Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027. Most manufacturers haven't seen that number because vendors aren't sharing it.But for manufacturers whose margins hinge on precision AI is a tool with specific value when applied correctly. Commerce analyst Heather Hershey put it plainly on OroCommerce’s recent AI Hype Detox for Manufacturers and Distributors podcast: "AI is a big, nebulous umbrella that can mean essentially everything and nothing all at once." That vagueness costs money.
Modern AI systems excel at generating language and identifying patterns. They struggle with layered rules, permissions, and context. B2B commerce runs on all three.
Some organizations that aggressively deployed AI agents across customer and internal workflows have found that the models often fail when confronted with complex operational rules or highly contextual decision making. In these environments, the system may produce confident answers that are incomplete, inconsistent, or simply incorrect.
This does not mean AI lacks value. The technology is powerful, but it performs best when paired with strong data foundations, clearly defined processes, and human oversight.
Commerce analyst Heather Hershey offers a useful test: "Would it actually make sense for someone to engage with a chatbot instead of the UI? If you can't answer that definitively, an LLM might be overkill."
For manufacturers and distributors, ignoring the technology entirely would mean missing an important shift in how digital systems evolve. At the same time, blindly trusting it to replace complex human judgment introduces unnecessary risk.
The organizations that succeed with AI will likely take a balanced approach: experiment with it, learn from it, and integrate it where it genuinely improves workflows.
After hype has passed, it is time to determine how to experiment with AI so that it will deliver real business results.
Integrating AI into workflows where workers are supported rather than replaced allows manufacturers to find greater value from AI and provides a safe environment for companies to learn via experimentation with AI capabilities while still ensuring that their operations are reliable.
In the case of digital commerce, utilizing AI in environments such as OroCommerce generally means creating solutions to explore ways that AI can help speed up the pace at which work is accomplished. AI can be used as a tool to assist in finding new insights through analysis of large, complicated data sets, help employees review product and pricing data quickly and efficiently and automate repetitive administrative tasks while enabling quicker, better decision making for employees.
“Our approach to AI is less Sci-Fi, more ROI. Our job as a technology vendor is not to promise futuristic possibilities, but to deliver real value to customers quickly. That means focusing on practical AI use cases, grounded in proprietary data, safely contained by native guardrails, and easy to deploy so teams can see results immediately,” says Jary Carter, CRO and Cofounder of OroCommerce.
In other words, the most productive way to engage with AI today is by identifying low-risk, high-ROI processes and experimenting there first. Let AI improve efficiency in controlled areas while your governance, oversight, and business rules remain firmly in charge.
This approach allows manufacturers to learn from AI, refine how it fits into their workflows, and gradually unlock its value without introducing unnecessary operational risk.
For many manufacturers, the biggest barrier to effective AI adoption is not the technology itself, but the operational foundations underneath it. Before introducing AI into workflows, organizations must ensure that the core systems governing their business are reliable, structured, and clearly governed.
Three elements tend to determine whether AI initiatives succeed or fail.
AI systems can only work with the information they are given. AI cannot help when data like product price, availability, or terms appear in multiple ways across ERPs, CRM systems, or even spreadsheets, if critical data is not all in one place.
Here's a basic example. If a customer asks for the contract price on a specific item and there is not a single source of truth to pull from within seconds (because the employees need to look across multiple systems), then AI will not fix the problem. In fact, it may amplify the confusion. Establishing a single, reliable source of truth for operational data is therefore the first prerequisite for meaningful AI use.
Data alone is not enough. Organizations also need clear rules governing how that data is surfaced to different users.
In B2B commerce environments, those rules can include customer-specific pricing, contract terms, distributor access permissions, and internal approval workflows. When these rules are fragmented or inconsistently applied, AI systems can behave unpredictably. The more conditional logic an AI system must interpret without a clear structure, the more likely it is to produce inconsistent results.
Establishing centralized commerce logic ensures that the right information is delivered to the right customer or employee consistently.
Finally, organizations need clear ownership of processes and systems.
Each operational workflow should have defined human and system owners, along with governance mechanisms that allow teams to track changes and audit decisions. Without this structure, AI can introduce additional uncertainty into already complex environments.
Strong governance allows companies to experiment with AI while maintaining control. Guardrails can be applied within core systems, changes can be tracked, and organizations can ensure that AI supports business operations rather than undermining them.
When these foundations are in place, AI becomes far more practical to explore. Without them, even the most advanced AI tools will struggle to produce reliable outcomes.
For manufacturing leaders evaluating AI, the most effective approach starts with operational questions, not technology selection.
AI investments tend to produce the strongest results when they address clear friction within existing workflows. Before selecting tools or platforms, identify where that friction actually exists.
Where is friction slowing down processes?
Many manufacturers still rely on manual steps in processes that could be streamlined. Tasks such as manual order entry, repetitive data reconciliation, or routine administrative work often create unnecessary delays across the organization.
Where are teams losing time on low-value tasks?
When employees spend significant time gathering information, generating reports, or responding to predictable internal questions, AI can often assist by automating data retrieval, generating summaries, or surfacing relevant insights faster.
Where does complexity create high error rates?
Operational complexity in pricing, product configurations, and order management can introduce costly mistakes when handled manually. AI can support these workflows by identifying anomalies, validating data inputs, or assisting teams in navigating complex product and pricing structures.
Where does scale exceed what humans can reasonably handle?
High-volume environments, such as large order processing pipelines or customer inquiries, can overwhelm teams during peak demand periods. In these scenarios, AI tools can help prioritize outreach, assist with guided selling, or support self-service capabilities that reduce operational bottlenecks.
Where can stronger relationships improve customer outcomes?
In B2B manufacturing, long-term customer relationships matter more than individual transactions. AI can help surface proactive insights, support sales teams with better customer intelligence, and enhance digital self-service experiences that make it easier for customers to do business.
When organizations begin with these questions, the most practical AI opportunities often emerge naturally. In many cases, the highest-value applications are not the most experimental ones, but rather the solutions that remove friction from everyday processes and improve how teams and customers interact with the business.
Artificial intelligence will matter for manufacturers, but only when approached with discipline. The technology’s promise is not in replacing people, but in amplifying human capability across data-intensive tasks, customer interactions and knowledge work.
“The most important question is not what the technology can do, but what business problem it is solving,” says Carter. “At OroCommerce, we view AI as an enabler of smarter commerce operations and a better customer experience. It should strengthen decision-making, streamline complex workflows, and remove purchase complexity for customers.”
The businesses that succeed will interrogate vendor claims, prioritize data readiness, and deploy AI where it supports strategic objectives.
AI is not a distant concept for manufacturing, it is an evolving set of capabilities that must be grounded in the realities of enterprise operations. When applied thoughtfully, it can unlock new efficiencies and deeper insights; when applied blindly, it risks distraction and wasted investment. The choice for manufacturers is not whether to engage with AI, but how to do so in a way that aligns with the realities of their business.