The right service answer — at the machine, in minutes

The manufacturer builds capital equipment with a 15–25 year service life and a growing service share. Its service team supports thousands of machines across multiple generations and variants — each with its own build year, its own variant, and its own software version.

At every machine, the technician searched five sources at once: manufacturer manuals in several versions, supplier component manuals, spare-parts lists in the ERP, past tickets, the wiki — 30 to 90 minutes of search time per call-out while the machine sat idle.

Fault codes were treacherous: the same code “E-23” meant different things across two controller generations. Half-right full-text matches led to the wrong repair procedure.

Knowledge of the older generations lived in the heads of experienced technicians — who are retiring. On top of that: multilingual supplier manuals and mobile technicians.

Every machine is given a curated identity — product line, variant, software version, location. All further knowledge (manuals, bulletins, tickets, spare parts) is bound to that identity, not to a full-text index.

The technician scans the nameplate; the machine is identified. For “fault E-23,” Piko returns the applicable procedure with the manual title, version, page, and bounding box of the cited passage — plus the matching spare part with its part number and availability.

The ERP, PDM, and helpdesk remain the systems of record. Piko links, searches, and answers, but changes nothing. Service technicians confirm the applicable procedures on the job; their corrections improve the mapping.

Pure search time per call-out dropped from around 60 to about 10 minutes — the machine is down for less time, and the customer waits less.

Dangerous cross-application errors — a procedure that applies to the similar-looking variant but not to this one — are prevented by the curated vocabulary.

Service engineering and design now learn from field data instead of anecdotes for the first time: the most common fault codes per product line, spare parts with above-average failure rates.