Skip to main content
Gothi Tech

3 articles

Field notes,
not thought leadership.

What we have learned putting models next to machines. Opinionated, specific, and occasionally uncomfortable.

  1. 01
    Compliance7 min read

    Compliance is a data model, not a folder structure

    Storing documents is not compliance. An auditable system binds each record to the specific regulatory clause it satisfies, so that evidence can be reconstructed rather than searched for.

    In short

    • A document management system answers 'where is the file'; a compliance engine answers 'which requirement does this satisfy'.
    • Audit readiness is a property of the schema, not of how well the folders are named.
    Read
  2. 02
    Industrial AI9 min read

    Why your industrial AI pilot never made it to the second line

    Most industrial AI pilots succeed technically and die commercially. The failure is almost never model accuracy — it is integration debt, unowned alerts, and a pilot scoped so narrowly it cannot generalise.

    In short

    • A pilot that does not write into the CMMS produces alerts nobody is accountable for.
    • Models trained on one line's sensor placement rarely transfer without recalibration.
    Read
  3. 03
    Predictive Maintenance8 min read

    What predictive maintenance actually costs, and what it actually saves

    The return on predictive maintenance comes from avoided unplanned downtime, not from reduced maintenance labour. Here is the arithmetic, and the three inputs that decide whether a deployment pays for itself.

    In short

    • Unplanned downtime typically costs 8–20× the direct cost of the repair itself.
    • Predictive maintenance ROI is driven by downtime avoided, not by maintenance headcount reduced.
    Read
Subscribe via RSS