A SMALL CONTENT WARNING / NO ID REQUIRED

AI can be
a bit gross.

This site contains AI. I used it to help make things, solve problems, and write some of this. I still had to check it.

Would you like the less-slop version?

Less copy. Actual work. Same human responsible.
Oops I AI’d AgainFIELD NOTES BY THOM HAMILTON
FOR BUSINESS OWNERS WITH ENOUGH ON THEIR PLATE

AI IS A LOT.
LET’S START WITH
YOUR PROBLEM.

Another tool. Another “you need to be using this.”
I’d rather start with the thing you keep doing twice.
I’m Thom. I’ve been figuring this out in my own businesses.

Find a familiar headache ↓
ACTUAL WORK.
HUMAN CHECKED.
REGINA, SASKATCHEWANREAL WORK. HUMAN JUDGMENT. A FEW DO-OVERS.01 / A PLACE TO START

01 — YOUR DAY, BEFORE THE AI TALK

You don’t need another thing to manage.

I’d start with a recurring frustration. Pick one that sounds like your week.

A SMALL PLACE TO START

See one real example →

FROM MY FILES. FROM MY WORKBENCH.

YES, THIS STUFF
BECAME REAL.

Here’s some of the work behind the explanations: a physical tag, artwork, a prepared letter, and the map that connects the steps. Open any piece for a closer look.

The physical KinTag.A printed tag, a programmed chip, and a record behind its URL. I still make, program, and check the physical result. The coin gives a sense of scale.
An idea with a personality.“Room for One More,” a Next to Kin colouring-book design from my files. AI-assisted artwork becomes a designed page that still needs my review.
A letter with a specific destination.A prepared museum outreach letter, written around the organization and the sample idea. This shows the document design; it doesn’t establish a send, partnership, or endorsement.
The connections, made visible.An overview from my 105-process working atlas. It records inputs, outputs, handoffs, evidence, and human decisions. It’s a reconstructed process map, not a live execution trace.
See why I rebuilt my own business site →

02 — LESS MYSTERY. MORE “OH, I GET IT.”

Follow one job.
All the way to the mailbox.

For Next to Kin, I connected the preparation of personalized sample packs. Here’s how the information moves, and where I stay involved.

BUILT + ACCEPTANCE TESTED Scheduled worker paused October 10, 2026. This is an explanation of the workflow, not a running automation. Shipment completion is unverified in the reviewed records.
RESEARCH → PREPARE → CHECK → MAKE → MAIL → LEARN

03 — HOW I ENDED UP HERE

My accidental journey to becoming your AI Bushwhacker.

I didn’t set out to become an AI expert.
I had businesses to run. Things to make.
And a growing list of “surely there’s a better way.”

I’m Thom Hamilton: a Métis business owner, filmmaker, maker, and founder of Next to Kin in Regina, Saskatchewan. I started using AI to solve practical problems. One experiment led to another, and the useful pieces started connecting.

I direct, connect, review, and refine work made with multiple AI tools. I still make the calls, check the results, and do plenty of the making.

ONE EXAMPLE YOU CAN GET YOUR HEAD AROUND

FOUR TOOLS.
ONE FRONT DOOR.

I run Whitetail Films as well as Next to Kin. Those subscriptions once made sense. Over time, they ballooned with features I didn’t need, and their prices grew faster than my bookings. I needed a setup that fit the business I actually had.

THE SEPARATE PIECES

SquarespaceMy public website
HoneyBookClient management & contracts
Pic-TimePhoto delivery
VidflowFilm delivery

THE CONNECTED BUILD

Whitetail
Films.

My website, project dashboard, contracts, planning portal, and photo + film delivery, brought into one site.

Visit Whitetail Films ↗BUILT · LIVE SITE · ADOPTION VARIES

THE ACTUAL INTERFACES

Here’s what I can open now.

Saved dashboard and delivery screenshots, plus the real portal interface rendered with sample data. Open each for a closer look.

My side of the work.

A saved October 13 dashboard view, cropped above client activity. The summary shows the state of the dashboard at capture, not my complete booking history.

CLIENT MANAGEMENT / HONEYBOOK ROLE
The couple’s shared plan.

The actual portal component and styles, rendered with sample data for this demonstration. No private client link, contact information, or contract is shown.

PLANNING PORTAL / HONEYBOOK ROLE
The finished work, delivered.

A saved delivery-verification screenshot showing the film player, photo controls, and downloads. The test collection is labelled as a verification view.

DELIVERY / PIC-TIME + VIDFLOW ROLES

I wasn’t trying to build more software. I was trying to stop paying for more than I needed.

That became the brief: bring the jobs I actually use into one connected site. Give couples one place to start, and give me a clearer view of the work behind it.

01 / THE CLIENT SIDE

A path for each couple.

A personal page connecting their wedding website and delivered films. Photo galleries and film-only delivery have their own presentation.

02 / MY SIDE

The job stays in view.

Leads, projects, payment records, to-dos, and planning information come together in a dashboard I helped shape around my work.

03 / MY ROLE

I directed. AI helped build.

I set the requirements and reviewed the result. Claude Code wrote the implementation, with Codex contributing a security review. I still own the decisions and the upkeep.

What’s built—and what I’d still check +
How the earlier tool roles map to the new build
Earlier toolCapability built into the siteWhat that doesn’t establish
SquarespacePublic website on Vercel; DNS migration records.Subscription cancellation or the old bill.
HoneyBookLeads, projects, contracts, payment tracking, to-dos, and a couple’s planning portal.Card processing or complete adoption across client work.
Pic-TimePhoto galleries with ordered sections, folder uploads, and reordering controls.That every old gallery has been migrated.
VidflowFilm cards, chosen thumbnails, a theatre player, and film-only delivery pages.That every film has been delivered through the new site.

The build record: 28 commits dated October 12–22, 2026 in my local project history. That is a development window, not a measurement of full-time work or proof of complete migration.

Underneath: Next.js on Vercel, Supabase for records, Cloudflare R2 for delivery assets, and Resend email helpers. Payment tracking is separate from payment processing.

The cost question: A fair savings comparison needs my actual old bills and current hosting, storage, and email usage. Owning the code still comes with maintenance and running costs.

Got a collection of tools that don’t quite talk? Start with that. →

04 — THINGS I ACTUALLY BUILT, USED, OR TRIED

A few things from the workbench.

Specific jobs, with honest labels. “Built” means code exists. “Used” means documented use. Neither promises today’s reliability.

Based on the process atlas and supporting records reviewed October 24, 2026. These are evidence labels, not a fresh audit of every live service.

05 — THE “OOPS” IS EARNED

Some of the useful work
was the do-over.

01 / REVISED

The rough cut isn’t the film.

I’ve fed rejected cuts and finished-film examples back into my wedding preparation rules. Generated timelines are editable starting points; I inspect the joins and finish the story, sound, music, and colour.

02 / PAUSED

Built doesn’t mean left running.

The sample-pack workflow was built and acceptance tested. Its scheduled worker was stopped October 10. Keeping that distinction visible matters more than making the diagram look impressive.

03 / PROTOTYPE

A good preview still needs a real test.

Midway Portraits and NFC instruction cards remain prototypes. A drawing preview or a generated file doesn’t tell me whether the physical product is ready for someone else.

06 — A GENTLE LOOK UNDER THE HOOD

The tools follow the job.

I break the work into steps, connect the information, and keep the owner involved where judgment matters. Some steps need AI. Others need ordinary, dependable software rules.

Information→Useful preparation→Defined action→A human decision
I’m curious about the technical bits +
AI
Helps interpret, research, draft, or generate. I check factual and creative outputs.
MCP
Gives an AI application a way to discover and use exposed tools. It doesn’t supply the business rules or guarantee correctness.
APIs
Let systems request specific actions, such as preparing a pack or updating a record.
SQL functions
Apply defined database rules, such as allocating unique tags and keeping related records consistent.
Webhooks & schedules
Start work when an event occurs or a set time arrives.
People
Approve, judge, manufacture, and handle exceptions. Every workflow doesn’t need every technical layer.

07 — START SMALL

Show me the thing
you keep doing twice.

Describe a recurring headache. My alchemy tool will offer a fresh angle and one small experiment to try. If it feels useful, we can talk.

Your description is sent to the AI service behind my KetchupSammy Rory tool. Leave out customer names, contact details, and confidential information. No email required.

A starting point to think with, not a promise of results.

Prefer a simple email? Write to me at Next to Kin ↗