AI is the buzzword across all industries right now, and manufacturing certainly is no exception. How will we use it to make things simpler and more efficient? Which parts of manufacturing will be most impacted? How much money should be invested into AI? Recently, it seems no one has wanted to talk about anything else.

In fact, 80% of manufacturing executives plan to invest 20% or more of their improvement budgets in smart manufacturing initiatives, according to a Deloitte survey.

See also: Has physical AI gone ‘mainstream’? One new survey says yes

And AI is bringing big changes to the industry. But when will AI reveal transformative operational models of the future? Most operations leaders (89%) report that their technology investments haven’t fully delivered the impact expected.

Manufacturers must do more, faster, with fewer errors and fewer resources. But the tools they are relying on to do that are failing them because the data isn’t correct to begin with. But finding where the breakdown of the data is? That’s the real problem.

The problem: A disconnected supplier network

It’s no secret that manufacturing supply chains are incredibly complex. A single product may have dozens or even hundreds of parts and materials coming from all over the world. Suppliers are as varied as the components they ship.

Many are large companies that are fully tapped into an interoperable network that gives partners good information. But many are not. The varied nature of manufacturing materials brings with it varied vendors, a long tail of smaller commodity suppliers, who may still be operating with manual ERP entry, spreadsheets, and PDFs attached to individual emails.

See also: AI has yet to reach full adoption across most GRC teams, report finds

That means vital documentation like purchase orders, acknowledgments, shipping updates and invoices are coming in from dozens of different sources in different formats, languages and delivery methods.

It’s like a giant game of telephone gone awry. When all that information is combined and fed to AI agents, the inconsistencies are magnified. Production lines stall, invoices don't reconcile, and the planning decisions manufacturers make every day, such as what to order, when to order it, how much to hold, are built on a foundation that can't be trusted.

According to McKinsey’s 2026 global survey of supply chain leaders, only 42% of companies have visibility into tier-two suppliers or beyond, and that number has declined since 2022 as companies cut back on supply chain digitization in favor of other technology priorities.

Inflection points make the problem worse. For example, companies going through an ERP migration that need their supplier network to move with them, or organizations launching a strategic initiative to drive efficiency through automation.

Supplier transaction data informs automation efficiency or invoice acceleration, and if the data isn’t there, the initiative is likely to stall out or fail completely.

Reliable inbound transaction data through a uniform intelligent network

The solution lies in connecting manufacturers to their full direct-supplier community, which won’t happen automatically. Even sophisticated software cannot solve the problem of connecting long-tail suppliers that are not yet EDI-capable or have never sent a digital transaction.

That takes good old-fashioned leg work of outreach, onboarding, and compliance work to get long-tail suppliers connected and transacting accurately into the manufacturer's ERP.

See also: Successful AI products win long after the sales contract is signed

We need to build systems that account for that last mile of the supply chain. Less-sophisticated suppliers still have valuable resources that manufacturers need access to, so we need to plan for the inevitable situations where they aren’t already connected via standard ERP systems.

When that layer is finally automated, the IT director managing a system modernization or the controller trying to accelerate invoice reconciliation and improve cash flow visibility will finally have a foundation worth building on.

Manufacturers have historically dealt in BI-oriented data, formatted to track KPIs and structured for human consumption. AI requires a more continuous stream of raw data to generate reliable predictions and provide context for decisions about what to order, when to expect it and what to pay.

That type of high-quality, real-time data cannot happen with manual, disconnected systems. It is up to partners up and down the supply chain to make sure we change that.

The pressure to show AI progress here is growing, and it creates urgency to move from early experimentation toward end-to-end automation.

The manufacturers that get there fastest will build trust in stages: proving where AI can improve decisions, expanding its role as the underlying data and processes become more reliable, and ultimately allowing it to operate with far less human intervention.

That journey also depends on context no single manufacturer can realistically create alone, including patterns drawn from millions of transactions, changing trading-partner requirements, and proven resolutions to common supply chain exceptions.

See also: The last unconnected asset on the plant floor: The case for ‘smart’ PPE

With that foundation, AI can better recognize what normal looks like, identify when something is off, and take on more responsibility without introducing hidden risk into production, supplier payments, or cash flow. The goal is still full automation. The advantage comes from reaching it on a foundation built to scale.

The upshot: Build for scale

Operations leaders who recognize and act now to close the long-tail supplier gap are building a durable competitive advantage. Not only are they fixing what is broken with current AI initiatives, but every ERP upgrade, every analytics investment, and every subsequent AI application compounds the value of getting that foundation right.

That groundwork also is setting up future growth. With a process in place to accommodate all types of suppliers, standardize their data and connect it to the broader network, the stress of scaling is alleviated.

See also: AI stokes debate over cloud-powered compute vs. on-prem

Most manufacturers have made progress automating the outbound side of their supply chain, but the inbound side has not kept pace. As AI infiltrates more of the industry, that will become more of an issue. AI models trained on clean, bidirectional, ERP-connected supply chain data make more accurate predictions than models running on isolated datasets.

Automated transaction flows on both sides of the supply chain, along with supplier communities that are aligned on requirements, tested, and operating consistently are the missing piece to taking AI technology from buzzword to the growth and return on investment we are all looking for.