AI in manufacturing SMEs: securing your experienced people's know-how, with ERP and MES rather than against them

The labour market is easing, but know-how is still getting scarce

The skills shortage index of the Adecco Group and the University of Zurich was around 22% lower in 2025 than the year before. Only four of 32 occupational groups still count as shortage occupations. Two of them affect manufacturing directly: engineering specialists and electricians.

The easing follows the economic cycle. Demographics continue: according to economiesuisse, from 2029 over 30,000 more people a year will reach retirement age than young people enter the workforce. Every retirement takes know-how that sits in no system: how to set up a machine for a particular material, how a fault announces itself, what a customer tacitly expects.

The bottleneck of the coming years is less the jobs than the know-how that leaves with experienced people.

Where AI helps in production

The most effective areas of use in manufacturing SMEs have one thing in common: they make the know-how of the best available to everyone, and the specialist keeps the decision.

Use	What the AI does	Impact
Knowledge assistant for faults and maintenance	Searches manuals, fault logs and maintenance history and suggests proven solutions with their source	Shorter downtime, new staff become independent sooner
Work preparation and programming	Suggests tools, parameters and sequences, learned from earlier jobs of the best specialists	Less time per job, more consistent quality
Quotes and costing	Finds similar earlier orders in the ERP and prepares a costing	Faster quotes, less dependence on individual estimators
Quality deviations	Evaluates inspection records and deviation reports and spots recurring patterns	Causes become visible earlier
Onboarding	Answers questions about routings and instructions in plain language	Experienced people are interrupted less

Our project with Hexagon shows what this looks like at scale: an AI assistant learns from the best experts and supports 30,000 technicians in selecting tools. That saves 700,000 working hours and CHF 4.2m a year. The pattern also works in a company with 80 employees, on a smaller scale.

ERP and MES: AI alongside, not instead

Many executive teams hesitate because they fear a new large IT project. That is not necessary. Five principles keep the effort small:

1. ERP and MES stay in the lead. Orders, routings and confirmations stay where they are today.
2. The AI only reads at first. It accesses the data through existing interfaces or exports and writes nothing back. A person still enters decisions in the system.
3. Start with the data you have. Order history, fault reports, inspection records and manuals are enough to begin. Perfect data is not a prerequisite.
4. Check what your ERP offers. Many vendors bring their own AI functions. What matters is whether they address your bottleneck.
5. Your know-how stays yours. Design data and customer drawings belong only in environments with a business contract and no training of the vendor’s models with your data.

Secure your experienced people’s know-how before it leaves

The most valuable step is often the simplest: capture what your most experienced people know while they are still there.

1. Identify key people. Who retires in the next five years, and whose know-how will then be missing?
2. Capture the know-how. Short conversations, annotated examples from daily work, explanations at the machine. The AI turns this into searchable instructions with source references.
3. Have the experienced people check it. They approve what is right. That builds trust and recognises their work.

Where Swiss SMEs stand today

According to the AXA study of September 2026, three out of four Swiss SMEs use AI in their daily work. 35% have deliberately built it into their processes. Its use for optimising work steps rose from 23% to 35% within a year. In the machinery, electrical and metal industry, AI-related skills are gaining importance mainly at technology leaders, according to BAK Economics, while the majority is still waiting. Those who start now build a lead.

Three steps to begin

1. Choose a bottleneck: Where do orders wait for a single person, where do faults take too long, where does onboarding cost the most?
2. A pilot with measurement: Depending on complexity, eight to twelve weeks, read-only access to ERP, MES and documents, and a baseline measurement before the start.
3. Scale what works: To further lines, sites or processes, with a short report to the executive team and the board.