SYS:// web_production > content_ops > ai_automation

I'm the person people ping when things don't work. (Usually Friday, 4:55pm.)

Twenty years at Microsoft and XBOX, keeping things running smoothly. Lately I've been redoing them to run themselves.

Selected work

Manual processes, turned into systems.

Ten projects from XBOX, Microsoft, an early-stage AI startup, and what I'm building for myself. Pick one and watch the before become the after.

Localization PipelineXBOX app for PC

3 languages

became

30 languages

Every surface had its own CMS, its own rules, and a lot of copy-and-paste.

My tooling generated localized content per surface from its own source and cut production turnaround from 30 minutes to 3.Case study › Localization Pipeline

Result30 min → 3 min · 122 locale variants

Hello. I'm Jon.

Jon Bostwick, smiling outdoors above green hills and the ocean in Maui

I've spent 20+ years as a web producer, mostly at Microsoft and most recently at XBOX, working across XBOX campaigns, player support, and store operations.

For the last few years, the most interesting part of my job has been watching AI change how I work. I've gone from hand-building websites and managing CMS publishing pipelines to building LLM agents, automations, and personal tools.

Sometimes it's simple. Find a repetitive action, define it, and then seek out solutions.

Before tech, I used to teach, and it's never really stopped being a part of me. I'm looking for a team rethinking how content ops works.

More about me

On the workbench

Tools I build for myself.

Personal builds. Each one started as a question — why not try this?

  • The job board dashboard in its default dark theme, a ranked list of job cards, each with a fit score, per-axis score meters, pay and dates, under a header with search, sort and a min-fit slider beside the band counts: Strong 60, Good 144, Maybe 106, New 27

    Personal Job Board

    multiple sites became one ranked job list

    A serious job search means watching a dozen-plus sources daily and reading fifty postings to find three that match my interests. Get AI to find your favorites for you!

  • The Doom Board in its fictional demo scenario, a dark ship-console dashboard under the tagline "The world is probably not ending today. But let's check." A purple advisory reads Condition Purple, Uncomfortably Coupled, above Mr. Doom's narrated update and panels for risk sectors, what changed, and what is preventing the worst case

    The Doom Board

    doomscrolling became doomreading

    A local AI model reads a dozen public news feeds, links every claim to its report and sorts the world into seven risk sectors. Mr. Doom, its sardonic narrator, brings the doom. Shown here is the fictional demo.

  • A weather report titled "The summer got longer. The drizzle stayed." over a row of 78 colored stripes, one per summer from 1948 to 2025, shading from blue for cool summers to deep red for hot ones, with the last decade almost solid red

    Seattle's Still Wet

    a hunch about summer became 78 years of data

    A friend asked if the Seattle weather was sunnier. While I thought so, I had AI research publicly available data and create reports. One dataset, six different interpretations, from a boring report to a Korean cartoon.

  • The Meeting Transcriber page, white and yellow headline type on cobalt blue. A black vinyl record labelled Drop it here takes the recording; beside it are empty fields for the meeting name, names and terms, people speaking and language, and status lights for speaker labels and LM Studio

    Meeting Transcriber

    a meeting recording became who said what

    Drop a recording on the record and get back a Markdown transcript with speaker labels, plus templatized minutes from a local model. For privacy and to save money, it all runs on my laptop, so the audio never leaves it.

A closer look · the job board

A job board that ranks itself.

One command, one ranked dashboard. It runs locally, just for me.

Read the case study

What I do

Four things I do, each with its proof.

Automate content pipelines

Find the repetitive work everyone tolerates and make it easier: publishing, localization, review.

Toolkit: CMS architecture, publishing pipelines, localization ops, multilingual delivery.

The localization case study

Build AI agents & tools

Copilot agents, workflow automations, local-LLM tools. Small systems that catch errors early and do the boring bits.

Toolkit: M365 Copilot agents, Power Automate, GitHub Copilot, LLM tooling, MCP. PowerShell, Python and C#, AI-assisted. Azure DevOps.

The Teams alerts case study

Teach teams to use AI

Former teacher. Programs and sessions that turn AI curiosity into shared team practice, and teammates into people ready for their next thing.

The AI Explorers case study

Ship web content at scale

Web production end to end: CMS architecture, publishing pipelines, campaign launches, front-end care. Twenty years at XBOX and Microsoft.

Toolkit: 20+ years: CMS publishing, campaign production, accessibility, cross-browser QA.

My background
The XBOX app for PC Store tab: featured sale banners, a Game Pass promotion, and rows of game tiles
The XBOX app for PC. Every campaign placement shipped in 30 languages and 122 locale variants.

How I build

Build it ugly, then go wild.

My process is less a fixed sequence than a loop. Roughly:

  1. Brainstorm

    Talk the idea through with an LLM. Years in product development help most with the back-end requirements.

  2. Prototype

    Ugly is fine. Within a day, maybe even one prompt, something is running.

  3. Build

    Build it until it works. Working is the starting line, not the finish.

  4. Iterate

    Ask what would add real value, rank the list, fix, and ask again. Some projects, for months.

  5. Design

    It's been happening all along, but this is where I go wild.

  6. Check it

    Measure it against things it can't see. A tool that ranks my job search needs checking.

Read “Build it ugly”

Outside the work

Colorful places, and very good dogs.

I'm based in Seattle. I've been bumbling along in kung fu and tai chi for several years, I love squash (the sport and fruit), and I travel to colorful places like Curaçao.

The Seattle waterfront at night — the Great Wheel lit up over Elliott Bay, boats in the marina, downtown towers glowing
Seattle waterfront Home.
An aerial view of Seattle at summer sunset — Elliott Bay and the Olympic Mountains under a burning orange sky, the Space Needle at right
Seattle in the summer Best summers in the world.
The Willemstad waterfront at golden hour, the whole pastel row glowing under a low, warm sun.
Willemstad, Curaçao The inspiration for the site’s palette, on location.
A vivid red-and-pink summer sunset over a silhouette of evergreen trees near Bellevue
Bellevue summer The sky showing off, or an alien invasion?
A wine-red pond winding through green brush at Keālia Pond, Maui — water turned deep magenta by halobacteria
Keālia Pond, Maui Wine-dark, regrettably not wine.
Emma, a tan dog with big floppy ears and a soulful, open-mouthed grin, lit by sun
Emma My ball-obsessed soulmate.
The pastel row of Willemstad, Curaçao — the Handelskade waterfront — framed through a stone fort window, a palm frond leaning in, turquoise sea beyond.
The Handelskade Framed through a fort window.
The Willemstad waterfront at night, framed through a stone fort window, the lit Handelskade facades reflected in the harbour
Curaçao after dark The waterfront, lit up.

Open to work

A future forward thinking producer available???

I'm looking for Web Production, Content Ops and AI automation roles: contract or full-time, Seattle or remote-US.