Why do companies hesitate to adopt AI? – A systemic view from production hell

Recently, I attended a startup association event in Berlin where they shared findings on AI use in companies and startups. One figure stuck: only 3% of companies use AI in their internal processes. When a trusted source cites a number, stay wary. A low figure is likely lower, a high one higher, except in obvious fakes. So, why 3%? Why so low? With all the hype and money poured into AI, why are companies – always eager to save time and money – so hesitant?

The answer ties to the production system: it’s no playground for new tech – it demands reliability, not experiments.

The Nature of a Production Process

I dug deeper – it’s less obvious than it seems. If AI can handle repetitive tasks and save costs, why not use it? A closer look reveals the system’s logic. Manufacturing is a finely tuned sequence of tasks – a balanced system. Supply chains, materials, and tools must mesh seamlessly. “Production hell,” a term from automakers, captures it: a great car design means nothing if production falters. Producers suffer because the design is frozen – change one part, and the whole breaks. You’re forced to make it work as planned. The system punishes deviations. It’s built for stability, not experiments.

False Assumptions About AI and Production

Finding manufacturing solutions can mean weeks or months of struggle.

Take a legendary story about a critical rocket part. The supplier quoted $100,000 for it. But the CTO said, “This is no different from a garage door actuator – it shouldn’t cost more than $3,000.” He assigned a sharp engineer to hit that target. Nine months later, the cost-cutting mission succeeded. The delighted engineer wrote a five-page letter to his boss, detailing the twists and turns of his creative process and how he finally brought down costs to $4,500. Five minutes later, he got a reply: “OK.” No thanks, no fanfare, no praise. Nine months for one part. Yet every rocket mission saved $95,500. That’s production hell. Now imagine: Could AI have helped here? Yes – but only before the manufacturing process is set and running. Once it’s locked in, there’s no easy way to improve this actuator process.

This shows why AI adoption in production is tough – it’s complex. The question of AI adoption in manufacturing rests on shaky assumptions. First, it presumes a universal, polished AI exists, able to work like a human. That oversells the tool, expecting it to slide into production as smoothly as a worker. Second, it assumes production processes bend easily. They don’t. Both assumptions fail. Production is a web of hundreds of compromises. Humans, machines, and processes form a resilient network. Assuming an outsider like AI can just improve it ignores this logic and risks disaster. Changes must fit the system’s rules – it’s designed to churn out uniform results at minimal cost, not to host experiments. “Never change a winning team” becomes: Don’t tweak a running process unless you want trouble.

The Cost of Disruption

Companies cling to their processes for a reason: a system change is a wrench in the gears – costs soar and sales vanish. No income, no work – the company and its factory hit “nirvana.” A day’s downtime can burn millions, cost jobs, lose customers. AI causes not just technical risks – they’re existential. It’s not simply about tweaking tools or steps; you’d have to rethink the product, tasks, process, and the entire supply chain. Destabilize the system, and it bites back fast.

The Illusion of Progress

Viewing AI as a single, mature tool is wrong. It’s many models, growing daily, target tasks foreign to production. The hype promises progress, yet the system’s logic is clear: any newcomer must outdo the current setup under all constraints. Until it proves that, it stays out. AI evangelists call data “the oil of the 21st century,” pitching: “Hand over your data, we’ll slash inefficiencies and costs.” But production veterans, forged in real “production hell,” remain wary. Their humility and caution make them doubt these grand promises.

The Role of IT and Pragmatic Skepticism

In businesses, information technology – IT, informatics, computer science – organizes data flow to keep production humming. Why aren’t they leaping onto the AI train? Are they failing? No – the system runs well, so they’re doing their job. When AI proponents push, “When do you add AI?” IT experts counter with hard, practical questions: How will AI optimize step 12? What does it do with our data, and what’s the outcome? Prove the gain. As guardians of stability, their caution shields the system from rash leaps into the unknown.

The Market as the Driving Force

The smartest move for companies now is to wait. Successful producers know their craft – or they’d have left the field long ago. Change comes from outside, from the market: customers abandon a product (think telephone books with the internet, digital cameras with smartphones), or tech shifts force adaptation. It’s never voluntary – remember: “Never change a winning team.” The system only bends to force; it doesn’t budge without cause. AI must wait until new circumstances demand it. That doesn’t mean IT ignores AI – they just take it with a grain of salt.

Conclusion: The Wisdom of Restraint

So, why the hesitation? A production process isn’t an experimental lab – it’s a system needing stability and change only when forced. Current processes hold steady – AI doesn’t fit. That shifts with new products and fresh “production hell,” not now. Manufacturers succeed for a reason: they master their process, technology, and the balancing act of a vast supply chain for quality output. The low adoption rate signals pragmatism, not lag. The system endures through adaptability, not rash tests – restraint is its power. Otherwise, production hell returns in a flash.

About the author: Silvio Gerlach, an economist with a sharp, systemic lens, breaks down the logic of how AI reshapes business, society, and our daily lives. Discover what AI means for us – and how to deal with the new realities. Subscribe now and cut through the hype.