95% of Supply Chain AI Initiatives Fail. Here's What the Other 5% Do Differently.
A practical guide to change management: navigating the human side of agentic AI adoption in logistics
We are living through a pivotal moment in supply chain.
Agentic AI is enabling limitless scale and operational impact, and organizations moving early are already pulling ahead. And yet, over 95% of supply chain AI initiatives fail to deliver sustained ROI. This isn’t because the technology fails, but because organizations underestimate what it takes to manage the change.
After 100+ AI implementations across the world's leading freight forwarders and customs brokers, including 6 of the top 10 globally - we've learned that the difference between a promising pilot and lasting transformation almost always comes down to people, process, and change management.
In this installment of our agentic AI in supply chain series, we draw on our experience deploying AI at leading logistics companies to cover the human barriers to adoption, why resistance runs deeper than most leaders expect, and how to ensure successful change management during your AI transformation.
Agentic AI: A Different Kind Of Change
Every generation of technology has changed how supply chains operate. The internet connected global trade partners. Cloud computing made real-time visibility possible. But agentic AI goes further - it doesn't just surface new insights. It acts on them, changing how decisions are made.
Unlike its predecessors, agentic AI plans, reasons, and executes tasks end-to-end. It takes autonomous action within defined guardrails - following the rules, SOPs, SLAs, and logic your business sets.
This isn't just an operational shift. It changes where human judgment gets applied, what your team is responsible for, and how work flows through your organization.
For operations teams that have been doing things the same way for decades - in an industry built on control, compliance, and manual oversight - that can be hard to adjust to. Even when the benefits are clear.
BEFORE Imagine James - a customs broker with 20 years of experience. He manually pulls data from commercial invoices, packing lists, and bills of lading. He looks up HS codes one by one, cross-referencing tariff schedules and compliance rules. He populates declaration forms field by field. And when information is missing, he chases shippers and waits for responses before he can progress. |
AFTER Today, an AI agent handles that process end-to-end. It ingests documents the moment they arrive, extracts the relevant data, and structures it automatically. It recommends HS classifications with confidence scores, flags compliance concerns, and drafts requests for missing information. When James opens his queue, the entry is prepared and ready for his review and sign-off. James's expertise is still required. But, he no longer has to spend time on data entry, drafting emails, and comparing documents - he's spending it on the exceptions, decisions, and customer relationships that only he can handle. |
Agentic AI is redefining roles. And that change - more than any technical challenge - is what organizations need to prepare for. This is not a pilot-phase experiment - it's the new operating model. Organizations that treat it as one risk falling behind those that don't. The shift is already underway. The question is whether your organization is prepared to move with it.
AI Resistance: The Fear Behind the Friction
Change is hard - even when the benefits are clear. For most teams, resistance to agentic AI starts with inertia and uncertainty. Processes that have worked for years, ways of working that feel second nature, and a lack of clarity about what "good" looks like with AI in the mix - these are enough to make even willing teams hesitate. For others, the concern runs deeper still - worrying that talk of AI driving efficiency and scale means their jobs are at risk.
Whether the resistance is emotional or practical, it can manifest as active pushback - and that can derail even the most promising deployment.
Resistance can take many forms - ignoring guidelines, opting out of training, refusing to use AI tools, and in some cases tampering with performance metrics to make agentic AI appear less effective than it is. And it's not just a frontline issue.
Increasing capacity with agentic AI isn't about reducing headcount - it addresses a challenge supply chain leaders know well: the pressure to hire when demand surges and cut when it drops. It's about building an operation that can handle more, with the same talented team.
"Agentic AI is enabling Hellmann to do 2–3x more business with the same workforce."
Jens Drewes
CEO, Hellmann Worldwide Logistics - TPM26
Getting It Right: The Framework for Lasting Adoption
These are the key sticking points that come up time and again in our AI deployments - and how to overcome them.
Before You Start | |
Hidden gaps in buy-in and readiness create blockers | Assess readiness across every team before you deploy. Customs, finance, and operations are often at very different stages - and misalignment creates blockers that could have been avoided. Identify the gaps early. |
Overambitious pilots damage confidence before value is proven | Start with a single, high-value workflow - invoice matching, document extraction, or customs entry prep. Prove the value there first. Then scale. |
The wrong first users can derail early momentum | Don't default to your most experienced operators - they're often the most invested in existing ways of working. Start with the team members most curious about AI. They'll move faster and build a stronger foundation. |
Change Management Tip Some leading forwarders choose to rebrand their Raft solution internally - giving it a name that feels native to their business and culture. It shifts the perception from "something a vendor installed" to "our platform." When people feel ownership over something, they are more likely to champion it and want it to succeed. A small change that can turn reluctant compliance into enthusiastic adoption. |
During Deployment | |
Trust - Teams won't hand off declaration prep or invoice approval to an agent they can't verify or explain. | Ensure your solution logs every agent action end-to-end. Give teams full visibility into what data was used and why a decision was made. Make review and override straightforward - nothing should happen in a black box. |
Role Redefinition - The shift from processing entries and invoices to supervising AI agents is hard without clear boundaries. | Appoint a strong internal champion from day one - someone who understands the operations, the technology, and the value AI can bring. Clearly communicate what agents are taking on, what stays with the team, and how day-to-day roles are changing. |
Scaling Beyond the Pilot - Getting one team or trade lane working is one thing. Scaling organization-wide is another. 23% of supply chain AI projects stalled in 2025 due to lack of cross-functional alignment. (Gartner) | Go live fast and deliver value quickly - nothing converts a skeptic faster than results. Redefine success metrics - move from manual throughput to decision velocity, exception resolution rates, and clearance times. Align operations, customs, finance, and leadership on the same vision before scale begins. |
The key is to focus on one principle: don't move faster than your organization's trust allows. When you are ready to deploy your agentic AI solution, the way to do that is to expand autonomy in stages - earning trust at each step before moving to the next.
Stage 1: Recommend | Stage 2: Approve | Stage 3: Execute | |
AI Role | Surfaces recommendations | Handles routine cases, flags exceptions | Executes end-to-end within defined parameters |
Human Role | Make every decision | Review and sign off on exceptions | Supervise and handle edge cases |
In practice | AI flags potential HS code mismatches on an import declaration for human confirmation | AI matches AP invoices against buy rates and accruals, auto-approving within tolerance - human reviews only the invoices flagged outside threshold | AI processes and approves invoices end-to-end within defined parameters, human monitors exception reports and audits outcomes |
Moving too fast to Stage 3 destroys trust. Moving too slowly kills ROI. The best approach is to let confidence - not a timeline - determine the pace.
From Execution to Supervised Autonomy
The supply chain organizations getting AI adoption right treat it as a transformation, not a tool. And they choose the right partner - one with a product built for the industry and the expertise to drive change.
At Raft, change management is built into our platform and how we partner with every customer. We work as an extension of your team from day one - delivering value fast, training your teams, agreeing a clear roadmap from narrow start to full-scale rollout, and maintaining shared accountability at every stage. Every deployment is configured to your SOPs, rules, workflows, and preferences - down to an operator level. The platform adapts to how your business works, not the other way around.
With over $150M in measurable savings delivered for freight forwarders and customs brokers worldwide, we know what good looks like - and we'll help you get there. Book a call to learn how Raft can help you deploy and adopt agentic AI.
