Screenshot your route in the field service app and drop it here. Faster than typing addresses, and every stop becomes a note you can finish in about a minute.
More than one screenshot is fine if the list scrolls. Duplicates get dropped.
Hold to talk
Asks for the survey without naming a score, so the numbers still mean something when they get quoted upstairs.
There is no self sign-up on purpose. Bryan creates every account.
The order matters. Experts read the whole lawn before they look at any one spot — novices dive at the first brown patch and miss the spreader stripe running through it.
Logging the knock takes a second and it is the thing the customer survey asks about. If nobody answers, the written report becomes your talk.
Straight to an answer. The photo is matched against your own confirmed library first, then the shortlist goes to the model to decide. Confidence is how often three passes agreed — not the model's opinion of itself.
One shot is enough. A tight macro helps most.
Pick the property. Everything below is instant — no waiting on the AI.
Say what you see and what you did. The AI treats your voice as outranking the photos.
Reading the lawn…
Drag a box around the leaf, lesion, or insect. This crops from the full-resolution photo — sharper than pinch-zooming with the camera.
You're the licensed expert — the AI drafts, you verify. Fix anything wrong, delete anything you didn't see yourself.
35 years of Georgia lawn experience, on tap. Ask out loud or type it.
The stuff a 35-year vet knows on sight. Tap any entry.
Your own confirmed IDs, shot on your own route. Legally clean, hyper-local, and the thing no competitor has. Tag a photo and it becomes both a comparison card for you and an example the AI can be shown.
Re-importing a backup on top of an existing library used to create repeats. Photos are now matched on their pixels, not their ID, so this cleans up what already slipped through.
A conflict is one photo filed under two different conditions. Only you can settle those — and until you do, they teach the matcher to confuse exactly those two.
Indexing is what makes a photo searchable by sight. Do it once per photo, on wifi. Everything runs on this phone — no photos leave it.
Target 15–25 confirmed photos per condition. Past 25, spread beats count — different lawns, severities, distances, light, seasons.
Holds out every photo in turn and asks whether the rest of your library gets it right. This is your real number, and it tells you exactly which two conditions to go shoot.
Your photos live only in this browser until you export them. Do this often.
The showcase file is stripped of GPS and notes and contains only confirmed photos. Drop it next to showcase.html as library.json.
Not connected. Photos live only on this device.
Connect it in Setup. Once connected, every device and every future technician sees the same library.
No account needed. Export a file here, email or AirDrop it to your other device, import it there.
This is your credibility document. Log what the AI said versus what it actually was. A tool you measured beats a tool you praised.
Stored only on this phone.
When the first pass comes back LOW confidence, the app automatically re-runs the same photos on the stronger model. That's where paid credits actually buy accuracy.
Attaches up to 3 of your saved library photos that match what it's looking at. Improves hard IDs, costs a bit more per report.
Leave blank to keep the library on this device only. Setup steps are in the guide — it takes about 10 minutes and the free tier covers thousands of photos.
The only add-ons the AI may recommend. Blank price prints "ask us for a quote."
Replace these with your company's real Salesforce codes when you get the list.
Crops matter: a lesion can be three percent of a wide shot. Cropping in stops it being averaged away.
Do this on wifi at the shop, not at a stop. It caches after the first time and then works with no signal.
Photos stay. Only the computed match data is cleared.
Working