It's nice to meet you

I help a small NFP work smarter than its size.

I'm Katie. I look after everything operational and technical that isn't frontline work at a not-for-profit domestic violence service. I take the things that have quietly become a mess and give them a structure that holds.

Connect on LinkedIn See my work ↓
Katie Morrison smiling in a garden, wearing a dark green jumper

About

The generalist holding it together

I'm the operational and technology backbone of a not-for-profit domestic violence service. I sit between leadership, frontline staff and vendors, and my job is to make sure the organisation around the frontline simply works: the systems, the sites, the contracts, the tools.

One day that means automating how a remote team books emergency accommodation. The next it's rolling out AI across the organisation, or redesigning facilities maintenance processes. I'm not a specialist who couldn't choose. I'm the generalist holding it all together, and in a small NFP that's where one person can have an outsized impact.

Every dollar and hour I save is resource returned to the mission. I think of the work as stewardship, not admin.

Selected work

A few things I've built

01
AISystems

An AI operating system for my working week

My week was scattered across notes, inboxes and memory, and things were starting to slip. So I designed a system that runs it for me: plain-text notes in Obsidian plus a scheduled AI assistant that plans my mornings, closes out my days, and keeps track of every commitment I've made.

Nothing falls through the cracks any more. I wrote it up as a full case study, and I'm now replicating the system for our finance team.

02
ToolsCost saving

Digital staff ID cards, built free in Canva

A rebrand and a round of staff turnover meant we needed new ID cards, but printing comes with minimums, costs and delays, and our turnover is too low to fill a batch quickly. I proposed building them in Canva instead. An external agency refined the design, and I rolled it out.

No printing costs and no waiting. The design stays in our hands, easy to adapt, and a new starter can have their ID on day one.

03
AutomationFinancial control

A spend tracker that changed the conversation

Nobody was tracking spend against one contractor, so overspend only surfaced after the fact. I built a tool that reads each invoice as you drop it in, pulls out the site and the amounts, and keeps a running total. One file, run locally, so no financial data leaves us.

I used it to flag July's overspend, and for the first time we had a constructive conversation about it. Call outs now need approval before they happen.

Have a go. Tap an invoice to send it across to the tracker.

Invoices (4)
Spend tracker
Drop an invoice here
Total spend
$0.00
Invoices
0
Site A$0.00
Site B$0.00
Site C$0.00
InvoiceSiteAmount
Nothing read yet

A demonstration with invented invoices. The real one reads the actual PDFs and never leaves my machine.

04
InfrastructureCoordination

Deputy on a three-site network upgrade

My manager set this project up and handles the quotes and negotiation. I'm her deputy on it: on site with the auditor, coordinating staff across the three sites, and the day to day contact for the vendor. The delicate part has been walking our incumbent IT provider through the handover.

Three sites moving onto infrastructure the service can rely on. My job is making sure the detail on the ground matches the plan on paper.

AI, put to work

The IT translator, now fluent in AI

Small charities don't get tech teams, so I became one. I've always been the translator between technical and non-technical people, and AI has become the most powerful thing I translate. I use it to run an organisation: an AI assistant that manages my working week, an org-wide rollout with a hub, training and monthly coaching, and systems I'm now building for other teams.

And I use it for the joy of it. Some of what I build with AI has nothing to do with work at all, which is exactly why I trust it at work. When you build things for fun, you learn what these tools are actually good at.

For fun

Things I built because I wanted to

The Idea Collider

A machine for having better ideas. It smashes together a niche audience, a hardware signal, an odd output medium and one awkward constraint, then makes you argue for the weird result before you're allowed to judge it. I built it with AI because my brainstorms kept landing on predictable ideas. They don't any more.

Roll a collision

A voicemail line for family stories

Call a phone number, leave a memory, and it appears on a website that collects the lot. Built so the least technical people in my family could preserve their stories through the one interface everyone already knows: leaving a voicemail.

See how it works

HOULP! looks after my house

A home maintenance app I built as a single HTML file. It knows every job the house needs and when, and it works entirely offline. Now I'm giving it a body: a palm-sized hardware companion, a tamagotchi for the house, that droops when the gutters are overdue.

Meet it

My Gym Coach

A pocket workout tracker built with AI in a single HTML file: my whole program with instructions, personal bests, a streak dot per exercise, and Pebble, the accountability pet rock. It nags me exactly as much as a rock can.

Train with it

Blob Runner, Game Boy edition

An endless runner in the original Game Boy's four shades of green: one blob, one button, and obstacles drawn pixel by pixel. Made with AI for no reason except that games are fun. Tap or press space to jump.

Play it

The pixel sprite lab

The game needed obstacles, so I built the tool that draws them: a tiny coding playground where rectangles become Game Boy sprites, in the same four greens. Change the numbers and watch the mushroom change.

Draw something

Writing

Notes on making things work

I write about operations, technology and AI in the not-for-profit sector. This collection will grow; here's the first piece.

How Claude and Obsidian run my working week

A case study: building an AI-assisted operating system from plain-text notes and a scheduled assistant.

+

Draft adapted from Katie's case study. Replace with the full write-up.

The problem

My role is deliberately broad: operations, technology, vendors, facilities, whatever needs a structure. The cost of breadth is that everything competes for the same attention, and the failure mode is quiet. Nothing dramatic goes wrong; things just slip.

What I built

The foundation is plain-text notes in Obsidian: a daily note, a commitments register, and a file for each initiative I'm running. On top of that sits a scheduled AI assistant. Each morning it builds my daily note: the decisions I need to make, the delegations to chase, and a short prioritised plan. Each evening it reconciles the day against my commitments register, updates the initiative files, and writes a debrief.

Why plain text

Because it holds. Plain-text files don't depend on a vendor, a subscription or an export tool. The AI does the moving parts; the notes stay simple enough that the whole system is legible to me at a glance.

What happened

Nothing falls through the cracks any more, and the time I used to spend reconstructing my week now goes into the work itself. The strongest signal it works: I'm now replicating the system for our finance team.

More pieces coming soon, here and on LinkedIn.

Let's talk

I'm always looking to build my network so if you think we'd have interesting conversations, I'd love to hear from you.

Connect on LinkedIn