---
title: I Ran My Light Show Prep Like a Production Office
description: "A week of light show prep, run like an event production office: one lead, six-part briefs, an append-only log, and a check that reported dark lights as on."
image: https://willcurran.com/hubfs/will-curran-brand/light-show-prep-production-office-cover.png
---

[Skip to content](https://willcurran.com/blog/light-show-prep-production-office#main)

[Will Curran](https://willcurran.com/)

[Home](https://willcurran.com/) [Portfolio](https://willcurran.com/portfolio) [Press](https://willcurran.com/media) [Blog](https://willcurran.com/blog) [Favorites](https://willcurran.com/favorites) [Speaking](https://willcurran.com/speaking) [Consulting](https://willcurran.com/consult) 

[Hire Will](https://willcurran.com/hire)

[Writing](https://willcurran.com/blog) Events — I Ran My Light Show Prep Like a Production Office [Subscribe](https://willcurran.com/blog/subscribe)

Events

# I Ran My Light Show Prep Like a Production Office

For a small festival, I ran my light show prep like an event production office: one lead calling the show, six-part written briefs, and an append-only log that ended the week at 1,818 lines. One miss came from a check that reported a set of lights as on while they sat dark.

![](https://willcurran.com/hs-fs/hubfs/will-curran-brand/will-curran-avatar.jpg?width=44&height=44&name=will-curran-avatar.jpg)

Will Curran Nov 9, 2026 · 7 min read

<https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwillcurran.com%2Fblog%2Flight-show-prep-production-office>

On this page

![An abstract dark graphic styled like a call sheet: one marigold box labelled Lead, a single line to a box labelled Live show, and a row of smaller desk boxes whose dashed lines run only to the Lead, next to the title Light show prep, run like a production office.](https://willcurran.com/hs-fs/hubfs/will-curran-brand/light-show-prep-production-office-cover.png?width=1600&height=900&name=light-show-prep-production-office-cover.png)

For a small festival I help put on, I ran the light show prep the way I'd run a production office. One lead called the show. Everyone else worked from a written brief. A shared log recorded the steps. The crew was my assistant team, a set of AI agents, and by the end of the week the log had 1,818 lines and the project had 178 merged pull requests. One miss came from a check that reported a set of lights as "on" while they sat dark.

This post is about the prep, and about the structure I'd copy again.

## The production office is the part that transfers

I started an events company at 17 and ran it for sixteen years. What I carry from all those shows is the structure. One person calls the show. Call sheets say who is where and when. There is one sheet everyone trusts on show day, and nobody improvises on the live rig because they had a good idea in the moment.

A week of lighting prep has the same problem as a load-in. Lots of people want to touch the same thing, some changes can't be taken back, and the deadline doesn't move. So I set it up the way I know how.

![An abstract call sheet with three rows. Calls the show: one lead, which owns decisions, the timeline and every live change. On the rig: one operator at a time, working from a six-part brief. At the desk: project manager, research, show paperwork, docs and librarian, which propose changes for the lead to accept. A side panel lists what reaches me: decisions batched as multiple choice, anything that needs my hands, and corrections. Underneath, one shared log with one line per step.](https://willcurran.com/hs-fs/hubfs/will-curran-brand/light-show-prep-production-office-call-sheet.png?width=1440&height=1040&name=light-show-prep-production-office-call-sheet.png)

## One person calls the show

One lead owned the decisions, the timeline and every change to the live show. On the rig, the rule was one operator at a time.

Everyone else worked at the desk: a project manager, research, show-day paperwork, documentation and a librarian. Desk roles proposed changes as pull requests, and the lead decided what went in. The project manager collected their updates and sent the lead a digest about every 45 minutes.

Short jobs went to single-job workers. Each one got a brief, did that one job and stopped. The rule was one job per worker and never two workers on the same file.

## Every brief had six parts

On a show, a call sheet that just says "do lights" gets you nothing. The briefs here all had the same six parts:

1. **Goal.** One sentence.
2. **Current state, with numbers.** What exists right now, counted.
3. **Exact steps.**
4. **Stop points.** Marked plainly. At each one, the worker stops and waits.
5. **Rules.** What never to touch.
6. **A final report, with numbers.**

The operator's steps in the log carried a count before and a count after, so anyone could check the result without redoing the work. When a worker had to hand back in the middle of a long export, the lead could read its last line and pick up from exactly there.

## The log outlasted every shift

The log was one plain text file, one line per step:

```
- HH:MM [Role] what happened
```

These assistants lose detail when a long conversation gets summarised, and a fresh one starts from nothing. The tail of the log said exactly where things stood, so a summarised or brand-new worker could read the last screen and carry on.

Three smaller habits sat around it:

- **A task board and a milestones file.** The milestones file was the day's timeline, one line per item, edited in place to "DONE" or "SKIPPED".
- **Times come from the clock, never from memory.**
- **A rebuild page.** I asked for one page that takes a brand-new session from nothing to a running team.

For show day there was one source of truth: a booth card and a runsheet. When the shipped show files were final, they got a fingerprint.

## What reached me, and what didn't

A good production manager protects the producer's attention. The rule for what reached me was short:

- **Decisions only I could make,** batched as multiple choice with a recommendation.
- **Anything that needed my hands:** plugging in drives and power-cycling players.
- **Corrections,** when I'd been told something wrong.

Routine progress, merges, doc fixes and back-and-forth between desk roles were absorbed and logged.

Some calls were taste, and those stayed mine. The rig had some cheap derby effects next to much bigger lights, and when they showed up in early demos as a mess of mixed colours I gave the rule: one pure colour that matches the palette, or off. In my words at the time, "simpler is better", because it's "a cheap party light".

## The check that said the lights were on

Midweek I asked whether we could read the whole show's programming. The lead had a narrator tool built that reads each track's lighting script and describes it in plain words, and it was checked against my own run-through notes. It matched everywhere it could be compared.

It said the derbies were on for most of each track it checked. The scripts still said "on", and the narrator read that faithfully, but a change earlier in the week had cut the derbies off from the parts of the script that drove them. The narrator was reading dead lanes.

The call was to sit the derbies out of the show and fix them afterwards. The lesson for anyone copying this: before you trust a checker, show it a known failure and make sure it catches it.

## The parts that were not magic

- **A planned export would have deleted most of a drive's music.** It was caught from an earlier entry in the log. The export order changed.
- **A pre-check is a forecast.** One cleanup step was predicted to remove nothing. The after-check showed it had removed a few hundred files and copied them back. The final count matched, so no harm was done, but only the after-check proved it.
- **Sync stopped quietly.** The shared copy of the project stopped updating, and desk roles worked from stale files until someone noticed.
- **Something built from memory missed a piece.** A DJ kit was missing its loop files. The librarian caught it on a second read, and the kit was rebuilt the same evening.
- **Two operators, once.** A second operator got started while the first was still working. The fix is to check before you start one.
- **Some roles were waste.** Long-running roles that kept re-reporting the same progress were not worth it, and neither was polish work late in the week or chatter during exports.
- **I don't know what it cost.** Nobody tracked exact usage, so I can't give you a number.

## Where I was the bottleneck

Everything physical waited for me. Plugging in a drive and power-cycling a player are the moments where a mistake is expensive.

## If you are about to try this

- **Write the delegation down first.** Decide what the lead may decide alone, and what always comes back to you with a recommendation.
- **Give a checker a known-bad case before you trust it.** A check that has only ever seen good results is unverified.
- **Always run the after-check,** even when the pre-check says nothing will change.
- **Read the log before anything irreversible.** The near miss above was caught that way.
- **Start with a way to measure cost.** I didn't, and now I can't tell you.

## Build it yourself

**What it does:** runs a multi-day project with a team of assistant sessions without you managing every step.

**The parts:** an assistant tool that can start helper tasks, a shared repository, and plain text files: a log, a task board, a milestones file, a rebuild page and a show-day sheet.

**How it fits:** one lead session owns decisions and every live change. Single-job workers take a six-part brief and stop when done. Desk roles propose changes for the lead to accept. Every worker appends one line per step to the log. The lead checks in every 30 to 60 minutes, re-reads the board and the log tail, and gives idle workers their next job.

**Time and cost:** about a week here, under a fixed deadline. I didn't measure the cost, so set up a way to track usage before you start.

Give this to your AIHelp me run a multi-day project with a small team of assistant sessions, organised like an event production office. One lead session owns every decision, the timeline and every change to anything live. Workers each get one job from a brief with six parts: goal, current state with numbers, exact steps, clearly marked stop points, rules, and a final report with numbers. One worker per job, never two on the same file. Other roles work at the desk and only propose changes for the lead to accept. Keep state in plain files: an append-only log with one line per step in the form "- HH:MM \[Role\] what happened, with counts before and after", a task board, a milestones file, and one page that lets a brand-new session rebuild the team. Take times from the clock, never from memory. Only these reach me: decisions batched as multiple choice with a recommendation, physical tasks, and corrections. Before trusting any checker, test it on a known failure. Always run an after-check, and track usage from day one.

If you have run a project like this, with people or with software, I'd like to hear what you would change. Reply and tell me.

Filed under [Events](https://willcurran.com/blog/tag/events)

![](https://willcurran.com/hs-fs/hubfs/will-curran-brand/will-curran-avatar.jpg?width=64&height=64&name=will-curran-avatar.jpg)

Will Curran

I spent sixteen years building one company, and I have been in the event industry since 2007. These days I write about the technology I actually use — trucks, radios, home infrastructure, and the tools that survive past the first invoice.

[All posts by Will →](https://willcurran.com/blog/author/will-curran)

Get the next one by email

One email per post. No digest, no sponsors.

[← Previous Favorite Things: Claude, and How I Actually Use It](https://willcurran.com/blog/favorite-things-claude)

## Related reading

[Events Walk the Room: What Live Events Taught Me About Checks That Lie 4 min read](https://willcurran.com/blog/walk-the-room) [Events Taking an RV to a Festival That Bans Generators 5 min read](https://willcurran.com/blog/rv-festival-generator-ban) [Events We Brought a Weather Bot to a Festival. It Died on Day Three. 6 min read](https://willcurran.com/blog/festival-weather-bot)

[Will Curran](https://willcurran.com/)

[LinkedIn — 29,339+ followers](https://www.linkedin.com/in/willcurran) [GitHub](https://github.com/rancur) [YouTube](https://www.youtube.com/channel/UCJgscc27l4fpkWx1SeK05xg/) [Spotify](https://open.spotify.com/user/rancur) [SoundCloud DJ mixes](https://www.soundcloud.com/thewillsee) [Twitch](https://www.twitch.tv/thewillsee) [Last.fm](https://www.last.fm/user/default) [Trakt — movies & TV](https://trakt.tv/users/willcurran) [Goodreads](https://www.goodreads.com/user/show/55302838-will-curran) [Will See Music](https://www.thewillsee.com) [Arizona Meshtastic](https://azmsh.net) [Electric Forest Meshtastic](https://efmesh.com)

© 2026 Will Curran — All rights reserved. [RSS](https://willcurran.com/blog/rss.xml) [Back to top](https://willcurran.com/blog/light-show-prep-production-office#main)

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Will Curran",
    "url" : "https://willcurran.com/blog/author/will-curran"
  },
  "dateModified" : "2026-11-09T18:29:16.880Z",
  "datePublished" : "2026-11-09T18:29:16.000Z",
  "headline" : "I Ran My Light Show Prep Like a Production Office",
  "image" : [ "https://willcurran.com/hubfs/will-curran-brand/light-show-prep-production-office-cover.png" ],
  "mainEntityOfPage" : {
    "@id" : "https://willcurran.com/blog/light-show-prep-production-office",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://willcurran.com/hubfs/image-1.png"
    },
    "name" : "Will Curran LLC"
  }
}
```