Two document types carried the value: quote letters with itemised scope, and debris-and-access terms. A third — storm response messaging — worked well under time pressure, and a fourth, anything about tree health or risk, we did not generate.
The reason is that tree work has two categories that look similar and are not. Describing what will be cut is administrative. Assessing whether a tree is hazardous is a professional judgement requiring inspection, species knowledge and an understanding of decay. A language model has none of those, and it will not say so unless you stop it.
Where it helped
Quote letters with itemised scope
A tree quote needs to say which trees, which limbs, to what remaining height, whether the wood is chipped or hauled, whether stump grinding is included, and what happens if a power line is involved. That structure is stable across jobs; only the specifics change. Drafting fell from about 22 minutes to 6 once the template carried the itemisation, and the itemisation itself is what prevented most follow-up disputes.
Debris and access terms
Where the wood goes, whether chips are left, what happens to large timber, whether the crew needs gate access, and what happens if the access is blocked. Generated drafts covered more of this than what we had been sending. Two clauses mattered most: what happens if access is obstructed on the day, and who is responsible for damage to surfaces that were not disclosed.
Storm response messages
After a storm, the phone does not stop. A short holding reply — current response window, how the queue is prioritised, what to do if a tree is on a power line or blocking access — is worth having written before you need it. Keep the safety instruction explicit and first: anything touching a power line goes to the utility, not to you.
Where it failed
Hazard and risk assessment
Drafts assessed whether a tree was dangerous, described structural weakness, or characterised decay. That assessment requires inspection by a qualified arborist, and getting it wrong in writing creates liability as well as physical risk. We removed all of it. In our testing this was the category where generated text was most confident and least reliable.
Species, disease and treatment
Generated text identified likely species from descriptions, named diseases and suggested treatments. Species identification and diagnosis are field tasks; treatment advice can be regulated. The safe version states what will be inspected and refers anything diagnostic to the arborist.
Permits and protected trees
Drafts stated whether permission was needed. Tree protection rules vary by council, by species and by size, and in many places works on protected trees require consent. The document should say that the customer is responsible for confirming requirements, and name the authority to check with.
Utility and line clearance
Text described working near power lines as a routine scope item. Any work within reach of a conductor is the utility's job. We made this a fixed, non-editable line in every storm message.
What customers actually read
Someone calling about a tree is usually worried about two things: is it going to fall on something, and what will this cost. Stating the prioritisation logic up front — dangerous and blocking access first, cosmetic work after — reduces the number of people chasing you, because they know where they sit in the queue.
Worth adding to every quote: whether the price includes hauling the debris away. In our testing this was the most common source of a disputed final invoice, more than the price itself.
Storm season, written in advance
Every tree business has the same week in the year when the phone does not stop, and it is the worst possible time to be writing policy. Prepare three messages before the season: the holding reply with your current response window, the prioritisation note explaining what gets attended first, and the safety note stating that anything touching a power line goes to the utility. Having these ready matters more than any drafting time saved.
Editing time, from our 30-day benchmark
Quote letters went from about 22 minutes to 6, near $8 saved at $30 per hour. Debris and access terms dropped from roughly 15 minutes to 4. Storm messages from about 8 minutes to 2, and having them pre-written mattered more than the time saved.
A workflow that holds up
Template the quote with mandatory fields for tree count, height, stump work and debris handling. Keep one standard debris-and-access clause block and reuse it. Write storm messages before storm season, with the utility instruction fixed. Hazard assessment, species diagnosis, treatment advice and permit requirements stay with the arborist — inspected on site, checked against local rules, and written by whoever carries the accountability.
Itemise the work, refer the judgement
Quotes, terms and storm replies scale. Whether a tree is dangerous does not.
Try Rytr →Frequently Asked Questions
Can AI write our tree service quotes?
Yes, provided the itemisation — trees, heights, stumps, debris — is filled from the site visit rather than generated.
Should it assess whether a tree is dangerous?
No. That requires inspection by a qualified arborist, and a wrong written assessment carries liability.
Can it identify species or disease?
No. Those are field diagnoses; treatment advice can also be regulated.
What about permits for protected trees?
Have the customer confirm with their council. Rules vary by species, size and location.
How should storm messages handle power lines?
With a fixed first line: anything on or near a power line goes to the utility.
Is this useful for a two-person crew?
The quote itemisation and pre-written storm replies pay back first.