About JY Solutions

Built from the field up.

15 years in construction. Still in it. Now building AI systems for the people who actually build things.

John York, founder of JY Solutions
John York
Founder, JY Solutions
The background

Construction first.
AI second.

John York has 15 years in construction — 10 of them in civil. He's not a project manager who moved to tech. He's a lead hand, still in the field, who sees every day where things break down: a bid that came in too tight because something got missed in the drawings, a fabricated piece that shows up wrong because the shop drawings weren't read right, work that has to be redone because the information upstream wasn't good enough.

That field-level view is what drives the builds. The problems aren't abstract — they're the same ones that show up on every job site, every week. Repeatable problems are automatable problems.

He started building AI systems to close those gaps — tools that help trade and service businesses stop losing time to the same avoidable mistakes. Five systems are running in production today, with clients in BC and Alberta.

15
Years in construction
10 years civil, field lead hand
5
Systems in production
Live and running, not demos
BC
BC & Alberta
Where current clients operate
The work

What this actually looks like

Not consulting decks. Not strategy sessions. Actual Python scripts, Claude agents, and Google Sheets integrations – built and handed over running.

No vendor lock-in

Everything is built open – Python scripts, plain JSON, standard APIs. You own the code. You can hand it to any developer. Nothing requires a SaaS subscription to keep running.

Human-in-the-loop by default

AI drafts, humans approve. Every system has a review step before anything touches a customer. You can loosen that over time as you build trust in the output – but you start with control.

Built to last, not to impress

No flashy demos that fall apart in production. Systems include error handling, checkpoint/resume logic, and rate limit management. They're designed to run Monday morning without anyone babysitting them.

Honest about what AI can't do

AI is good at classification, drafting, summarizing, and pattern recognition. It's bad at judgment calls that require real-world context. Systems are scoped to what AI actually does well.

The framework

How every system
gets built

Every build follows the WAT framework – Workflows, Agents, Tools. The same structure that makes it reliable, maintainable, and actually hand-off-able.

W / Workflows

The Instructions

Markdown SOPs that define what to do step-by-step. Written like you'd brief a competent teammate. When the system breaks, you read the workflow to understand why.

A / Agents

The Decision-Maker

Claude handles the reasoning – classification, drafting, summarizing, extracting. It reads the workflow, calls the right tools, and handles edge cases. Probabilistic where it should be.

T / Tools

The Execution

Python scripts that do the actual work – API calls, data transforms, Sheets writes, Slack pings. Deterministic, testable, and fast. If each step is 90% accurate, five chained AI steps = 59% success. Tools fix that.

Work together

If it's repeatable,
it can be automated.

Free 30-minute call. We figure out what to automate first, scope it honestly, and decide if it's a fit.

Book a free call See past builds

john@jysolutions.ca