For small offices where paperwork, follow-up, and trust all live in the same room.
I help relationship-heavy professional teams find the hidden structure under intake, documents, handoffs, status updates, and repeated admin.
The point is not to bolt a chatbot onto a broken process. The point is to map the work, find the leaks, build one safe improvement, and leave behind a plan a real office can use.
Delivered · in production
Every document was its own island. The same job existed in four places with four slightly different sets of numbers, and the only thing holding it together was one person remembering which file was current.
I built the truth layer first: one canonical record per job, with every document generated from it and none of them allowed to become a second source. Then the documents themselves, off that single record, formatted the way the firm already sent them to clients.
It lives inside the client's own Google account, not on my infrastructure. No server to maintain, no subscription, no dependency on me still being available in two years. Owner and admin tools stay hidden. The daily path is open the workbook and click the document you need.
Two weeks after handoff it was still in daily use by two people who had no part in building it. That is the measure I care about, and it is the reason there is no screenshot here: the system belongs to them now.
Longer history, and how the work has been done since 2005, lives in the archive.
Most teams do not need another AI demo. They need clearer intake, better handoffs, cleaner notes, safer retrieval, and fewer “where is that?” moments.
My work starts with the process before the model. I look for the truth source, the permission boundary, the human decision point, and the one change that would make tomorrow less brittle.
By day, I work inside a relationship-heavy professional-services environment where accuracy, discretion, and follow-through are the job. Outside that lane, I build and audit practical AI systems for small offices, caregivers, and operators who cannot afford theatrical software.
Before AI, I spent nineteen years building practical systems for clients who needed the work to run without hand-holding: custom portals, workflow tools, CRM architecture, secure web applications, and documentation non-technical teams could actually use.
I take on a small number of fixed-scope projects for professional-service teams, consultants, and operators who need a clearer path through messy information or a repeated workflow that keeps wasting time.
A fixed-scope look at one paperwork-heavy workflow: a working session to hear how the job actually moves, a map of where materials enter and stall and repeat, the recurring losses ranked by what they really cost, one bounded template or checklist you can use immediately, and a 30-day plan that does not require turning the office upside down.
Building the system, and mapping what an AI should be allowed to know before automation touches it, both follow from the audit or stand on their own once the problem is already mapped. Those are conversations, not menu items.
If the problem is still fuzzy, we start with a conversation. If the work is not bounded, we do not start.
When I work inside a live client account, a wrong write is not a bug, it is somebody's missing file. So on that work the guard gets built before the feature does: automated writes confined to a sandbox declared in advance, delete and move and share blocked outright at the tooling layer, production read-only by default, and a record of every action taken.
None of that is visible to the person using the system. It is the reason they can keep using it without thinking about me.
The same principle runs through everything: make the system honest about what it knows, what it does not know, what it is allowed to touch, and when a person needs to decide. In practice that means source-of-truth rules, retrieval boundaries, audit trails, consent gates, failure checks, and documentation that survives handoff, built on Google Apps Script and Sheets, on current AI models with deliberate routing between them, and on nineteen years of PHP, JavaScript, CRM architecture, and secure client portals underneath.
The audit work is new. The pattern is not.
From 2005 to 2024, I built and ran a systems and design consultancy with no advertising, no investor capital, and no marketing budget. Just sustained referral from clients who trusted the work enough to send other people to it.
Retention and collection figures are counted from the practice's own invoicing records, which cover 2012 to 2023.
Representative work spanned healthcare-adjacent portals, multi-division talent agency systems, commercial real-estate operations, nonprofit arts platforms, and payment workflows. Different industries, same pattern: people brought a fragile process, and the answer was a clearer system.
The current work arrives the same way it always did. Someone who trusted the last system sends the next person.
I am interested in practical AI systems for teams whose work depends on trust: professional services, healthcare-adjacent operations, family-care coordination, financial services, insurance, and small offices with too much knowledge trapped in people’s heads.
The goal is not to replace people. It is to make the work easier to find, hold, hand off, and trust.