Level 4 taught a landscape company to govern AI. Level 5 is about building the part that acts. A drafting assistant suggests; an automation does. It books a site visit, publishes a route sheet, sends an arrival notice, prepares an irrigator's morning packet or drafts an invoice, and when it is wrong the error has already reached a customer, a crew or the ledger by the time anyone looks. This level teaches automation whose authority is bounded in writing, whose every action is recorded, and whose failures are caught by design rather than by a customer's phone call.
The first three modules set authority and orchestrate the two flows where automation pays off and fails most visibly. Module one sorts every proposed automation into suggesting, acting inside written bounds or waiting for a named person, writes permitted actions, forbidden actions and required approvals, keeps the agent authority register through a busy season, and names the decisions no landscape agent may make. Module two automates the path from enquiry to booked site visit while the estimator decides scope and price and an authorised person signs, and treats the federal calling and texting rules, revocation of consent and AI voices as hard requirements. Module three orchestrates the recurring cycle from schedule to route sheet, arrival notice, closeout and invoice, handles weather exceptions and skipped visits out loud, and runs the winterization batch so that no property is dropped.
Modules four to six are the integrations. Module four draws every system an automation touches, from the CRM to the controller portal and payroll, writes a data contract at each boundary, keeps biometric data and identity documents out under the state and federal rules that apply, and shows how one wrong property record travels through an automated cycle. Module five starts every automation read-only and makes it earn each write, separates proposed records from posted ones, keeps invoices, credits and refunds behind a named approver and card data out entirely, and sets the reconciliation every write needs. Module six treats route, weather and controller feeds as inputs to a proposal a person accepts: heat and storm thresholds are set by a named safety lead, a labelled controller is not a measured saving, and changing a controller can be licensed irrigation work in some states.
Modules seven to nine are human control. Module seven designs approval gates that produce a decision rather than a reflex, explains what the research on automation bias found and where its limits lie, fits gate volume to a crew leader's morning, and measures whether each gate works. Module eight gives every automation its own identity, a named owner and the narrowest rights, sets roles for the office, crew leaders, seasonal workers and subcontractors, and runs onboarding, offboarding and access reviews at the speed the trade hires and lays off. Module nine designs levels of stop that work at five in the morning, reconciles work in flight and half-sent notices, keeps the manual route sheet and paper closeout alive, and plans the return to automation before it is needed.
Modules ten to twelve are defence and observation. Module ten threat-models customer mail, review sites, bid specifications and safety data sheets as carriers of instructions, from the defender's side only, explains why injection remains a residual risk once an agent holds tools, and reads the 2025 and 2026 editions of a voluntary risk list side by side. Module eleven specifies the agent action record, what must never be logged, integrity and retention, and how the trail sits beside the wage, injury and other records the company must keep anyway. Module twelve finds failures that raise no error, retests every vendor accuracy claim on the company's own properties after each change, caps cost, and names who answers an alert at seven on a Saturday.
Module thirteen responds when an automation has already failed: declaring the incident, containing and recovering a missed winterization batch, a wrong-property message or an exposed gate code, checking which notification duties reach the company and which injuries must be reported to OSHA, and running a review that changes the system rather than blaming a person. Module fourteen measures time, quality and cost against a baseline taken before the change, refuses double-counted savings, seasonal illusions and borrowed figures, and takes a bounded pilot to a recorded decision. The capstone is the Automation Implementation Project for a fictional landscaping company. The level ships with a printable workbook and eleven templates, and the examination draws forty scenario questions from a reviewed bank.
Everything here is professional education. It is not legal, agronomic, arboricultural, engineering, security, safety or insurance advice, and it does not replace the product label, a certified pesticide applicator, a licensed irrigator, contractor or tree expert where a state requires one, the employer's safety programme, legal counsel, or the law of the state where the work is performed. Completing the level earns an independent educational certificate issued by AI Coalition Network with a public verification page. It is not a licence, an applicator certification or a safety or security credential, it carries no professional education hours or recertification points, and it satisfies no licensing, certification or safety-training requirement.