A one-person surveyor built an AI operating system that works while he's away
A quantity surveyor turned AI into the system two practices run on, including a weekly task that scans the press and updates his pipeline unprompted.
James Brown runs two one-person practices in construction, and over a few months he quietly rebuilt how he works. Instead of one impressive demo, he built a daily AI system with skills, memory and a scheduled task that runs on its own. It is a grounded picture of what steady practice can turn into.
P
Philip Brown
Chartered quantity surveyor and Passivhaus consultant running two one-person practices.
Before
James runs James Brown Projects, a quantity surveying and cost consultancy, and Futurehaus Consulting, a Passivhaus consultancy. Both are one-person practices, so fee proposals, cost checks, valuation reviews, correspondence and school governance work all landed on one desk. Lessons from AI training risked fading a week after each class.
“I treat AI less like a search box and more like briefing a new graduate surveyor in their first week.”
In their own words
The turning point
Across a year of AI training, he changed how he thought about the tool. He began treating it less like a search box and more like briefing a new graduate surveyor in their first week. He also wrote a recall index of everything he learned, so each lesson built on the last.
“The tool does the grind and I own the output.”
In their own words
After
Today AI is part of his everyday work. He has standing instructions, a project space for each business, a growing library of skills, an editable memory and connections into Airtable and Microsoft 365. A weekly scheduled task scans the construction and AI press, checks ideas against his workflow maps and flags anything that suggests he is on the wrong path. He also built brand books and websites for both businesses, and is shaping the setup into something teachable through GroundMark.
What it lets people do
The system handles the grind of proposals, cost checks, design checks and correspondence, while he keeps the judgement and owns the output. Its background tasks keep new ideas and evidence flowing across his ventures even when he is not at his desk.
What they learned
For James, the habit matters more than the tools. A failed attempt is not a reason to drop a workflow, and the gains come from saving what works rather than waiting for a perfect system. He also trained the generic AI voice out of his setup so the results sound like him.
“The habit is the hard part, not the tools.”
In their own words
See the work
His custom skills library, from valuation checks to governanceMore skills, including brand voice and a Passivhaus checkerFurther skills for presenting, story and project building
Personal AI operating system was built during AI Mastery.
Read the full submission
This isn't a single project. It's a working habit. I run two one-person practices: James Brown Projects, a quantity surveying and cost consultancy (the professional who manages cost and contract on a building project), and Futurehaus Consulting, a Passivhaus consultancy. Over the last few months I've built AI into how I work every day, as an operating system rather than a novelty, and some of it now runs on its own without me lifting a finger. Most award entries show one impressive thing. This is the system a business actually runs on.
Achievements
• A personal AI operating system: standing instructions about who I am and how I work, a separate project space for each business, a growing library of skills, an editable memory, and connections into the tools I actually use (Airtable and Microsoft 365).
• A recall index of everything I learned across a year of AI training, written up so the learning compounds instead of evaporating a week after each lesson.
• Real daily use, not demos: fee proposals, cost checks, valuation reviews, Passivhaus design checks, correspondence, and my school governance work. The tool does the grind and I own the output.
• A scheduled task that runs on its own every week and does the whole loop for me. It scans the construction and AI press, pulls out possible new workflow ideas for my GroundMark consultancy, then cross-checks each one against my existing workflow maps. From there it acts: genuinely new ideas get added to the pipeline, supporting articles get logged as evidence for the direction of travel, and anything suggesting a workflow I'm building is the wrong call gets flagged. AI working in the background across all three of my ventures, not only when I'm sat at the desk.
• Brand books and websites for both businesses, built from the ground up.
Tools: Claude (standing instructions, projects, skills, memory, connectors, scheduled tasks, and local-file access), Airtable, Microsoft 365, and a rotating set of image, voice and research tools chosen by task.
Methodology: I treat AI less like a search box and more like briefing a new graduate surveyor in their first week. Assume it knows the textbook but nothing about how I actually work, then give it that context before expecting anything useful. Anything that works once gets saved so it pays back every time after. And one line I hold to: the tool does the labour, but I keep the judgement, show my sources, and own everything that leaves my desk. For a chartered professional that isn't negotiable.
Learnings: The habit is the hard part, not the tools. Some attempts flop, and I've stopped treating a dud as a reason to abandon a workflow. The gains compound from capturing what works, not from waiting for a perfect system. And house style matters: I've deliberately trained the generic 'AI voice' out of my setup, so what comes back sounds like me rather than a chatbot. And the best automation is the one you don't have to remember to run.
Direction of travel: Tightening the whole setup into something repeatable enough to teach, which is where GroundMark came from. Next is fuller automation of the connections between tools, and an operational data backbone to make fee proposals sharper.
Written 9 August 2026·Submitted by the storyteller
