Many of the recent headlines that cast doubt on the future of AI were courtesy of a 2025 MIT study that found 95% of GenAI pilots fail to generate a positive financial return.
Based on what I’ve seen working with clients over the last 12 to 18 months, this sounds about right.
Many IT and supply chain leaders are quick to buy into the excitement and possibilities of AI—and rightfully so. But their companies often struggle to move beyond the hype and convert pilots into full-scale programs—something that can be measured and improved upon.
If you’re curious like me, then you must ask yourself: Why? Where is the disconnect between the promise and the payoff? And what can supply chain leaders do to close this gap?
Because let’s face it: Supply chain AI isn’t going anywhere. The market for AI-based supply chain applications is estimated to grow from $14.5 Billion in 2025 to over $50 Billion in the next five years. AI availability is not the issue. It’s whether your technology ecosystem enables or hinders AI-enabled innovation.
Getting supply chains ready for AI
In supply chain management, where demand spikes, tariffs, geopolitics and supplier glitches are everyday headaches, the last thing you need are technology flops that waste serious money and precious bandwidth.
Now more than ever before, supply chain leaders are realizing that if you can get the basics right—from sourcing and supply-demand balancing to warehouse optimization and last-mile delivery—you’ll be ready to flex when volatility hits.
This same fundamentals-first mentality must be used as you embark on an AI journey. In my experience, there are five technology essentials that companies need to get right before AI delivers value.
1. Cloud-based platforms and accessible data
To make GenAI hum using large language models (LLMs), you need data. Lots of it. And you need the ability to combine data from multiple sources and manage it effectively. Realistically, this means your data needs to be in a cloud environment that can be rapidly and securely accessed by many groups and able to integrate with various applications.
You should capture historical plan data (i.e., business plans, operational plans, and execution plans) alongside past performance and actuals, but also real-time data that sits within your supply chain systems.
In fact, exception data is some of the most valuable information you have to uncover root causes, bottlenecks, and minor disruptions. Instead of truncating and discarding it like many of today’s solutions, keep it. The economics of data storage and processing have made “data hoarding” a viable strategy, so there’s no excuse not to grab and hold on to everything.
Ideally, you are creating a solid data foundation made up of every supply chain decision you’ve ever made – past and present – to inform your next decision. While a human couldn’t possibly contemplate all this information, this is the type of task that LLMs excel at. This extra context is what will enable your future AI-based systems to identify issues, recommend solutions and self-heal. Models will learn from exceptions, sharpen over time and become more autonomous.
2. Composable systems and applications
You don’t have to rip out your entire technology stack to harness AI in your supply chain. A composable approach is much more effective to achieve higher solution utility and rapid ROI. Use your AI transformation initiative to rethink how your systems connect and interact.
If you are like most of my clients, then you will have a few legacy enterprise systems that don’t play well with others. Let AI agents bridge those gaps. AI agents effortlessly tie together legacy applications, AI-powered add-ons, advanced analytics and reporting tools to give you end-to-end supply chain visibility.
A modular, composable style delivers value quickly. A2A, MCP and other enterprise-compliant workflows enable you to layer new capabilities on top of what you already own, avoid massive overhauls, eliminate data and process silos, and roll out improvements incrementally. Plus, you will be able to leverage and incorporate workflows provided by third party agents such as those developed by the vendors of the legacy enterprise software applications.
3. Modern code and development tools
Yes, AI agents can be used as a workaround to connect legacy systems. Additionally, you can leverage your AI journey as the impetus to modernize these systems. Simplify customizations. Eliminate decades of technical debt and reduce operational overhead. This could be through a migration to a new platform or just an upgrade of an existing enterprise platform with less technical debt.
Not only do aged legacy enterprise systems require a significant amount of support, but they erode user confidence, decrease business agility as a result of poor system performance, and the total cost of ownership gets bigger as they get older.
Modernization tools like Sapient Slingshot are a good place to start. These SDLC accelerators automate the full cycle—from legacy code interrogation and solution upgrades to code development, remediation, regression testing, and rollouts.
Modern code creates a flexible technical ecosystem that makes data accessible for analysis, boosts output speeds, improves data quality, and accelerates decision-making timelines. Plus, it increases rigor when it comes to actively managing data accuracy, contextual integrity and security.
4. Well-defined supply chain technology strategy and KPIs
Taking a step back to look at the bigger picture, one of the biggest indicators of whether your GenAI pilot will fail or succeed is if it enables your overall business and supply chain strategies. Or is it a one-off experiment untethered to business outcomes?
You’ve got to stay grounded in your strategic direction and North Star. Only then can you determine how AI supports your broader technology strategy and, more importantly, your Target Operating Model. Without a realistic, strategic supply chain capabilities roadmap and measurable KPIs, you're just tinkering.
(Related: My colleague Brad Murach laid out a helpful three-part framework in his recent blog, Roadmap Reality Check: How to choose the right supply chain technology.)
5. Change management and change-ready mindset
Digital transformations should not begin or end with technology. Especially not AI. The pace of change is no doubt speeding up. However, what many large enterprises are struggling to do is keep up their pace of change management.
As with any new tool, you will need guardrails, SOPs and even a task force or Center of Excellence to ensure adoption and proficiency. Because if I’ve learned anything in my 30+ years as a technology strategist, lack of adoption kills more projects than any half-baked strategy or slow deployment.
For AI, which is burdened with real and perceived threats to job security, consider adopting a "human-led, AI-enabled" approach. Elevate your workers and their work, don't just automate them away. Foster end-to-end, systems thinking and ownership of high-value activities. Train, retrain and upskill as needed.
Unlocking AI's true potential
Whether you’re chasing AI’s promise of enhanced demand sensing, dynamic inventory analysis, autonomous supply chain planning, or even predictive equipment maintenance within warehouses and distribution centers, if you can get these five technology essentials right, your GenAI pilots will be built to deliver.
Your organization won't just survive the AI era—it will lead it by turning data into a true competitive advantage. The time for supply chain AI readiness is now. Spinnaker SCA is here to help.
