What AI knowledge systems actually deliver in mechanical and plant engineering: service, spare parts, documentation, regulation. With real figures, real costs and honest limits.
According to Aquant, one in seven on-site visits is avoidable and a third of all service requests are plain information questions. Which requests the hotline can resolve, where remote diagnostics ends physically, technically and legally, how to find the suitable cases in your own service reports - and when reducing visits is not worth it.
German machinery manufacturers earn 27 percent of their revenue from service at a multiple of the new-machine margin - and invest 7 percent in it. Why the potential goes unused, how a fifth of revenue becomes half the profit, what the installed base is worth, and when expanding service is not the first step.
More than 90 percent of service technicians search for information for at least half an hour a day; in machinery manufacturing one in eight searches for more than two hours. What that costs a service organisation with 100 technicians, why the knock-on costs exceed the search time, why PDF search is not enough - and when a knowledge system is still not worth it.
A good quarter of the employees in German mechanical and plant engineering are 55 or older, and 41 percent of machinery manufacturers expect to lose company-specific knowledge when older employees leave. What a service technician knows after 30 years, why manuals and handover meetings do not catch it - and how a living knowledge base grows out of real service cases.
Across the industry, 77 percent of service visits are resolved at the first visit; the top 20 percent reach 88. What the first-time fix rate really measures, why repeat visits happen, what they cost - and which six levers reduce them.
37 percent of companies expect to be affected by the AI Act as users, up from 23 percent a year earlier - machinery manufacturers are often providers as well. What the Digital Omnibus deferred in July 2026, which obligations attach to which role, and where a deployer turns into a provider unnoticed.
Since 11 September 2026 a 24-hour clock runs on actively exploited vulnerabilities - for every connected machine in the field. Only 29 percent of companies know what the CRA means for them, according to Bitkom. What machinery manufacturers must organise now as manufacturers and as NIS2 entities.
55 percent of German SMEs name reliable results as the top purchase criterion for AI - and Gartner predicts at least 30 percent of projects end after the pilot. The checklist service and IT leaders use to test a system before they buy it.
On 20 January 2027 the Machinery Regulation replaces the Machinery Directive - and allows purely digital instructions for use for the first time. The three conditions, where paper remains a right, and why the switch is more than compliance.
Around a quarter of revenue, but 30 to over 50 percent of profit: after-sales carries machinery manufacturing. Where AI demonstrably works in service, what it requires - and which 2026/27 deadlines set the pace.
Our MCP server brings Outlook into MCP clients - email, calendar, folders and rules over the Microsoft Graph API, with OAuth and deeply nested folder hierarchies. MIT licence.
sevdesk integration for AI agents - contacts, invoices, receipts, orders, bank accounts. Written entirely in TypeScript, under an MIT licence.
MOCO's time and project management as an MCP connection for AI agents - projects, activities, invoices, master data. Written entirely in TypeScript, under an MIT licence.
codestra has been around since 2017 - and a great deal has shifted in the last two and a half years. The new site is the consequence of that, and the beginning of something we did not have before: a place where we think out loud.
A 30-minute first conversation. We listen and tell you whether an approach like one of these fits your case.
