable or inconsistent, leading to misleading outputs or excessive “noise” from irrelevant risk flags. Focus on strengthening data foundations, embedding human-in-the-loop validation, and prioritizing practical integration and usability across internal teams and supplier networks.
–Marissa Licursi
Director
Grant Thornton Stax
Data distrust. If you don’t own and trust your data, AI just scales bad inputs into bad decisions—no matter how polished the dashboard looks. Stakeholders need strong data ownership, governance, and validation before layering on AI.
–Carson Joyner
Digital Strategy Lead
Gnosis Freight
Accountability drift. When something goes wrong, blaming “what the AI said” doesn’t resolve disruptions. AI works best as an assistive tool—not an authority—when its recommendations are validated, actions are supervised, and humans remain accountable for outcomes.
–Doug DeLuca
Product Marketing Manager
SAP Business Network
Ambiguity. Applying AI to processes that aren’t well defined or repeatable introduces ambiguity and hallucinations. Another risk is hype outpacing reality, creating confusion about what AI can do. Leaders should define the type of AI being used and ensure data is curated and structured.
–Jack McCrum
Director, Optimization and Analytics
Reveel
Unclear ownership. Many AI pilots succeed locally but stall during scale up. The barrier is rarely the technology. More often, it’s unclear ownership, undefined decision authority, or processes that were not designed to absorb AI-generated outputs. Automation and AI will deliver sustained value when they are scaled enterprise-wide and aligned to execution roles, decision rights, and performance accountability.
–Matt Derganc
Senior Director
SSA & Co.
Overestimation. A common pitfall is assuming AI understands how logistics really works. Software can optimize routes and forecasts, but it struggles with real-world issues like customs rules, packaging compliance, or unexpected inspections. Use AI as a support tool while experienced operators keep control of final decisions.
–George Wicks-Farr
Head of Operations
Pallet2Ship
Avoid the Overreliance Trap
Overreliance on AI is a big risk. Investing in automation without fixing underlying processes creates new failure points. Chains break down when AI ignores real-world conditions and frontline decisions. The goal isn’t to automate everything, it’s to enhance how decisions are made. Companies getting this right modernize workflows first, then embed AI to improve service, transparency, and trust.
–Zach Jecklin
Chief Information Officer
Echo Global Logistics
Relying too much on AI can become perilous because every decision directly affects sourcing, pricing, and tax outcomes. Without validation, errors can quickly scale into material, financial, and compliance issues. As tax authorities raise expectations around auditability, transparency and traceability are mandatory to ensure every AI-powered decision can be explained and defended.
–Chris Hall
Senior Tax Officer, Global Tax & Compliance
Vertex Inc.
Applicability. Its use in inventory management with Customs remains limited. Trust is a key concern—stakeholders can’t yet rely on IT to fully run FTZ operations or interface with U.S. CBP systems like ACE. Today, AI is best suited for administrative tasks. As the technology evolves, stakeholders should prioritize validation, oversight, and phased integration.
–Jeffrey Tafel
President
National Association of Foreign-Trade Zones (NAFTZ)
Rushed deployment and incomplete data. Shippers must understand where AI data comes from and whether it reflects today’s real-world operating conditions. AI won’t solve capacity or service challenges, but with strong data governance, it can improve predictions and real-time insight.
–Mika Majapuro
VP, Product Management
TransmetriQ
No competitive edge. Anyone can add generic AI tools to their tech stack. Differentiate by leveraging unique data, refining AI for specific needs, and aligning it with strategic goals from the ground up. Offer a solution that no one else has.
–Kevin McMaster
SVP, Customer Success & Operations
Flock Freight
Data gaps. Assuming AI can compensate for gaps in operational data is one pitfall. Many supply chains still depend on periodic scans or checkpoint tracking, leaving gaps in visibility. If AI is fed fragmented signals, it can produce misleading outputs at scale. First, improve how data is captured.
–Simon Ford
Founder & CEO
Blecon
Missing inputs. Algorithms may not fully account for temperature, chain-of-custody, or compliance, and AI decisions can be opaque. Overreliance on historical data may miss sudden spikes in urgent shipments. Stakeholders should use AI to support, validate recommendations, train staff, and maintain oversight for high-priority or sensitive shipments.
–Lorena Camargo
President
Customized Logistics and Delivery Association (CLDA)
Autopilot mode. If AI makes a mistake based on bad data, it can snowball before anyone notices. Stakeholders must keep humans in the loop to vet big decisions and ensure the “math” aligns with real-world common sense.
–Bradley Barry
Director & Partner, Supply Chain Services
St. Onge Company
Protect and Validate Your Data
Adoption is outpacing management. Two risks stand out. First, security: Sensitive data is moving across more systems, often without proper oversight. Start by auditing access and reviewing it regularly. Second, signal integrity: AI outputs depend on data quality and degrade over time. Monitor inputs and validate models.
–Scott Stonys
Head of Sales & Customer Success
Spotter AI
AI in supply chains is creating new routes for cascading breaches. Poorly governed tools risk leaking data or linking systems in unintended ways; firms need tighter supplier oversight, clearer permissions, and human sign-off for critical automated actions.
–Melissa Carmichael
Head of U.S. Cyber
Beazley
Lack of customer service. Freight forwarders shouldn’t assume that deploying AI tools can replace the value of customer service excellence. An experienced operator can read a customer’s urgency, stress, and expectations in a way no algorithm can. The real advantage lies in combining human insight with AI for route optimization, shipment visibility, and pricing options.
–Sean Yanok
VP Regional Development
Gebrüder Weiss
Fragility. Optimizing just for cost strips out buffers, creating fragility. Algorithms that human operators don’t understand lead to errors without accountability. Human intuition gets weaker over time like an unused muscle. Solutions: Train AI to value redundancy, not just efficiency. Demand tools and workflows where AI offers recommendations and a human makes the final call.
–Nick Rakovsky
CEO
DataDocks
Change management whiplash and job security fears, plus AI agents that negotiate without real transportation context, eroding trust. Fix it by leading with strategy: map where humans need the most lift, pilot with clear guardrails, and measure outcomes for all parties, not just speed.
–Carly Gunby
VP Revenue
Transfix
Wrong answers. AI amplifies everything—including inconsistent, outdated, or duplicated information from trading partners. The result: confident wrong answers at machine speed. Stakeholders should automate data validation at the point of ingestion, before feeding AI systems. Clean, governed data flows are the foundation for any AI initiative.
–Michael Bevilacqua
VP AI Product Management
Adeptia
Drivers feel watched, which erodes trust. At our company, we address this by focusing on real-time risk prevention rather than constant recording. When AI identifies risky patterns before a crash without invasive surveillance, it becomes a supportive tool for safety. This approach helps stakeholders build a culture of trust while protecting the bottom line.
–Dr. Stefan Heck
CEO
Nauto
AI in supply chains can amplify bad data, create opaque decisions, introduce bias, overfit to disruption patterns, and expand cyber and privacy risk. Stakeholders should use strong data governance, human oversight, model monitoring, scenario testing, security controls, and clear accountability for decisions and outcomes.
–Jason Kasper
Senior Director of Product Marketing
Aras
Many risk rules, registers, and triggers sit in the backend systems of regulatory authorities, that AI engines haven’t managed to scrape and learn from. Also, rules and regulations evolve continuously, and AI engines will need to re-learn and adapt while the supply chain engine keeps running 24×7.
–Siddharth Priyesh
Vice President and Head of Group Commercial
CrimsonLogic
Treating AI as a one-time fix is a key risk; autonomy must be earned through training, guardrails, oversight, and updates. Stakeholders should prioritize data governance, clear use cases, and human oversight; validating and refining recommendations to ensure outcomes align with business strategy.
–Rachelle Butler
Director of Strategy
JBF Consulting