Operations: From Chaos to Checklists
11 min read
Nobody starts a landscaping company because they love writing procedures. Maya certainly didn't. But 14 employees means 14 people who might do a thing differently than you would, and every spring her phone proves it — crew leads calling about mulch depth, sprinkler flags, whether to bill for the gate repair. Operations prompts are the least glamorous use of AI and the highest return. This lesson covers the three that matter most.
The voice memo that became an SOP
Maya's spring cleanup process lived in exactly one place: her head. Writing it down had been on her to-do list since 2023.
So she stopped trying to write it. On a site visit in March, she recorded a six-minute voice memo, just talking through the job while walking a property — rake this first, check irrigation heads before the mower goes anywhere, photograph winter damage before touching it — out of order and full of tangents. Her phone transcribed it. Then:
You're an operations manager who writes clear procedures for
field crews.
Context: Below is a transcript of me talking through our spring
cleanup process. It's rambling and out of order. My crews know
the equipment, but new hires don't know our standards.
[PASTE TRANSCRIPT]
Task: Turn this into a step-by-step SOP a new hire could follow.
Format: Numbered steps in order of execution. Bold anything that
protects us from a damage claim. Flag any spot where I contradicted
myself or left a gap, and list your questions at the end.
Constraints: Plain language. One page, maximum.That last format request is the pro move. Asking the AI to flag gaps turns it into an editor, not a typist — Maya's transcript, it turned out, never said what to do with the debris. Six minutes of talking, ten minutes of editing, and a job that had waited two years was done. Structurally, this is just The Five-Part Prompt pointed at unglamorous work.
Meeting notes into action items
Same trick, smaller scale. After her Monday crew-lead meeting, Maya pastes her messy notes and asks for a table with three columns — the task, who owns it, the deadline — plus a separate list of anything discussed that ended without an owner. That second list is where dropped balls hide. Most meetings don't fail on decisions; they fail on the four items everyone nodded at and nobody took.
The decision tree: your judgment, written down
Now for the concept this lesson exists to teach, because the rest of the course builds on it.
A decision tree is a written if-this-then-that script an employee can follow without calling the boss. That's the whole definition. No software, no whiteboard diagram — just your judgment, captured as a short series of questions where every path ends in a clear instruction.
Here's the scene: it's 7 a.m., it's raining, and a crew lead has to decide whether today's jobs run. That used to mean calling Maya, who was already fielding two other calls. So she brain-dumped how she actually makes the call and prompted:
Context: I own a 14-person landscaping company in Calgary. On rainy
mornings, crew leads must decide whether to run or reschedule jobs.
Here's how I think about it, unfiltered:
[BRAIN DUMP: light rain we mow anyway unless the property is steep...
heavy rain ruts the lawns, never worth it... some clients are
flexible, it's noted on the job sheet... don't push work into a day
that's already full...]
Task: Turn this into a decision tree my crew leads can follow on
their phones at 7 a.m. without calling me.
Format: Numbered questions. Each answer leads to an action or to
the next question. End every path with one clear instruction. If
my logic has a gap, ask me about it.One slice of what came back:
- Is the rain heavy enough to leave standing water or wheel ruts? No — run the job as scheduled.
- Yes — check the job sheet. Is this client marked flexible? No — call Maya.
- Flexible — does tomorrow have fewer than six jobs booked? Yes — move it to tomorrow and text the client before 8 a.m.
- Tomorrow is full — move it to the next open day and text the client before 8 a.m.
Four lines, and an entire category of morning phone call disappeared. Notice what a decision tree really is: one big judgment call broken into small, explicit steps. Hold that thought — it comes back in the next chapter as the key to running big projects with AI.
Try it now
Pick the question your team asks you most often. Talk your answer into your phone for three minutes — rambling is fine, that's the point. Paste the transcript into Claude or ChatGPT and ask for an SOP or a decision tree, and make sure to ask it to flag gaps in your logic. It will find at least one.