Systems Thinking Series 2: Fix the thinking, not just the tools

From the outside, manufacturing looks like a marvel of coordination. Designs become products. Raw materials become precision parts. Teams work across continents to launch complex systems at scale.

Look closer, and the cracks begin to show. 

A design change triggers a costly delay downstream. A sourcing team buys a near-duplicate part because they couldn’t match it to a previous one. Quality spends days creating a control plan from scratch. A new hire spends months decoding tribal knowledge no system captured. It’s not that people aren’t doing their jobs. It’s because the system between the systems isn’t working.

The Silent Killer: Siloed Thinking

Modern manufacturing is built on extreme specialization — a mindset rooted in one of the oldest economic ideas: Division of Labor, as described by Adam Smith in The Wealth of Nations. Smith argued that breaking work into smaller tasks would boost productivity. And he was right. Specialization enabled workers, and later departments, to become highly efficient in narrow domains, driving output and quality.

But over time, this logic didn't just shape how factories operated, it shaped how software was designed. PLM for design. ERP for procurement. QMS for quality. Each tool evolved to serve its own function, but few were built to interact fluently with others. We didn’t just divide labour; we divided logic. And in doing so, we gradually lost the thread of systems thinking.

This loss shows up in the everyday dysfunction of manufacturing operations:

• A spec change is made without understanding its impact on quality procedures.

• A part is redesigned from scratch, despite a similar one already existing.

• A supplier is selected for cost savings, only to be rejected due to tolerance issues.

Everyone is solving their problem — but in isolation. The tools aren’t broken; they’re just blind to what happens outside their silo. What results is a disconnected decision-making chain, where good choices in one area lead to poor outcomes in another. It's the manufacturing version of the Blind Men and the Elephant — each department feeling one part, but no one seeing the whole.

When Tools Reinforce the Problem

Enterprise systems didn’t just inherit functional silos — they entrenched them:

• PLM and PDM tools manage drawings and revisions, but rarely speak to cost or manufacturability.

• ERP systems focus on transactions, not design intent.

• QMS platforms ensure compliance, but don’t trace quality issues back to engineering assumptions.

• And shared drives hold what doesn’t fit anywhere else.

  
  

Beyond the siloed nature of design, these software’s weren’t built to handle unstructured, visual, or cross-functional data. They store documents. They move workflows. But they don’t interpret, correlate, or explain. This passive data flow leaks intelligence at every handoff.

Here’s what it looks like on the shop floor:

• Sourcing can’t see the tolerance ranges that drive cost, buried inside drawings.

• Procurement re-quotes a part that already exists under a different name.

• Quality rebuilds control plans manually, disconnected from design specs.

• New engineers spend weeks learning undocumented tribal knowledge that no system captures.

  
  

Each function moves quickly, but in slightly different directions — driven by its own KPIs, tools, and timelines. What’s missing is the ability to see the whole system, understand ripple effects, and respond with coordinated insight.

The legacy systems didn’t fail — they just weren’t designed for this level of complexity. But in today’s environment of fragile supply chains, tighter timelines and exploding product complexity, that disconnect is more than inefficient. It’s a risk.

This is not a workflow problem — It’s a thinking problem

Much of manufacturing’s digital transformation has focused on automation, cost optimization, and data capture. But the bigger leap is not mechanical, its mental.

We need to shift from:

• Managing files to understanding relationships

• Completing tasks to building feedback loops

• Documenting activity to building shared intelligence

That requires a new kind of approach — one where drawings, specs, quotes and QA plans are no longer static artifacts, but living components of an integrated system.

Why this matters more than ever

Today’s manufacturing challenges aren’t single-variable equations. They are system-level puzzles:

• Fragile global supply chains

• Compressed product launch cycles

• Increasing regulatory scrutiny

• Rising design complexity

• High workforce churn and loss of tribal knowledge

More tools won’t solve this. Better systems thinking will.

How AI changes the equation

In many industries, AI has already transformed how teams work:

• In customer service: 80% of queries are handled by bots.

• In sales: Emails are auto generated and leads scored and prioritized.

• In software: Code written by AI copilots.

But in manufacturing? We’re still trying to figure out if the drawing we’re using is the right version.

This isn’t a tooling gap. It’s a thinking gap. 

AI doesn’t just automate tasks, it thrives when tools are built around context, connection and continuous learning. It enables the kind of intelligence our legacy systems were never designed for, such as:

• Understanding unstructured data like drawings and specs

• Connecting insights across functions

• Preserving institutional knowledge

• Predicting downstream effects before it hits the floor 

  
  

10 things next-gen manufacturing tools should do

Here's what modern, AI powered systems should enable:

1. Sync BOMs automatically

2. Detect change impacts across functions

3. Auto-generate QA documents from drawings

4. Build Digital Twins from real-world data

5. Flag manufacturability issues in real-time

6. Capture and reuse tribal knowledge

7. Extract specs and suggest suppliers from drawings

8. Classify parts by function for promote reuse

9. Generate work Instructions and digital threads

10. Suggest variants based on history and context

    
    

Systems Thinking — at Software Scale

For decades, we’ve tried to make manufacturing more efficient by automating parts of the process. But the biggest gains are no longer in speed. They’re in understanding and building systems that can connect the dots between design, sourcing, quality and production. AI gives us a real shot to think like a system.

Not to replace experts — but to amplify them. To capture what they know. To connect what they do. And to make that intelligence accessible across the organization.

AI, complexity, and the way forward

Legacy enterprise systems tried to solve complexity by standardizing it — breaking it into fields, forms, and workflows. It worked for the information era; it worked with static processes and predictable supply chains when one could manage structured data in siloed functions. They weren’t built for the intelligence era. They weren’t built to make sense of interconnected complexity or empower teams to navigate it. 

The complexity we face today in manufacturing, design, sourcing and quality doesn’t sit neatly in rows and columns. It lives in drawings, emails, PDFs, tribal knowledge, and informal handoffs. It is unstructured, visual and dynamic. It’s where most intelligence lives.

AI, especially Vision and Language Models (VLMs), offers a radically different way forward:

• Interpret drawings like an experienced engineer would

• Spot patterns across revisions, suppliers and QA records

• Summarize tribal knowledge from unstructured documents 

• Bridge gaps between technical and non-technical stakeholders

• Reveal system-level issues before it becomes failures

  
  

AI can help us design systems that think like systems — not just track what we’ve done but understand why we did it and what happens next. AI gives us the first true tool for context and interconnection. If we want to fix manufacturing, we can’t just fix the tools. We must fix the thinking that created them. And that begins by seeing the system as a whole and designing for how it really works.