In the past few years, AI copilots have been widely adopted by organizations to improve productivity. Deloitte’s 2026 enterprise AI survey found that sanctioned AI access grew from under 40% to around 60% of workers in one year, but only 25% of organizations had moved 40% or more of their AI pilots into production. McKinsey’s 2025 survey similarly found that 88% of organizations use AI in at least one business function, while only about one-third had begun scaling AI programs enterprise-wide.
The market is now shifting rapidly toward AI agents. McKinsey found that 62% of organizations were at least experimenting with AI agents, including 23% already scaling an agentic system somewhere in the business. Deloitte found that 85% of companies expect to customize agents to their own business needs, suggesting that enterprises increasingly see generic copilots as a starting point rather than the end state.
NitroLens recently joined Taiwan Tech Talk during SF Tech Week, and one topic that stood out was how AI is moving beyond software workflows and into the physical world.
As AI becomes more vertical and enters environments like manufacturing, a bigger question emerges:
How does AI perform in the real world, where operational context matters?
The state of AI adoption for manufacturing companies
More than 70% of manufacturers surveyed by the Manufacturing Leadership Council are already using generative AI tools such as ChatGPT or Microsoft Copilot, up from 46% in 2024, and nearly 90% expect to increase usage over the next two years.
But manufacturers still describe themselves as early in AI maturity, with data quality and workforce skills among the biggest barriers. Manufacturers are also still underusing the data they already collect. Rockwell reports that only about 44% of the manufacturing data collected is being used effectively.
The biggest gap is between using AI tools and embedding AI into the operating system of a manufacturer.
Manufacturers already have a huge amount of useful information across ERP, MES, quality systems, spreadsheets, email, machine data, and experienced employees. The problem is that this context is fragmented.
That makes operational AI much harder than deploying a chatbot.
This is why the bottlenecks are increasingly shifting toward:
Data readiness → system connectivity → business context → workflow redesign → governance → human adoption
What challenges do manufacturing companies need to overcome to win in the AI race?
To win in the AI race, manufacturing companies need to overcome a different set of challenges than software-first companies.
The biggest issue is whether the organization has the systems, business context, and governance required to make AI work reliably in real operations.
1. Legacy systems that were never designed for AI
Many manufacturers operate with decades of accumulated systems and equipment. Connecting AI to those environments without disrupting production is significantly harder than deploying a standalone copilot.
Legacy infrastructure, inconsistent standards, and disconnected workflows remain major barriers to industrial AI.
2. Fragmented data and missing business context
Manufacturing data is scattered across ERP, MES, CRM, spreadsheets, email, machine systems, and employees’ heads.
Even when the data exists, it often lacks the context AI needs to understand relationships, constraints, commitments, and exceptions. Nearly half of manufacturers in a recent Manufacturing Leadership Council survey rated their operational data as having low AI readiness.
3. Capturing domain knowledge that lives in people’s heads
Manufacturing depends heavily on implicit knowledge: which supplier can actually deliver, which material works best on a particular machine, when production can break the normal scheduling rule, or which customer commitment cannot move.
AI needs access to this operational knowledge, not just structured system data.
This becomes even more important as experienced workers retire and manufacturers try to preserve institutional knowledge.
4. Trust, governance, and operational reliability
The tolerance for AI error is very different when software is affecting a production line, delivery commitment, quality decision, or physical equipment.
Manufacturers need clear permissions, human approval points, auditability, cybersecurity, and mechanisms for handling exceptions.
How NitroLens will play a role in this transition
NitroLens helps manufacturers move from fragmented systems and isolated AI tools to Central Intelligence that understands how the company operates as a whole.
Rather than adding another standalone copilot, NitroLens connects existing systems, business rules, decision history, priorities, and operational knowledge so AI agents can reason across teams and help move work forward.
1. Connect fragmented systems and context
Bring together the information already spread across ERP, MES, CRM, spreadsheets, documents, and employee knowledge.
2. Capture how the business actually works
Make policies, priorities, relationships, exceptions, and past decisions available as context AI can understand and reuse.
3. Apply AI to real cross-functional decisions
Start with workflows where multiple teams need to coordinate and no single department has the full picture.
Mini use case: an urgent custom order
Today, an urgent custom order can trigger back-and-forth across sales, production, procurement, and management.
Each team sees only one part of the problem. Sales wants to respond quickly. Production may already be at capacity. Procurement may not be able to secure materials fast enough. Management has to balance the customer opportunity against other commitments.
By the time everyone aligns, a customer quote may take 2+ days.
With NitroLens, an AI agent can identify the opportunity, pull the relevant company context, check production capacity, supplier availability, management priorities, and past decisions, then bring the recommendation to the right person for approval.
Once approved, the agents can coordinate the downstream work — enabling the company to respond in as little as 30 minutes.
4. Build from one workflow into company-wide intelligence
Each successful use case adds more context, decision history, and operational knowledge to become the company-level intelligence that understands how the business actually works and helps teams make better decisions together.
Learn more at nitrolens.ai