Why almost every manufacturing company wants AI, but only a few are ready – and what that means for your strategy.
The Gap Nobody Talks About
A recent study by Redwood Software surveying 300 manufacturing companies worldwide delivers a sobering number: 98% are exploring AI-driven automation. But only 20% feel ready to deploy it at scale.
78 percentage points difference. That’s not a gap. That’s a chasm.
And the details don’t make it any better:
- 70% have automated less than half of their core operations
- 78% have automated less than half of their critical data transfers
- Only 40% have automated exception handling
The study calls this phenomenon the “Mid-Stage Automation Maturity Trap” – companies have invested in individual systems but remain stuck there.
The Island Problem
Imagine your IT landscape as an archipelago. Each island is a system: ERP here, MES there, QMS on a third island, production planning on a fourth.
On each island, automation works beautifully. The production scheduler optimizes automatically. The ERP generates purchase orders. The QMS logs inspection results. Each system on its own – a well-oiled machine.
But as soon as something needs to move from one island to another?
Someone gets in a rowboat. Manually. With Excel. Via email. Through copy-paste.
Automation ends at the waterline. And that’s exactly where value gets lost.
Why AI Fails in This Landscape
Here’s the critical point many overlook: Artificial intelligence is not a magic wand that connects islands.
Quite the opposite. AI amplifies the problem.
A large language model can be as powerful as you like – if it has no access to current production data because that data is locked in the MES, it’s blind. If it gets no feedback from the ERP because data transfers run manually, it works with outdated information. If exception handling isn’t automated, the AI doesn’t even know a problem exists.
AI can only be as good as the data streams that feed it.
The companies that feel “ready” – the 20% – don’t have better AI. They have better orchestration. They built bridges between their islands before they started thinking about AI.
The 5-Stage Model: Where Do You Really Stand?
In our work with mid-sized manufacturing companies, we’ve developed a maturity model that maps this journey. Not as an academic exercise, but as a practical tool for positioning.
Stage 1: AWARE
The company knows AI is relevant. There are initial experiments – a chatbot here, a pilot project there. But no strategy, no integration, no measurable ROI.
Typical indicator: “We tried ChatGPT.”
Stage 2: ACTIVE
An AI strategy exists. First use cases are identified, budget is allocated, one or two projects are running. But the projects are isolated – they solve individual problems, not system problems.
Typical indicator: “Our predictive maintenance pilot is running on one line.”
Stage 3: OPERATIONAL
AI runs productively in at least one core process. ROI is proven. Governance exists – policies, responsibilities, processes for new use cases.
Typical indicator: “60% of inquiries are answered automatically by our bot.”
Stage 4: SYSTEMIC
AI is integrated into the majority of workflows. There’s a central platform. Cross-functional teams work AI-assisted. New projects use AI by default.
Typical indicator: “Every department has at least one productive use case.”
Stage 5: AI NATIVE
The organization is fundamentally built around AI. Processes, roles, decision pathways are AI-first designed. Without AI, the company wouldn’t function this way.
Typical indicator: “AI agents orchestrate our core processes.”
The Honest Assessment
Where does mid-sized manufacturing stand today?
Based on our assessment:
| Stage | Share Today | In 3 Years |
|---|---|---|
| Aware | ~60% | ~30% |
| Active | ~25% | ~35% |
| Operational | ~12% | ~25% |
| Systemic | ~3% | ~8% |
| AI Native | <1% | ~2% |
The majority starts at Stage 1 or 2. That’s not a flaw – it’s reality. The Redwood study confirms this with hard numbers.
But here’s the crucial point: The difference between “Aware” and “Operational” is not primarily technological. It’s organizational.
The 70/30 Rule
In AI transformation, an uncomfortable truth applies: 70% of success comes from people and processes. Only 30% from algorithms and technology.
This contradicts the narrative the tech industry sells us. There, it’s all about model sizes, benchmark scores, new architectures. All important – but not decisive.
What’s decisive:
1. Data integration – Does your data flow between systems? Automatically, in real-time, bidirectionally?
2. Process clarity – Do you know how your processes actually work today? Not on paper, but in practice?
3. Organizational readiness – Is someone responsible for AI? Not as a side project, but as their main job?
4. Change capability – Can your organization absorb change? Or do new tools fail due to resistance?
Companies that can answer “yes” to these four questions are the 20%. The rest struggle with the island problem.
The Way Out of the Trap
If you recognize yourself in the “Mid-Stage Maturity Trap”: Don’t panic. The way out is known. It just requires a shift in perspective.
Step 1: Map Your Islands
Before you think about AI, understand your current landscape. Where are the system boundaries? Where are the manual handoffs? Where does data disappear?
A simple test: Follow a customer order from placement to delivery. How often does it change systems? How often does a human touch the data?
Step 2: Build Bridges, Not New Islands
The reflex with AI projects is often: “We need a new tool.” Wrong. You need connections between existing tools.
APIs. Middleware. Data platforms. Event streaming. The technologies exist. The question is whether you prioritize them – or whether you keep investing in isolated point solutions.
Step 3: Automate the Handoffs
Exception handling isn’t sexy. But it’s decisive. If 60% of your exceptions are handled manually, then 60% of your AI outputs will need manual post-processing too.
Start where it hurts: at the handoffs, at the exceptions, at the escalations.
Step 4: Think in Workflows, Not Tools
The question isn’t: “Which AI tool should we buy?”
The question is: “Which end-to-end process do we want to transform?”
The difference sounds subtle but is fundamental. Tools solve partial problems. Workflows solve business problems.
The Target Vision: Human Above the Loop
When we talk to clients about “AI Native,” a concern often comes up: “Does that mean AI takes over everything?”
No. The opposite is true.
Our target vision is called “Human Above the Loop” – the human above the loop. AI operates within defined boundaries. It optimizes, it automates, it suggests. But strategic decisions, ethical considerations, customer relationships – those remain human.
The irony: The better the AI integration, the more valuable human work becomes. Because people can focus on what only people can do. Instead of copying data from A to B.
The Homework for 2026
The Redwood study shows: The problem is known. The solutions exist. What’s missing is execution.
If you do only one thing this year: Map your data flows. Not your systems – you know those. But the flows between them. The manual handoffs. The Excel bridges. The email workarounds.
This isn’t glamorous work. But it’s the work that makes the difference between the 98% who want AI and the 20% who can actually use it.
The question isn’t whether AI will transform your industry. The question is whether you’ll be among the companies that benefit from it.
Source: Redwood Software, “Manufacturing AI and Automation Outlook 2026”, January 2026. Based on a global survey of 300 manufacturing companies.
About the author: Tim is Associate Partner at Mesakumo, supporting mid-sized manufacturing companies with AI transformation. Previously, as Strategic AI Lead at KION Group, he was responsible for over 48 AI projects with a targeted EBIT contribution of €500 million.