Where AI helps
| Work | What has been shown or reported | Where it comes from |
|---|---|---|
| Finding the cause of scrap | At a Hitachi Energy chip plant, an AI model pointed to two process steps that were spoiling chips. A controlled test cut yield loss by 21.7%. On a second product it fell by 51.3%. | Hitachi Energy case. A peer-reviewed field study. |
| Visual inspection | At a Siemens electronics factory, inspectors who saw where the AI had spotted a problem caught 93% of defective boards. Inspectors given only the AI’s score caught 82%, fewer than the AI alone. | Siemens case. A peer-reviewed experiment with 48 factory workers. |
| Quoting and costing | At Herges, a steel fabricator, an AI model’s estimates of production time were five times closer to the real figure than those of experienced staff. EIST, a six-person metalworking shop, reports quotes in about 10 minutes that used to take up to two hours. | Herges case, measured by a public support centre. EIST case, the owner’s own estimate. |
| Product development | At Procter & Gamble, one person with AI produced product ideas as good as a two-person team without it, in about 16% less time. | Procter & Gamble case. An experiment, not yet peer-reviewed. |
| Sales, marketing and administration | EU manufacturers that use AI apply it most often to marketing and sales (30%) and to business administration (27%). | Eurostat, official statistics |
| Engineering software | Siemens added AI agents to its engineering assistant in 2025. One customer, thyssenkrupp Automation Engineering, reports better code quality and faster development. No figures were given. | Company announcement |
| Maintenance | Rolls-Royce says digital models of its aircraft engines extended the time between maintenance for some engines by up to 50%. | The company’s account, in a 2021 press report |
| Keeping production running in a crisis | Among 237 German manufacturers, those more ready for AI coped better with the Covid-19 lockdowns. The effect appeared only above a certain level of readiness. | Peer-reviewed survey study (Lerch and others, 2022) |
No controlled test yet for maintenance and production staff. No published controlled study from 2025 or 2026 has tested generative AI with maintenance or production workers. The figures in the table on maintenance and engineering are the companies’ own.
Inspectors’ own skill can fade. In a study at four endoscopy centres in Poland, doctors found pre-cancerous growths in 28.4% of colonoscopies done without AI before AI was introduced. Three months after it was introduced, the rate without AI was 22.4%. Staff who inspect alongside AI every day may lose the skill they need when the AI fails. Test inspectors without the AI now and then. See the Siemens case.
Size matters less than skills. In Germany, 20% of small and medium-sized firms used AI in 2022–2024, from 19% of firms with fewer than five employees to 36% of firms with 50 or more. Once the state development bank KfW took account of graduates employed, innovation, research and a digital strategy, size and sector mattered little.
A plan is not a result. In 2025 Volkswagen said it would invest up to €1 billion in AI by 2030 and expected savings of up to €4 billion by 2035. See what does not count as evidence.
Maturity scores are common here. Researchers have published AI maturity models for smart manufacturing and for the process industry. A commercial index for industry, AIMRI, is run by a paid, certified assessor and is not a self-assessment. A score describes where you stand. It has not been shown to predict results. See do not treat a maturity score as a forecast.
Risks and rules
- AI that keeps a machine safe. In the EU, the AI Act classes an AI system as high-risk when it is a safety component of a product covered by listed EU product-safety laws, and that product must be checked by an outside body before sale. A safety component is a part whose job is to keep the product safe. This classification applies from 2 August 2028.
- Machinery is handled through its own law. A 2026 amendment moved the Machinery Regulation to a separate part of that list. For products in that part, most of the AI Act’s own requirements do not apply directly. The requirements come through the product law instead. Ask your machine supplier which rules apply to your equipment.
- AI that only improves performance. The amended Act says AI used only for things like performance optimisation, automation or quality control is not a safety component. There is one exception: if its failure would endanger health and safety, it counts as one.
- AI that manages workers. In the EU, AI used to allocate tasks based on a person’s behaviour or traits, or to monitor and evaluate workers’ performance, is on the high-risk list. So is AI used in recruitment. These rules apply from 2 December 2027. The Act allows some exceptions, for example for narrow procedural tasks. In a survey of workers across the EU by the European Commission’s Joint Research Centre, 24% had their work schedules set by an algorithm. Where such a system set the pace of the work, workers reported higher stress.
- People still do the physical work. AI reaches office and engineering tasks first. Map tasks before you set targets. See map tasks before deciding what to automate and the worked example below.
Practices for this industry
All 30 practices in the library apply in this industry. None of them is specific to one sector. For the ones below, a study, a company case or a worked example from this sector shows why they matter here.
- Map tasks before deciding what to automate The worked example maps 375 task families at a 118-person manufacturer. About a quarter of the work is tied to a site or to doing something by hand.
- Budget for the work AI creates The worked example on a 118-person manufacturer counts 132 new checking and logging tasks.
- For each kind of task, test whether a person with AI beats either one alone At Procter & Gamble, one person with AI did about as well on product ideas as a pair without it.
- Do not treat a maturity score as a forecast Manufacturing has many AI maturity models. A score describes where you stand. It has not been shown to predict results.
- Make the AI show what its answer is based on Tested on circuit-board inspection at a Siemens factory. A steel fabricator shows the ten most similar past jobs beside every AI time estimate.
- Put your own records in order before you ask AI to predict from them Two metalworking firms had to clean or collect their production records before AI could estimate times and prices for one-off jobs.
Pick your company size
Size changes what is practical more than sector does. Each link opens the library filtered to that size.
Company cases
Each case is a named company with numbers and a public source. All cases.
- EIST: a six-person metalworking shop cut quoting time from two hours to ten minutes A metalworking shop with five employees and an owner used AI to price custom parts from 3D files and its own production history. A quote that could take two hours now takes about ten minutes. Most of the project was spent cleaning the firm’s data, and the owner still checks every quote.
- Herges: an AI model estimated production times five times more accurately than the firm’s experts A family-owned steel fabricator prices one-off jobs from its experts’ estimates of how long the work will take. A public support centre helped it record real times and train an AI model on them. The model’s estimates were five times closer to the real figure than the experts’ own.
- Hitachi Energy: an AI model pointed to the two production steps that were spoiling chips Researchers trained an AI model on the production records of a chip plant and asked it which process steps did most to spoil chips. Engineers tested its two suggestions on a live batch, and losses fell by 21.7%. On a second product the same method cut losses by 51.3%. The plant’s usual analysis had missed both causes.
- META-Regalbau: sensors and AI analysis of order picking paid back in about a year A shelving manufacturer with about 300 employees had staff wear small sensors so that AI could analyse how orders are picked in its warehouse. The findings shaped the layout of a new warehouse and save about €22,000 a year. Managers, the works council and all staff were told the purpose from the start.
- Procter & Gamble: one person with AI matched a two-person team without it Researchers ran an experiment with 776 Procter & Gamble staff on real product-development tasks. People working alone with AI produced ideas rated as highly as those of two-person teams without AI, and did it faster.
- Siemens: inspectors caught more defects when the AI showed what it had seen Researchers gave 48 inspectors at a Siemens electronics factory an AI aid for checking circuit boards. Half saw only the AI’s score. Half also saw a picture marking where the AI had found a problem. The second group caught more defects, and most of them did better than the AI working alone.
Studies from this sector
These studies were run on work that is typical of this sector.
Worked example
This example applies a task-mapping method to a made-up company that designs, assembles, installs and services industrial monitoring systems.
Sources
- Eurostat. Use of artificial intelligence in enterprises, 2025 data. ec.europa.eu
- Siemens (12 May 2025). Siemens introduces AI agents for industrial automation. Company announcement. businesswire.com
- CIO (2021). Rolls-Royce turns to digital twins to improve jet engine efficiency. A press report of the company’s account. cio.com
- Lerch, Heimberger, Jäger, Horvat and Schultmann (2022). AI-readiness and production resilience: empirical evidence from German manufacturing in times of the Covid-19 pandemic. International Journal of Production Research. doi.org
- Volkswagen Group (September 2025). Boosting innovation, reshaping mobility: Volkswagen Group invests in AI. Company announcement. volkswagen-group.com
- Chen, Liu, Xing, Peng and Yang (2021). Establishment of a maturity model to assess the development of industrial AI in smart manufacturing. Journal of Enterprise Information Management. doi.org
- Fornasiero, Kiebler, Falsafi and Sardesai (2024). Proposing a maturity model for assessing Artificial Intelligence and Big data in the process industry. International Journal of Production Research. doi.org
- INCIT. AIMRI: Industrial AI Maturity Readiness Index. incit.org
- European Union. AI Act, Regulation (EU) 2024/1689, Article 6 (classification rules for high-risk AI systems), as amended in 2026. Text on the European Commission’s AI Act Service Desk. ai-act-service-desk.ec.europa.eu
- European Union. AI Act, Article 2 (scope) and Annex I (list of EU product-safety laws), as amended in 2026. ai-act-service-desk.ec.europa.eu
- European Union. AI Act, Annex III (high-risk uses, including employment and worker management). ai-act-service-desk.ec.europa.eu
- European Union. AI Act, Article 113 (dates of application), as amended in 2026. ai-act-service-desk.ec.europa.eu
- European Union. Regulation (EU) 2023/1230 on machinery. eur-lex.europa.eu
- European Commission Joint Research Centre (2025). Digital monitoring, algorithmic management and the platformisation of work in the EU: Data from the AIM-WORK survey. Science for Policy brief. publications.jrc.ec.europa.eu
- Mariscal de Gante Martin and Villani (2026). Algorithmic management and working conditions in Europe. Evidence from the AIM-WORK Survey. JRC Working Paper 2026/05. publications.jrc.ec.europa.eu
- Zimmermann (February 2026). Einsatz von Künstlicher Intelligenz vor allem in Unternehmen mit hohen Innovations- und Digitalisierungsaktivitäten. KfW Research, Fokus Volkswirtschaft Nr. 533. kfw.de