Software
The software around the test
SNOW began as a way to read a rapid test from a phone photo. The same software now keeps the record of each test, handles a practice's conversations with pet owners and colleagues, and maps its results beside public case reports for the region.
SNOW uses AI in a few places, such as taking a history from an owner and suggesting updates to the record. Nothing the AI suggests reaches the record until a clinician accepts it.
Everything except second opinions between practices works inside a single practice, and those are waiting for other practices to join. The regional picture grows more useful with every practice that does.
Reading and recordsConsultationsAIPractice softwareYour regionStatus
Reading a test and keeping the record
A technician photographs the cassette, and SNOW measures each band or spot and files the reading with the patient, the clinic, the time and, when the owner's ZIP code is on file, the county. The photo stays on file for a clinician to look at again.
The record
Each record keeps the photo, the software's measurement and the clinician's call. A clinician's correction is saved beside the original reading and never replaces it.
A treatment reference for veterinarians
When a screening result is positive, veterinarians see a reference beside it covering confirmatory tests, the workup, management and common pitfalls, summarized from published guidelines. It covers heartworm, Lyme disease, anaplasmosis and ehrlichiosis in dogs, and feline leukemia virus and FIV in cats.
It is fixed reference material, the same for every case, and no AI writes it. Only veterinarians and practice administrators can open it, and pet owners never see it. We are building AI-assisted treatment options to sit beside it.
Consultations and second opinions
The practice keeps every conversation about a case in one place, from an owner's question to a second opinion from a colleague.
Questions from owners
Pet owners can ask a veterinarian at their practice or start a live chat with an AI assistant. Every thread lands in one queue for the team to claim, and the team can open consults of its own.
Second opinions
A clinician can ask a colleague for a second opinion. The software suggests who to ask and shows why. Once other practices join, colleagues there can be asked too, with the owner's agreement. That part is built and waiting for practices to join.
From a phone
Consults work fully in a phone's browser, so clinicians can read and reply wherever they have a connection. A companion iPhone app is in development, with notifications that reach the phone even when the app is closed.
Where SNOW uses AI
The AI in SNOW asks owners questions, writes summaries and suggests notes for the record, and what it writes is marked as the assistant's until a clinician accepts it. A veterinarian makes every clinical decision.
AI intake for owner chats
When a pet owner at a participating practice starts a live chat, the assistant replies in plain language. It asks a few focused questions, such as how long the problem has lasted, appetite, energy, parasite prevention and current medication. It skips prevention and medication when the record already has them, and anything the owner has already mentioned. It tells the owner that a veterinarian will read the thread and stops replying once a clinician takes over.
Urgent signs come first
Fixed rules, separate from the AI model, check what the owner writes for signs of a possible emergency. When one matches, the reply opens by telling the owner to call their clinic or an emergency service now, and the thread is marked urgent in the practice's queue.
A hand-over for the team
As the chat goes on, the assistant keeps a short hand-over for the clinicians, with what the owner reported, how soon follow-up looks needed and anything that seems unusual for the test. The queue is sorted by how soon follow-up looks needed, so the team can start the morning with the threads that look most pressing.
Record updates, one at a time
From each chat, the software suggests short notes for the patient's record, such as a current medication or parasite prevention. A clinician adds or declines each one, and there is no button that accepts them all. An added entry says the owner reported it in a chat and names the clinician who approved it.
Trends in plain language
Beside a forecast, a veterinarian can ask for a plain-language explanation of a regional trend. The AI is given only figures the software has already computed. It never sees patient records, cannot change the statistical call, and is instructed to say so when the data is too thin to mean much.
Treatment options for veterinarians
We are building AI assistance that suggests treatment options for the case in front of a veterinarian, to review beside the treatment reference. It is for veterinarians only, and the assistant that talks to owners will still never name a drug or a dose.
Where SNOW does not use AI
Reading a test, the treatment reference, forecasting and the check for unusual rises use no AI. They run on measurement, published guidelines or statistics that can be checked. Suggesting a colleague follows fixed rules, which also read the words and urgency of the assistant's hand-over. The assistant that talks to owners is instructed never to diagnose, name a drug or a dose, or say whether an animal needs treatment.
Owner chats, with the parts of the record the assistant needs, are sent to an outside AI service, which writes the assistant's replies and the team's hand-over.
Working with your practice software
Every test result can be downloaded as a one-page result sheet, ready to attach to the patient's record in the practice-management software your clinic already uses. The sheet writes out the control and every band or spot in words, includes the photo it was read from, and puts any caveats above the numbers.
Each result is also saved in a structured format designed for filing into practice software. An invalid test is never recorded as a negative, and neither is a borderline result.
Direct connections
We are working toward direct connections with veterinary practice-management systems, so a result can file straight into the patient's chart without being attached by hand. Until then, the result sheet does that job.
Your region
A national disease map
Each result filed by a clinic carries its date and, when the owner's ZIP code is on file, its county. For each disease, in dogs or cats, a county map shows the share of clinic results filed in SNOW that were positive. Beside it are human case reports that the CDC and state health departments publish, which run a year or more behind.
Each source stays on its own layer and is credited by name. SNOW's own layer works in rates, so a practice that tests more does not look like one seeing more disease. A county with too little testing is left blank so that it never reads as disease-free.
Regional views never show who the owner or the patient is.
Your region
Short-range forecasts
For each disease, the software projects the weekly share of positive tests a few weeks ahead from the clinics' own testing, for one state or the whole country. It uses standard statistical methods, set out beside each forecast, and no AI.
Each projection shows its uncertainty band, a control limit and the plain reasons for its call of Stable, Watch or Rising. When there is too little testing to call a trend, it says so. It also notes that a rising share can reflect a change in who is being tested as well as a change in disease.
Your region
Checking for unusual rises
The software compares each state's latest complete week of results with the weeks before it. It works out how many positives those weeks would predict at this week's level of testing, and flags a week that comes in well above that as Watch or Alert.
Because the expected count rises with the number of tests, a practice that simply tests more does not raise a flag. The check skips the week just before, so a rise already under way does not become its own baseline, and it makes no call when there are too few tests to judge.
The check runs on its own every hour as new results come in. A flag is a reason to look again and does not confirm an outbreak.
Current status
SNOW is in development and is not a cleared or approved diagnostic device. Every reading is presented for a clinician to confirm, and nothing the AI suggests reaches a record without a clinician's approval. The table shows which parts are built into the software today.
| Part | Status |
|---|---|
| Reading a test | Reading commercial tests today |
| Records and one-page result sheets | Built |
| Treatment reference for veterinarians | Built |
| AI-assisted treatment options for veterinarians | In development |
| Questions from owners, by Ask a vet or live chat | Built |
| AI intake, hand-over and record suggestions | Built |
| Second opinions within a practice | Built |
| Second opinions between practices | Built; waiting for practices to join |
| Recognition for colleagues who answer | Built |
| Payment and software credits for answering | In development |
| Consultations on a phone | Built, in the phone's browser |
| Companion iPhone app | In development |
| Disease map and forecasts | Built |
| Plain-language trend summaries | Built |
| Hourly checks for unusual rises | Built |
| Direct connections to practice software | In design |