Manufacturers are facing a decision point with AI. The technology has moved beyond experimental pilots into production deployments that deliver measurable results.
Companies deploying agentic AI now are reporting productivity gains that seemed impossible just months ago. Those still evaluating options risk falling behind competitors who’ve already transformed their operations.
CEO of Propel Software
Why This Time Is Different
AI promises have circulated through manufacturing for years. Early implementations focused on predictive analytics: forecasting demand, scoring leads, anticipating equipment failures. And, the results were incremental at best.
Agentic AI represents a fundamental shift. Rather than analyzing past patterns to predict future outcomes, agents take autonomous action within workflows. They don’t just identify that a component price increased; they calculate impact across your product portfolio, flag affected customer commitments, and initiate sourcing alternatives.
What makes manufacturing uniquely suited for agentic AI? Manufacturing operates in constant flux. For example, our manufacturing customers are often faced with daily realities like these:
- Supplier pricing changes without warning
- Quality issues surfacing in the field
- Regulatory requirements shifting unexpectedly
- Customer demands continuously evolving
- Supply chain disruptions threatening production schedules
As Forrester research on agentic AI in supply chain management notes, simple rules-based automation struggles to cope with the complexity of supply chain exceptions. In contrast, agentic AI has the potential to solve these challenges when properly implemented with governance and security controls.
Success depends on adaptive response, and response speed depends on eliminating the gap between recognizing problems and executing solutions. Traditional automation helped with repetitive tasks, but it often failed or gave misleading results when the inputs changed unpredictably. Analytics identified trends, but left execution to humans.
Agentic AI bridges both. It is autonomous enough to act, while intelligent enough to make contextual decisions.
The Business Case Manufacturers Demand
Manufacturing executives set high bars for technology investments. Most won’t proceed with AI without clear paths to significant returns. Many of the executives Propel is speaking with are targeting improvements in productivity metrics or cost reduction exceeding 20%.
These expectations aren’t theoretical. According to our State of Product Innovation 2025 survey of 800 manufacturing employees, companies actively using AI are reporting measurable results like:
- 52% – productivity gains
- 50% – competitive advantage
- 32% – reallocation of resources to higher-value work
How does agentic AI deliver on these types of returns?
Agentic AI delivers returns through two mechanisms: enabling existing staff to accomplish more by eliminating low-value work, or directly reducing headcount in roles focused on manual coordination and data management.
Consider the operational impact across functions:
- Supply chain operations: Agents monitor inventory requirements and lead times, identifying potential shortages before they disrupt production. The value isn’t a perfect prediction, it’s a faster response. Supply chain software has promised that for 30 years. When disruptions occur, agents can autonomously initiate procurement workflows and flag alternatives rather than waiting for someone to notice the problem.
- Quality management: Complaint data that requires hours of manual analysis gets processed instantaneously. When case volumes spike unexpectedly, agents identify root causes and route corrective actions before issues multiply. Historical performance analysis happens continuously, detecting failure patterns that enable preemptive maintenance.
- Change management: Engineering changes that once required days of impact analysis are now completed in hours. Agents automatically assess how modifications affect bills of material (BOMs), quality procedures, supplier relationships, customer commitments, and manufacturing processes, then route approvals to appropriate stakeholders with complete context.
- Training and compliance: Instruction manuals spanning multiple languages get transformed into region-specific training materials and assessments in minutes rather than days. Completion rates improve when content becomes accessible and relevant to local teams.
The Infrastructure Reality
Here’s where many manufacturers stumble: they assume powerful AI eliminates the need for robust underlying systems. Reality proves otherwise.
Agents require structured environments to operate reliably. Without enforced business rules and defined workflows, automation becomes unpredictable and problematic in production.
Why regulated industries face unique challenges
Our customers in manufacturing environments must maintain rigorous process adherence for compliance. When agents lack proper constraints, they optimize for efficiency in ways that violate mandatory procedures.
Here’s a real-world example around a quality scenario: An agent without business context reviews a decade of inspection records and notices certain documentation fields get completed inconsistently, maybe only 40% of the time, without apparent business consequences. Following efficiency logic, the agent stops prompting completion of those fields.
But those fields could document regulatory reporting requirements that haven’t faced recent audits. The moment inspection occurs, incomplete records become compliance violations capable of halting operations. The agent didn’t malfunction, it lacked the business rules to define which processes are non-negotiable.
Effective deployment leverages the business process rules of the underlying application, thus providing the applicable guardrails for the process. Then automation becomes powerful because it respects critical constraints.
Governance Requirements
Manufacturing environments need sophisticated access control. Not everyone should see everything that’s available.
Who needs access to what?
- Engineers require design specifications, but not customer pricing
- Marketing needs positioning information, but maybe not quality test details
- Sales wants delivery commitments without visibility into engineering work-in-progress
Platforms must enable agents to operate within these permission structures automatically. Can your systems ensure AI respects role-based access? When permissions change, do agents immediately reflect those new boundaries? If not, you’re potentially exposing data inappropriately.
What regulators actually require
Audit requirements add another layer. Regulators don’t accept “the agent handled it” as documentation. You need complete decision trails showing:
- What actions occurred
- Who authorized them
- What data informed decisions
- How processes followed mandatory procedures.
Mature platforms deliver this accountability consistently.
The Data Foundation
Automation requires careful execution built on comprehensive data foundations. Combining agent intelligence with thorough data analysis drives genuine business differentiation, but that data layer demands sophistication.
What does comprehensive data connectivity actually mean?
When product lifecycle management (PLM), quality management (QMS), computer-aided design (CAD), enterprise resource planning (ERP), and supplier systems feed agents comprehensive information, they optimize business efficiency across the enterprise. This end-to-end connectivity forms the foundation for reliable automation.
With structured data atop secure, rule-driven architecture, agents make real-time decisions within existing environments. Rather than pulling isolated information from disconnected systems, agents drawing from unified product data across items, BOMs, changes, documents, and metadata deliver context-aware responses that account for relationships and dependencies throughout product operations.
Propel One: AI Agents for PLM & QMS Software
In building Propel’s agentic AI solution for manufacturers, Propel One, we’ve observed that the organizations seeing the most meaningful results share one common trait: they’ve already unified their product data across PLM, QMS, and supplier systems. Those still operating on fragmented data struggle to realize AI’s full potential, regardless of how sophisticated their models are.
Data completeness directly determines agent output quality and the productivity gains realized across operations. The more structured and connected your foundation, the more reliable and valuable your agent capabilities become.
The Competitive Moment
Manufacturers implementing agentic AI now are seeing faster execution, fewer errors, streamlined operations, and more responsive adaptation to market changes. These improvements compound into accelerated time-to-market and improved margins.
What Agentic AI really represents
Agentic AI represents more than incremental productivity improvement. It’s a fundamental shift in how manufacturing operations execute, from reactive problem-solving to proactive process orchestration. Companies building on infrastructure designed for this shift will define competitive standards for the next decade.
The decision isn’t whether to adopt agentic AI. It’s whether to build proper foundations now or scramble to catch up after competitors establish operational advantages.
About the Author
Ross Meyercord is CEO of Propel Software and former Global CIO of Salesforce. He has over 35 years of experience leading enterprise technology strategy and scaling SaaS companies.
Learn more about how manufacturers are implementing agentic AI at www.propelsoftware.com.
Related Categories