Shift Handover, Work Instructions, Downtime: AI in Production and Maintenance

Only half of what the night shift noticed reaches the early shift. The work instruction at the machine is three years old, reality has long since moved on, and the same fault occurs for the fourth time without anyone having spotted the pattern in the logs. When a project ends, what was learned disperses along with the team.

Production and maintenance were not among the options in our survey, as we asked about administrative processes. Even so, the four applications here are office work, except that it happens at the machine: logging, updating, analysing, recording.

Where the time goes

In the reference company with a hundred employees, the four activities account for 31 hours a week. In hours, this is one of the smaller groups in the library, and it contains the one application that does not pay off through time savings at all. We included it anyway, and we explain why.

Four applications, costed

Logging the shift handover

Notes, voice recordings and machine data are turned into a structured shift handover report. A licence, no infrastructure, and the early shift knows what happened overnight.

In the model company: 12 hours a week, 40 percent of which can be automated, €14,040 a year. Setup €4,000, running costs €360 a month, payback after 4.9 months. Implementation: licence and training, no infrastructure of your own, 2 to 4 weeks. Tools: Microsoft 365 Copilot, Whisper, n8n. Data protection: not critical, internal data with no personal data involved.

Keeping work instructions up to date

The tool detects discrepancies between the instruction and the feedback from production and suggests updates. Mediocre in the numbers, but in practice the application that takes the sting out of audits.

In the model company: 8 hours a week, 35 percent of which can be automated, €8,190 a year. Setup €3,500, running costs €345 a month, payback after 10.4 months. Implementation: builds on the shared knowledge platform, 3 to 6 weeks. Tools: SwipeGuide, Claude, Azure AI Search. Data protection: not critical, internal data with no personal data involved.

Analysing downtime

Machine and fault logs are analysed, and recurring causes are named in plain language. This is a project that needs a connection to the machine data, and it does not pay off through the analysis time saved. It pays off through hours of downtime avoided, and those depend on your machine fleet, not on our model. That is why it appears in the library without a payback figure.

In the model company: 6 hours a week, 30 percent of which can be automated, €6,885 a year. Setup €35,000, running costs €635 a month, no payback from time savings. Implementation: dedicated project with an interface to ERP or machine data, 3 to 6 months. Tools: Azure AI, AWS Bedrock, Grafana, FORGE Insights (our own product). Data protection: needs review, settle the data processing agreement and deletion policy in advance.

Capturing lessons learned

The insights are extracted from minutes and changes and filed so that people can find them. Small, unspectacular, and after three projects the reason why the fourth runs more smoothly.

In the model company: 5 hours a week, 40 percent of which can be automated, €7,650 a year. Setup €4,000, running costs €180 a month, payback after 8.7 months. Implementation: licence and training, no infrastructure of your own, 2 to 4 weeks. Tools: Claude, Onyx, Microsoft 365 Copilot. Data protection: not critical, internal data with no personal data involved.

Where to start

Shift handover first: it is up and running in four weeks and paid back in five months. Work instructions and lessons learned are maintenance rather than a project, with the work instructions as an add-on to the knowledge platform. Downtime analysis only if you know what your downtime costs. If you do not know, measure exactly that first. Then the project either pays off at a glance or not at all.

Application	Hours per week	Savings per year	Setup	Payback
Logging the shift handover	12	€14,040	€4,000	4.9 months
Capturing lessons learned	5	€7,650	€4,000	8.7 months
Keeping work instructions up to date	8	€8,190	€3,500	10.4 months
Analysing downtime	6	€6,885	€35,000	no payback from time savings

How the figures are calculated

All figures come from the calculation model of the use case library, not from the survey. Reference company with 100 employees, 45 productive weeks a year, fully loaded costs of €65 an hour for commercial staff and €85 for technical staff, licences at €45 per user per month. The model counts only working time saved against the full costs, and the savings rates are deliberately set at the lower end of what is known from projects and studies. No quality gains, no avoided cost of errors, no effects on lead time. Replace the assumptions with your own figures and the model calculation becomes your calculation.

Further reading

The full calculation for all thirty applications is in the Use Case Library for Machinery Manufacturing, and the managing directors’ answers are in the survey results (in German).

Where is your biggest lever? Let’s talk it through briefly, no obligation.