Keep the knowledge moving.

Why separate knowledge capture from deterministic systems

Tacit knowledge—site quirks, client preferences and routine workarounds—often lives in experienced employees rather than recorded systems. The industry guidance is explicit: "In most landscape companies, the details that make a job run live entirely in people's heads: the site quirks, the client's preferences, the reason you never park on that side of the driveway," .

Routing and mower systems use different combinations of optimization, sensing and control. Ask each provider how its system handles travel times, boundaries and safety rules. Keep those controls separate from generated text so changes that affect safety or daily sheets flow through a named human approver rather than an unconstrained generative edit.

What to capture and how to capture it

Start by collecting the veteran knowledge that creates single-person bottlenecks: access notes, habitual exceptions, material and color preferences, and the reasons behind routine parking or routing choices. Capture during existing crew patterns rather than asking crews to adopt a new typing habit: "When the crews gather at the shop morning and evening, they're already talking through the job," .

Make capture low-friction: voice or short bilingual inputs at morning huddles and end-of-day checks so crews keep current habits. The same guidance notes crews can ask questions in Spanish or English and receive answers in the same language, broadening access and reducing a single-person knowledge chokepoint .

Store entries as time-stamped, queryable records. The knowledge base should be read-only to deterministic systems unless a dispatcher or named approver reviews and signs off on any change that affects routing order, mower boundaries, or safety parameters.

Routing, autonomous mowers and exception management

Make a route engine the single source of truth for daily sheets and ETAs. Route optimization is operationally valuable: "The more optimized your routes are, the more job sites you can visit in one day. Additionally, optimized routing will minimize back-and-forth travel of your team," .

Commercial autonomous mowers rely on explicit boundaries and sensing stacks—GNSS/GPS, LiDAR and computer vision—to keep machines inside safe perimeters and respond to obstacles. Product descriptions explain that these systems can save a perimeter and adapt when site obstacles change, "It remembers where different obstacles are, but if something new comes up, it will move around it and rechart its path for next time and moving forward," .

Operational rule: when a generative system suggests an exception (temporary gate code, stay-out zone, client-requested avoidance), require a named human approver before the route engine or mower controller accepts a write. That keeps safety and contractual constraints auditable and separates descriptive knowledge from determinative control.

A bounded pilot you can run next quarter

Pilot scope: two crews, roughly 120 recurring stops and one site with autonomous mowing enabled. Duration: an eight-week engagement—two weeks onboarding (ingest, crew training) and six weeks live operation.

Practical steps: ingest 8–12 months of job notes and fixed customer data into a searchable knowledge base; enable voice capture at morning huddles and end-of-day; integrate the route engine as a read-only consumer of exception flags; require dispatcher approval for any exception that would change route order or mower boundaries.

Approval workflow: dispatcher reviews exception flags before applying changes; technicians escalate with photos/voice notes for sub-24-hour decisions. Treat vendor descriptions of autonomous-mower behavior as product characteristics, not guaranteed operational outcomes; this pilot is a bounded operating hypothesis to test process and measurement.