Capability demonstration for Apex Water
AI-assisted engineering tools — accelerating DWSP authoring.
An interactive first-draft generator for New Zealand Drinking Water Safety Plans (DWSPs). Built as a proof of what an AI-assisted engineering workflow can produce — where research, tooling, and domain integration collapse from weeks into hours.
What this demonstrates
The tool has two axes of engineering leverage, both visible in the interactive preview:
Enter Cl₂ dose, contact time, temperature, pH — the tool computes CT and looks up delivered log-reduction from the USEPA SWTR / DWSNZ tables. UV validated dose + UVT gives delivered log per pathogen. Compared against the source-type-specific target with a traffic-light verdict.
The tool derives a candidate hazard register from the source type and catchment character, cross-checks whether the declared treatment chain has a matching barrier for each hazard, and flags gaps. The engineer refines instead of starting blank.
Interactive on the preview
- • Risk register (Section C) — likelihood + consequence + barrier status editable per hazard; add site-specific hazards; scores recompute live.
- • CCP register (Section D) — critical limits, monitoring, and corrective actions editable per treatment step.
- • Disinfection log-inactivation (Section D.1) — interactive Cl₂ + UV parameter panel with live pathogen-by-pathogen verdict.
- • All edits flow through to the printable PDF and the Markdown / JSON export.
Boundaries
- • Not a substitute for qualified engineering, legal, or regulatory review.
- • No integration with Hinekōrako, SCADA, laboratory LIMS, or CAD / P&ID sources.
- • Rule references and CT / UV tables are indicative — verify against DWQAR 2022 (current) or DWQAR 2026 (from 1 July 2027) before use.
- • Runs entirely in the browser — no project data leaves the client.
Why we built it this way
Consulting engineers draft DWSPs project-by-project in Word, using the Taumata Arowai template. The repetitive scaffolding — multi-barrier narrative, hazard tables, monitoring schedule, CT calculations — is a natural target for structured generation. A tool of this shape reduces first-draft effort by an estimated 60–80% and improves consistency of rule references across projects.
The rule / hazard / treatment / monitoring / disinfection catalogues are held as versioned data files, decoupled from the generator logic — so a rules revision (for example the DWQAR 2026 transition) becomes a data update rather than a code rewrite.