Manufacturers depend on people whose experience is difficult to replace. The buyer who knows which supplier will answer the phone when a shipment slips. The maintenance technician who recognizes a machine’s behavior before an alarm sounds. The quality engineer who remembers investigating a similar defect five years ago.
Yet look at how those people spend their days. Much of their time goes into finding records, chasing responses, reconciling information, and preparing documentation. That non-value-added work may be necessary today, but it deserves the same scrutiny manufacturers apply to waste elsewhere in their operations. Where can it be reduced or eliminated so experienced people have more time to apply their judgment and expertise?
Adding another dashboard doesn’t necessarily change that. A dashboard might show that a purchase order is late. The buyer still needs to find the supplier’s latest response, determine which production orders depend on the part, check available inventory, and decide whether expediting the shipment is worth the cost. Each system holds part of the answer. The buyer supplies the context and carries the work between them.
This is where the conversation about agentic AI becomes relevant to manufacturing. There is substantial work surrounding every judgment call, and much of it follows patterns that a digital worker could help manage.
Where the Work Accumulates
Consider supplier management. A buyer may have hundreds of open purchase orders, with confirmations and delivery updates arriving through email, attachments, and supplier portals. Even when the ERP is current, the explanation behind a changed date may be sitting in an inbox.
A digital worker assigned to supplier follow-up could reconcile those responses with open orders, identify exceptions, and prepare proposed updates for approval. The buyer would still manage the relationship and decide how to respond to a disruption. But the morning could begin with a clearer view of what needs attention.
That same burden appears elsewhere in the manufacturing value stream.
Engineering needs field feedback, quality findings, and prior design decisions to inform product changes. Production teams need maintenance history and operating procedures alongside current machine conditions. Quality engineers need evidence from several systems before they can prepare a credible corrective action. Field technicians need the equipment’s history, applicable service guidance, and the right parts before arriving at a customer’s site.
These are different jobs, but they share a dependency: someone has to gather the relevant information and make it usable before the next person can act.
Give the Digital Worker a Real Job
Calling an agent a digital worker is useful because it changes how we evaluate it. A new employee needs a job description, access to the appropriate systems, a manager, and a way to determine whether the work is being done well. An agent needs the same clarity.
“Help procurement” leaves too much undefined. Following up on unconfirmed purchase orders is a job with identifiable inputs, exceptions, and outcomes. You can measure how much time it consumes today. You can specify what the agent may prepare, what it may update, and when a buyer needs to step in.
The distinction between assistance and accountability matters. A supplier response may be straightforward to interpret, but accepting a delivery change could affect production commitments. The workflow needs to recognize that boundary.
It also needs to use the tools suited to the task. A language model can interpret an email. Inventory availability comes from the business system. A calculation should run in a tool that produces repeatable results. The agent coordinates that work so the buyer can evaluate the implications.
Start Where Capacity Is Constrained
The best first use case may be a recurring administrative burden that everyone has learned to tolerate.
Look at where your experienced people spend time on non-value-added work, gathering information or following up on tasks that should already be moving. Then ask what they could focus on if that work were reduced or eliminated. Would buyers get ahead of supplier risk? Would quality engineers spend more time preventing repeat failures? Would technicians have more capacity to support production?
Those answers give the project a business purpose. They also help distinguish a useful digital worker from an impressive demonstration.
CNXN Helix® Center for Applied AI and Robotics is Connection’s center dedicated to helping organizations identify, develop, and implement practical AI solutions. It brings applied AI expertise together with Connection’s broader technology and services capabilities to help manufacturers build solutions around their operations and business goals.
Our white paper, Your Next Hires Never Sleep: 5 Agentic AI Use Cases for Manufacturing, examines opportunities across supplier management, product design, production, quality, and customer service, including the digital worker roles and Microsoft capabilities supporting each.
Talk to your Connection Account Team or visit connection.com/ai to explore where a digital worker could add capacity to your manufacturing business.
Ryan Spurr
Ryan Spurr is the Director of Manufacturing Strategy and a Helix AI Advisory Director with 25 years of experience in manufacturing, information technology, and portfolio leadership. He leads Connection's Manufacturing practice, go-to-market strategy, client engagement, and advisory services focusing on artificial intelligence, information technology, and operational technology that make manufacturers more digitally excellent.
