Infor Shows a Manufacturing Room How to Move AI Off a Pilot

Applications · Manufacturing
Dr. Henning Dransfeld spent the session on the move off a pilot. Name one plant job. Settle who can see the data. Keep a person accountable for the decision. Pick an agent from the shared library.
By Shashi Bellamkonda · October 7, 2026
Oct 6
Velocity Week, Orlando
57%
named production planning
34%
Infor: US at full scale
500
agents he said are in the library

The session followed the morning overview

Dr. Henning Dransfeld, industry principal for automotive at Infor, led the manufacturing session on October 6 at Rosen Shingle Creek. The title on the screen was "AI in Manufacturing: What's Actually Working - and How to Scale it?"

He opened with a sailing race. The start is not the race. About a third of the way down the course, the boats in front keep the lead because they reach the good wind first. He tied that picture to Infor's Enterprise AI Adoption Impact Index. In the United States wave, full-scale deployment rose from 21 percent in April to 34 percent in the August survey of 550 decision-makers, Infor reported on October 6. He also said efficiency gains customers report back rose 7.1 percent from April to October, with the climb concentrated in North America.

The room poll set the starting point

About 94 people answered a live poll on operational artificial intelligence in manufacturing. Forty-three percent were still evaluating. Thirty-eight percent were piloting one or two use cases. Fifteen percent were live in production with a limited scope. Four percent said agents were embedded across several functions. Dransfeld read the gap between a pilot and a scaled deployment as the work of the hour.

A second poll, about 99 people, asked where they see or expect the biggest impact. They could pick three. Production planning and scheduling led at 57 percent. Finance, procurement, and back-office automation followed at 55 percent. Supply chain planning and demand forecasting came in at 48 percent. Inventory was 33 percent. He said that two years ago finance and procurement would have led. The one-point gap is small. The order is the change he wanted the room to notice.

He gave three steps, then four moves

The working sequence was diagnose, automate, then hand work to agents. Process mining finds the slow point. Robotic process automation takes the steps where nobody has to decide. Agents then support the operator, and in some jobs one agent can pass work to another.

The slide for the 34 percent listed four moves. Start with a business problem narrow enough to staff. He told a supply-chain case without naming the customer. The opening idea was a vision. The team sent the customer back for two weeks. The workable job was a ship-west or ship-east decision made from Singapore, using tools the customer already ran. Agent work came after that, on the next problems.

Address data access early. The slide cited the August index: 32.6 percent named data security as the top priority for success. The public write-up of that index says 33 percent cite data security, sovereignty, or compliance as the single greatest barrier. Keep a person in the loop. The slide cited a Connect Global 2026 report that 83 percent of the workforce will not trust artificial intelligence without human oversight. Govern in proportion to the risk. The slide cited a Grant Thornton 2026 survey that 78 percent of executives lack confidence they could pass an independent governance audit within 90 days. He said governance is what blocks the move from a pilot to the 34 percent.

He mapped the person-in-the-loop point on a car. Alerts first. Then a recommendation, such as lane assist. Then a confirm-before-it-moves step. Then an override. Then a monitor while the vehicle drives. A car with no steering wheel is the last box. He does not expect a discrete factory to reach that last box in the next 20 years. Too many points still need someone who owns the call. Job profiles shift toward coordination.

He named two published customer jobs

Combilift, the multidirectional forklift maker, was the service example. Infor has published it since September 13, 2022. Parts recommendations inside CloudSuite Industrial were predicted to raise first-time fixes 30 percent and cut service-job cost 40 percent. A later customer account says first-time fixes rose 30 percent, service jobs finished 30 percent faster, and the time to find related parts fell from 20 minutes to 5. On stage, the point was one visit with the right parts instead of several return trips.

Ring Container Technologies was the document example. Infor's account says robotic process automation on signed bills of lading made customer resolution on shipment questions 96 percent faster, with value in under 60 days. The customer story page puts the annual saving at about 12,000 hours.

I asked how the library gets built

I asked how Infor creates the agent library, and whether that library is common across the system. Dransfeld answered from automotive. The first agents started three or four years ago from problems the industry councils in the United States and Europe kept repeating. Quality inspection and invoice processing came up first. People from the business wrote clearer cases than the technical submissions.

When strategists collected favorite agents across industries, he said about 80 percent of what the industrial team already had was reusable. Automotive added its own layer for electronic data interchange and exchange platforms. This year he said the council no longer walks through the 500 agents in the library. The question is the business problem, then which agent to pick. A shortage-prevention agent for material constraints, tied to accounts receivable, is one he named as moving into the standard set.

The adoption path on the last content slide runs from a roadmap session to ranked use cases, a plan with pricing on the table, a proof if one is needed, then a kickoff on that same price. Sprint one is forward deployed engineers building beside the customer's team. The footer said the room leaves with a roadmap it owns.

The last poll was about the spend

After the case for speed, about 76 people answered. Forty-one percent said they do not yet have clarity on the right investment level. Twenty-six percent are increasing investment over the next 18 months. Twenty-one percent said artificial intelligence still has to compete with other priorities and is not a top three. Twelve percent said spend already matches the stated goal.

He closed with a line from Erik Brynjolfsson, who directs the Digital Economy Lab at Stanford. Stanford Graduate School of Business published it this year: "Which of your tasks does AI substitute for? Which does it amplify? Shift your weight toward the second list before someone shifts it for you."

For the buyer

Take one job on the schedule. Ask which step on the car chart that job is allowed to reach, which agent in the library already covers it, and what the industry council still has to build. The session's offer is a roadmap from that conversation, with forward deployed engineers on the first sprint.

Sources

  1. Dransfeld, Henning. "AI in Manufacturing: What's Actually Working - and How to Scale it?" Infor Velocity Week, Rosen Shingle Creek, Orlando, 6 Oct. 2026. Session slides and live polls, slido.com #3830 479. Author's notes.
  2. Infor. "US Businesses Are Scaling AI Fast. Now Comes the Harder Part." Infor, 6 Oct. 2026, www.infor.com/blog/scaling-ai-governance.
  3. Infor. "Infor Unveils Its Expanded Infor Industry AI as New Global Research Finds Off-the-Shelf AI Falls Short for Two Out of Three Businesses." Press release, 6 Oct. 2026. Survey of 2,111 decision-makers, August 2026, conducted by YouGov.
  4. Infor. "Combilift Boosts Equipment Uptime and Service with Infor AI." Press release, 13 Sept. 2022, www.infor.com/news/combilift-boosts-equipment-uptime-customer-service-infor-coleman-ai.
  5. Infor. "Ring Container Saves 12,000 Hours Annually with Infor AI and Enterprise Automation." Customer story, www.infor.com/customer-stories/ring-container-saves-time-resources.
  6. Brynjolfsson, Erik. Quoted in "Almost Every Job Has Tasks That AI Can Change." Stanford Graduate School of Business, 2026, www.gsb.stanford.edu/insights/almost-every-job-has-tasks-ai-can-change.
  7. Bellamkonda, Shashi. "Infor Says Enterprise AI Has to Start with the Process, Not the Chatbot." shashi.co, 6 Oct. 2026, www.shashi.co/2026/10/infor-says-enterprise-ai-has-to-start.html.
Disclaimer: This blog reflects my personal views only. Content does not represent the views of my employer, Info-Tech Research Group. AI tools may have been used for brevity, structure, or research support. Please independently verify any information before relying on it.