Lesson objective

Diagnose where responsibility for AI decisions actually rests across the office, the crews, the seasonal workforce and the vendors of a landscape company, and separate responsibility, accountability and authority for each.

Plain-language explanation

Why this matters now

Level 3 built operations: a service line with an owner, a written procedure a crew leader can follow, a confirmation sample and a stop condition. This level moves up a layer. The question is no longer "is this spray work order process sound?" It is the question that someone outside the company will eventually put to its owner: who in this company answers for how AI is used, and what would you show me?

The people who ask that question in this trade are specific. A state pesticide inspector following a drift complaint. An OSHA investigator after a reportable injury on a tree crew. An insurer at renewal. The board of an HOA that has read about AI in the news. A customer's lawyer after a dispute about what the proposal promised. None of them will ask which app you bought. They will ask who decided, on what basis, and who was supposed to check.

It helps to be clear about what exists and what does not.

  • No green-industry regulator or professional body has published AI guidance. Not EPA's pesticide programme, not any state pesticide agency, not a contractor or landscape-architecture board, and not the trade associations and credentialing bodies. That is the true shape of the map, not a gap in this course.
  • The ordinary law still applies to whatever a tool produced. Law / rule (United States, federal; binds employers covered by the OSH Act): under 29 U.S.C. 654(a)(1), "Each employer shall furnish to each of his employees employment and a place of employment which are free from recognized hazards that are causing or are likely to cause death or serious physical harm to his employees". Law / rule (United States, federal; binds persons, partnerships and corporations in or affecting commerce): section 5(a)(1) of the FTC Act declares "unfair or deceptive acts or practices in or affecting commerce" unlawful. Law / rule (United States, federal; binds certified applicators who rely on non-certified applicators for restricted use pesticides, and employers of commercial applicators): 40 CFR 171.201 sets the conditions for direct supervision. None of these mentions AI. Each applies no matter what drafted the schedule, the proposal or the training sheet.
  • Professional guidance (NIST, AI RMF 1.0, January 2023; voluntary, binds no company): Appendix C of the NIST AI Risk Management Framework states that human roles in decision making and oversight "need to be clearly defined and differentiated." It is a voluntary framework, NIST has announced that version 1.0 is being revised, and nobody can be certified to it. The sentence is still the right starting point.

A question to sit with

If an inspector asked on Monday who approved the photo identification app your crew leaders used to decide which beds to treat last season, would the honest answer be a name, "the office", the vendor, or a silence?

What goes wrong

Each part of the company assumes another part owns it. The office manager set up the field-service platform and thinks the spray-programme manager checked whether its AI notes were fit for application records. The spray-programme manager thinks the office manager approved the whole platform. The owner thinks both did. Everyone did part of the work; nobody made the decision.

Decisions drift to whoever is good with phones. The person who chooses tools is often the office manager, a dispatcher or a young crew leader who is quick with software. They may be very capable. They usually do not hold the applicator certification, the irrigator's licence or the arborist's credential that the work in question requires, and they have no authority to tell a certified applicator or a branch manager no.

The crews adopt tools on their own phones. A crew leader uses a free plant identification app to decide what a weed is. Another uses a chatbot to translate a safety talk. Nobody approved either use, so nobody checked either output. The company believes it has no AI in the field. It has several.

The seasonal workforce erases what the company knew. A foreman who understood the limits of the routing tool leaves in November. The spring hires never hear why the tool is not used for storm-day schedules. Knowledge that lived in one person leaves with that person.

The vendor is treated as the owner. "The software is built for lawn care companies" becomes the reason nobody checks. A vendor cannot hold the company's safety duty, the applicator's supervision duty or the company's responsibility for what it tells customers.

The owner is named as the owner of everything. In a small company the owner is accountable by default. That is not the same as being qualified. An owner who is not a certified applicator cannot make the application decision, however much the buck stops with them.

The practical answer

Separate three words that companies blur.

Word Meaning in this module Who can hold it
Responsibility Doing a piece of the work properly Anyone doing the work: office staff, crew leaders, seasonal hires, a vendor
Accountability Answering, by name, for the outcome A named person, and a qualified one where the work needs a credential
Authority Saying yes or no and having it stick A named person whose authority the owner has written down

Most landscape companies have responsibility everywhere, accountability with the owner by default, and authority with whoever is loudest on the day. Governance means putting accountability and authority in the same named place and writing it down.

Map the layers where AI decisions are actually made. A landscape company is not one workplace. It is an office, a yard, a set of trucks and a list of properties, and its workforce changes with the season. Walk each layer and ask where AI touches the work.

Layer Typical AI touchpoints Who usually decides today Question to ask
Office and sales Proposal drafting, estimating, customer messages, reviews The office manager or the salesperson Who checks what the proposal promises?
Dispatch Route proposals, weather rescheduling, crew assignment The dispatcher, or the software by default Who can override the route, and is it recorded?
Field crews Plant and pest identification, translation, photo notes Each crew leader, on a personal phone Which tools are allowed at all?
Spray programme Work order drafting, application record entry, training material The spray-programme manager Does a certified applicator own every application decision?
Seasonal workforce Onboarding material, translated safety talks Whoever runs onboarding that week Who reviewed the translation?
Vendors and subcontractors Platform features, subcontracted crews using their own tools Nobody Who agreed to what the vendor or subcontractor does with AI?

The last column is the one that matters. A blank answer is a finding.

Run the truck test. Ask five people in different layers, separately, the same question: "Who decides whether a new app can be used on a customer's property?" Ask one office person, one dispatcher, two crew leaders and one seasonal hire. Write the answers down word for word, without correcting anyone. The spread between the answers is a fair measure of how diffused accountability is.

Put three questions to the next management meeting, exactly as written.

Question A good answer A warning answer
Who approves a new AI tool for use on customer work or customer data? A named person, with a record of the approval "The office handles that" or "each branch decides"
Who can stop a branch manager or a crew leader from using a tool? A named person with written authority Nobody, or "we would talk about it"
Who answers an inspector, an insurer or an HOA board about the company's AI use? A named person who can produce records Whoever is free that week

Apply four tests to anyone proposed as accountable. The four tests are an AI Coalition Network recommendation.

  1. Named. A person, not a role, a department or a committee. The programme describes the role; day-to-day operation needs a name and a deputy.
  2. Qualified for the judgement. Whether a tool is fit for spray records, irrigation scheduling or tree assessment notes is a judgement for someone who holds, or directly supervises, the relevant credential. The office and the vendor advise on how the tool behaves.
  3. Authorised. The person can refuse a tool, pause a use and overrule a branch manager within a written scope the owner set in advance.
  4. Resourced. Time in the week, a small budget and a backup for the busy season. A title with no time attached is not accountability.

Remember this

  • A landscape company spreads work across an office, crews, a seasonal workforce and vendors, and that spreads responsibility for AI until nobody holds it.
  • Responsibility, accountability and authority differ; governance puts the last two with one named, qualified person.
  • No green-industry regulator or body has issued AI guidance; the obligations come from ordinary safety, supervision and deception law.
  • The owner is accountable by default but is not automatically qualified; a certified applicator still owns every application decision.
  • A blank in the "who decides" column is a finding, not a formality.

Landscaping example

Marlowe Property Services is a fictional commercial grounds company with 46 year-round staff, about 30 seasonal hires, two branches and a spray programme. At renewal, its insurer's questionnaire asks: "Who at the company is accountable for the use of artificial intelligence, and how is use approved?" The owner, Dana Whitcombe, does not know, so she asks around. The office manager says she switched on the field-service platform's AI job notes last spring because the vendor offered them free, but was never asked whether those notes were fit for application records. The spray-programme manager, a certified applicator, says he assumed the office had checked the platform. The north branch manager says his crews do not use AI, then remembers that two crew leaders identify weeds with a free app on their own phones before deciding what to treat. The onboarding coordinator has been translating the heat and equipment safety talks with a chatbot for the seasonal crew; nobody reviewed the translations. Dana lays the answers side by side: four positions, and a company that believed it had none. She runs the truck test with five people and gets five different answers to "who decides whether a new app can be used on a customer's property?" She answers the insurer honestly that the company is naming an accountable person and will confirm within 30 days. At the management meeting, the longest discussion is not about software. It is whether one named person may stop a branch manager's crews from using an app. Dana agrees: yes, within a written scope, with appeal to her. The operations director is named accountable for the programme, the spray-programme manager owns every AI use touching pesticide work, and the office manager is named adviser on the platform.

Responsible-use warning

Do not answer "who is accountable for AI?" with a department, a committee, the office or a vendor. None of them holds an applicator certification or an irrigator's licence, none signs the proposal, and none can answer an inspector in its own name. Naming one feels like progress and stops the question being asked. Do not assume that because the owner is accountable by default, the owner is qualified to decide every question. Name the qualified person, give them written authority and a deputy, and record it.

Human-review requirement

The owner, or whoever the owner names in writing, is accountable for the company's AI programme as a whole and for the safety duty that falls on the employer. Within it, the certified applicator owns every AI use that touches a pesticide decision, the licensed irrigator owns irrigation design or programming where the state licenses it, and the person who signs a proposal owns what it promises. Office staff and vendors inform those judgements and never hold them. The accountable person records who decided each use and on what basis.

Practical exercise

Use a fictional company, or your own company with nothing confidential.

  1. List every person, branch or function that has ever approved, restricted or informally allowed an AI tool in the company, including crew leaders using their own phones.
  2. Mark each entry R (does the work), A (answerable by name) or Au (can say no and have it stick). Note any entry with authority but no accountability, and the reverse.
  3. Complete the six-layer table from this lesson for your fictional company. Leave a cell blank if nobody decides; do not fill it with a guess.
  4. Run the truck test with five people in different layers and record their answers word for word.
  5. Write the company's current AI position in one sentence as an outsider would infer it from behaviour. If you need more than one sentence, list each position and who holds it.
  6. Pick the AI use with the highest consequence now running. Record who approved it, when and on what evidence. If nobody did, write that.
  7. Draft the four-test profile (named, qualified, authorised, resourced) for one accountable person, and note which test the owner would find hardest to agree.
  8. List what that person would need to answer an insurer's questionnaire with records rather than intentions.

Knowledge check

  1. What is the difference between responsibility, accountability and authority in this lesson?

    Answer guidance: Responsibility is doing part of the work properly and can sit with anyone. Accountability is answering by name for the outcome, which needs a qualified person where the work needs a credential. Authority is saying yes or no and having it stick. Governance places accountability and authority with the same named person.

  2. Has any green-industry regulator or professional body published guidance on AI, and what follows for a company?

    Answer guidance: No. None has. The company's obligations come from the ordinary law of safety, supervision, labels, licences and deception, which applies to whatever a tool produced. That absence is the true shape of the map, not a gap to be filled by guesswork.

  3. Why is the owner of a small company not automatically the right person to decide every AI question?

    Answer guidance: The owner is accountable by default but may not hold the credential a decision needs. An owner who is not a certified applicator cannot make an application decision; the certified applicator owns it.

  4. What does the truck test measure?

    Answer guidance: How diffused accountability is. Five people in different layers are asked separately who decides whether a new app can be used on a customer's property, and the spread between their answers shows how far the decision has drifted.

Sources and further reading