ML Journal

AI is enabling factories to respond faster to customer demand, adapt to disruption, and operate more profitably. 

TAKEAWAYS:
● Modernize smart factories via digitization; choose technologies based on business outcomes and validate data availability before designing AI solutions.
● Prioritize use cases for business impact, short time‑to‑value, and data availability; target 60–90 day pilots that can scale.
● Embed AI into human operational workflows and governance so intelligence drives actions while people retain accountability and judgment.  

In manufacturing across the globe, a consistent theme emerged in 2025: volatility is now an operating condition. Geopolitical friction, supply disruptions, regulatory shifts, and rising customer expectations have combined to make speed, resilience, and real‑time visibility essential. Manufacturers best prepared to absorb disruption had unified data, process automation, and operational transparency, while those hampered by fragmented systems and manual workarounds struggled to respond effectively. While modern factories often conjure up images of physical automation such as robotics and automated lines, digital automation is equally important, with AI taking center stage.

This year, the conversation has shifted from AI experimentation to embedding intelligence and resilience into daily operations and integrating coordinated capabilities into core workflows. Agentic AI is one of the most consequential developments. These systems go beyond providing insights and recommendations by autonomously planning, deciding, and taking actions to achieve a goal.

Principles to Help Pilots Scale

Three practical principles separate pilots that scale from those that stall: business outcome before technology; prioritize for impact, speed, and data; and industrialize with humans in the loop.

1.     Business outcome before technology

Successful projects begin with a measurable business problem, such as boosting first‑time service fixes, shortening order‑to‑cash, reducing obsolete inventory, or lowering energy consumption. Define the key performance indicator (KPI) first, then assess whether you have the data required to solve it. If the data isn’t available or trusted, the immediate focus should be data readiness: integration, master‑data remediation, and governance.

Real-life example: A materials‑handling original equipment manufacturer (OEM) used historical service records to predict likely faults on forklifts and provisioned the correct spare parts for field engineers. The result was an approximately 30% increase in first‑time fixes, directly improving margins and customer service through a rapid pilot.

2.     Prioritize for impact, speed, and data

Manufacturing is target‑rich for AI. Prioritize use cases using a simple triage:

• Impact: Will solving this issume significantly improve something important to the business—margin, throughput, quality, lead time, or customer service?
• Time‑to‑value: Can a meaningful pilot be delivered within 60–90 days?
• Data: Do you have the data needed to support AI and automation? Prioritize use cases where data readiness exists, even if imperfect, over those requiring time to build a data repository.

Small, fast wins build momentum and confidence.

Real-life example: A mid‑market equipment manufacturer still receives many of its orders via PDF email. Capturing these with optical character recognition (OCR), validating inventory with simple rules, and auto‑confirming when possible produced immediate ROI and relieved customer service teams hours of repetitive work. Adoption was high because it added true value and removed tedious steps.

3.     Industrialize with humans in the loop

While we’re moving toward more agentic capabilities, full autonomy entails a long journey to build trust. Frame AI as a “digital co-worker,” reducing cognitive load, surfacing prioritized actions, and allowing people to make final decisions in safety‑critical or high‑risk contexts. Human oversight shortens the trust curve, improves training data, and preserves accountability.

Foundations matter, so integrate before you automate. AI can only scale on a solid digital foundation and only be effective when it has context within the business processes it is acting on. Organizations that invested in integrated platforms connecting machines, logistics, planning, and service are able to activate AI more quickly. Removing data silos turns visibility into recommended actions: predictive inventory shortfalls trigger auto‑replenishment, plans adapt in real time, and exceptions are addressed before they cascade.

A recommended sequence might be: process mining to diagnose variability and bottlenecks first; automate repetitive steps with robotic process automation (RPA); and finally, apply AI for pattern recognition and optimization.

Real-life example: One contract steel manufacturer used process mining to reveal hundreds of return‑handling variations. Process redesign plus automation subsequently delivered measurable improvements in cycle time and cost.

Manufacturers need a pragmatic rollout playbook:

• Diagnose with low‑friction visibility tools (process mining, transaction audits).
• Automate routine, documented tasks first, before adding predictive layers.
• Prioritize AI use cases that can deliver results in 60–90 days with impactful KPIs.
• Keep the human in the loop to build confidence and trust.
• Measure business KPIs and iterate.

2026 is about realizing AI value, not by chasing novelty, but by industrializing proven approaches on solid digital foundations. Start with the business problem, move quickly on high‑value, short‑cycle pilots, keep people at the center, and treat agentic capabilities as workflow amplifiers. That’s how factories become smarter, more adaptive, and more profitable in measurable time.

To learn about how Infor Velocity Suite helps customers achieve sustained business value and more customer use cases, click here.  M

About the author:

Andrew Kinder is Senior Vice President Industry Principal for Manufacturing at Infor.