Six AI agents for a landscape construction contractor. A phone photo of paper-pad notes goes in; a branded Google Doc proposal comes out for the contractor to review and approve.
The contractor walked each property, measured the patio, planter beds, retaining walls and grades, and wrote it all on a paper pad. Then the working day ended at the desk.
What followed was one to three hours per proposal: deciphering handwriting, typing measurements, looking up material costs, sizing the job, building the proposal from scratch and emailing it. At twenty proposals a month, that is about forty hours of office time on work that produced the same kind of output every time.
The problem was not a missing tool. It was a missing pipeline. Notes lived on paper, pricing lived in his head, templates lived in old Word files, and client history was scattered.
Before any tooling, the contractor's own rules were written down: the rate card, sizing thresholds, complexity multipliers, the words he uses for scope, and the order work really happens in. Every proposal he had written contained those rules; the system needed them to be explicit.
Then came six agents, each with one job, and the contractor reviews and approves every proposal before a client sees it. It runs on commercial building blocks (n8n, Claude, Supabase, Google Workspace), so any piece can be replaced without rebuilding the rest.
Each agent hands checked output to the next. Low-confidence fields are flagged, never passed on silently. Click a stage.
Move the slider. Every figure is recalculated in Canadian dollars.
Model, not a measurement: two hours of office work per proposal before, fifteen minutes of review after. Operating cost follows the vendor price tiers (US list prices converted at 1.4188).
The stack is ordinary. The leverage is in the discipline that ties it together.
The pipeline automates the contractor's decisions, not a generic idea of what a proposal looks like.
Each agent fails safely and visibly. Nobody has to debug the whole pipeline.
Every proposal is reviewed before it reaches a client.
Phase one delivers notes-to-data in a month, before the full system is in place.
The same pattern fits claims intake, support triage, approvals and most work that moves information from one place to another. Start with a conversation about yours.
📅 Book a call