How we work.
Our methodology is not a brand. It is what is left after nine years of projects in production - and we revise it as soon as a project shows us that a step is missing.
Four phases, clearly separated.
The same four phases the homepage covers in a sentence - here with what actually happens in each one, and with the gate where we stop if it does not hold up.
Taking stock (2-4 weeks)
On site. We look at what actually exists: operating manuals, service reports, drawing archives, bills of materials, exports from ERP, PLM and DMS. In parallel we measure the baseline - search time, first-time fix rate, the share of cases resolved remotely. Measured, not estimated.
Outcome: Data audit · Use-case matrix · Effort estimate
First case in live operation (6-10 weeks)
One machine type, one service team, a clearly bounded space of questions. Before the first line of code there is a question set drawn from what your people actually ask - we build and measure against it from the first week, not at the final presentation.
Outcome: Running system · Metrics · Risk report
Live operation (3-6 months)
Connected to ERP, PLM, DMS and the service portal. Permissions are checked against the source system at query time - nobody sees more through the system than they would in the original. Logging, monitoring and keeping new documents in sync are part of it.
Outcome: Integration · Access control · Monitoring
Handover (4 weeks)
Your team operates and maintains the system itself: runbooks, training, a named owner and a process for new content and corrected answers. Without that process a knowledge system goes stale - and a system technicians have distrusted twice will not get a third use.
Outcome: Training · Runbook · Documentation
Things we never leave out.
Question set before prompt.
We define what good looks like before we build the first system. A project without a question set from the shop floor is a project without a notion of truth - you end up debating impressions instead of quality.
No answer without a source.
Every answer shows document, section and revision, checkable in seconds. Where the evidence is thin the system says so - a plausible wrong answer costs more in service than no answer at all.
A model is a dependency.
We version models like code, write regression tests and plan for replacement. Swapping models after go-live is a configuration change, not a new project.
Want to go deeper?
30 minutes: we walk your current plans through the phase model.