Field experience · case study

Industrial commissioning automation

I turned the commissioning paperwork of a large industrial project into software.

As the data specialist on a large minerals-processing expansion in its commissioning and handover phase, my day was repetitive, high-stakes data work: pulling records, validating them, formatting reports, signing off readiness packages. So I built a toolkit of automations to do it. Across the project, that toolkit is projected to remove on the order of 1,000+ hours of manual work. That figure is an estimate, projected to project end.

Large minerals-processing expansion · commissioning & handover

1,000+hours saved(estimate)
1,000+hours

of repetitive manual work projected to be removed over the project

Estimate, projected to project end. The project is still running.

By the numbers

Flagship

Energization documents (NoE), the single biggest win

By hand

~6 hrs/day

With the tool

under 1 hr

Saved

~5 hrs/day

Estimate from real before/after. The full story is in the NoE sub-case below.

~1,300+

subsystems on the project

~14,000

equipment tags

~13,000+

checksheets

The layer that ties it together

An AI assistant that runs the whole toolkit

The automations above started as discrete scripts. The layer that turns them into one system is an AI assistant I built on top of them.

It delegates work to background agents, runs the tools on command, and keeps a versioned memory that travels across machines. Instead of remembering which script to run and how, I talk to one assistant and it orchestrates the rest.

  • Delegates bigger jobs to background agents and merges the results.
  • Runs the discrete automations as a single, callable toolkit.
  • Keeps a version-controlled memory shared across machines.
runreportsyncsignone assistant, one toolkit
The toolkit

Thirteen automations, doing the repetitive work

The four that did the most heavy lifting, then the rest of the toolkit.

Readiness-certificate (RFCC) sign-off bot

Logs into the document register, downloads the handover docs, uploads files and metadata to the commissioning database, and signs the readiness certificates unattended.

Commissioning-progress dashboard refresh

One command pulls the data, runs the browser exports, syncs the readiness sheet, and refreshes the Power BI model.

Boundary-certificate (BIC) custody highlighter

Resolves every tag to its subsystem and shades the boundary document by custody: grey for commissioning, red for construction.

Overnight unattended orchestration

Runs the whole morning pipeline headless overnight and leaves the draft outputs ready for review.

Energization (NoE) linker

Bulk-links energized subsystems' tags to their commissioning numbers, hundreds of associations per run.

Checksheet field sync

Diffs an update file and syncs only the changed rows across roughly 13,000 checksheets.

Live readiness data pipeline

Pulls subsystem readiness from a shared sheet so there is one source of truth for custody.

Document-register (HOP) exporter

A browser bot filters the construction handover register and exports it to Excel.

Subsystem-readiness (RFWCC) reporter

Generates PDF and Excel readiness reports straight from the commissioning database API.

Daily report generator

Turns completed task-board cards into formatted Word reports for any date range.

Timesheet auto-fill

Reads the daily reports and fills the timesheet web app, hours and descriptions per day.

Task-board triage

Audits the board for missing labels, wrong lists, and overdue cards, and fixes them on confirmation.

Fast desktop PDF editor (rapid-pdf)

A fast Windows PDF editor for page management and markup, built for the large energization documents.

Go deeper

Two pieces, in depth

Two parts of this toolkit have their own full case study. They are sub-cases of the work above, not separate projects.

On the numbers

Every figure on this page is a conservative estimate from real before/after observation, and several are projected to the end of a project that is still running. They are presented as estimates, not measured fact, on purpose. The point is the shape of the work removed, not a precise hour count.

Real field work, taken off my hands by code.

The code stays private. The capability is the point. See the rest of what Lucas builds.

Back to projects