Use cases
What people build with it
Every workflow below ships inside AWO as a recipe you fill in — these are not illustrations of what might be possible. Pick one, answer three or four questions in your own words, and it arrives working. There are 49 of them; these are the ones people reach for first.
Work that comes to you
Leads, tickets, orders and applications are not things you watch a page for — they arrive. Point your form, shop or helpdesk at one address and AWO reads every submission against a rule you wrote in plain English.
Qualify new leads
“Companies with 10+ staff in retail, budget over $500 a month. Not students, not agencies.” Good leads reach your team with a one-line summary; the rest never interrupt anyone.
your form→ask the AI→only if…→Slack / Discord
Triage support requests
The ticket that is actually on fire stops sitting behind forty that are not. You describe what urgent means for you; everything else waits.
helpdesk→ask the AI→only if…→alert
Decide refund requests
Requests are read against the policy you describe. The ones that clearly fit pass quietly; the judgement calls come to you with the reason already stated.
shop→ask the AI→only if…→alert
Catch cancellations worth saving
A downgrade comes through, AWO reads the reason, and you hear about the ones worth a personal email today — while it is still today.
billing hook→ask the AI→only if…→alert
Screen applicants and RFPs
Applications read against what the role actually needs; inbound briefs read against the work you actually win. You stop writing proposals you were never going to get.
form→ask the AI→only if…→alert
Flag orders and expenses
Mismatched countries, a first-time buyer taking four of one thing, an invoice from a supplier you have never used. The routine ones go through untouched.
shop / finance→ask the AI→only if…→alert
Making money on the side
The recipes people run for themselves: finding the work, spotting the margin, and not being the last to know.
Freelance gigs that fit you
“React builds, $70/hour or $2k a project, remote. No unpaid trials.” You hear about the ones worth your evening, before fifty other people have applied.
job board→ask the AI→only if…→alert
Deal radar across marketplaces
The same hunt on several sites, merged into one list of what is actually worth buying — with a note on why the margin survives fees.
3 marketplaces→merge→ask the AI→one list
Clearance worth reselling
Watches a liquidation page and tells you when stock lands cheap enough to move at a profit — in quantities that make it worth the drive.
clearance page→ask the AI→only if…→alert
Grants, bounties and paid calls
Funding you actually qualify for, paid bounties in your stack, speaking and writing that pays. Three recipes, one rule each: is this real money for work I can do?
funding page→ask the AI→only if…→alert
Sort the offers on what you sell
Haggles read against your floor. Serious offers get answered; “is this still available” and the pay-via-an-agent scams do not reach you at all.
marketplace→ask the AI→only if…→alert
Client enquiries worth taking
Budget, scope and the red flags you have learned the hard way — “simple job”, no budget, wants to pay on results. You spend evenings on the ones that pay.
your form→ask the AI→only if…→alert
One report instead of forty alerts
Several pages read together, so you get a picture rather than interruptions. This is what several inputs feeding one graph is for.
Keep an eye on your competitors
Pricing pages, changelogs and blogs across the field, in one round-up of what actually changed — not five separate alerts to piece together.
5 pages→merge→ask the AI→one report
Watch your suppliers together
Who raised a price, who went out of stock, who quietly discontinued the line you resell. One message, in the order that matters.
supplier pages→merge→ask the AI→one report
A morning brief
The pages you would check yourself, read for you, in a few lines before the day starts. Quiet when there is nothing to say.
news + blogs→merge→ask the AI→brief
Watching a page
For anyone who currently keeps a tab open and checks it.
Watch a page for changes
Checks a page on a schedule and tells you when something you care about changes — with an AI step in between deciding whether the change was worth a message.
schedule→fetch page→ask the AI→only if…→alert
Tell me when the price drops
Watches a product page and alerts you when the price falls below a number you set. The comparison is a real condition, so it does not message you about every reprice.
schedule→pull out price→only if…→alert
Watch a status page
Alerts you when a service you depend on posts an incident, so the first you hear of it is not a customer.
schedule→fetch page→ask the AI→alert
Chat that answers itself
Where AWO's graph earns its keep: a busy channel would bankrupt a one-run-per-message tool, and batching turns fifty messages into one model call.
Triage a Discord or Telegram channel
Reads a batch of messages, classifies what people are asking for, and routes each kind somewhere different — a branch per outcome rather than one catch-all reply.
discord · batched→ask the AI→only if…→reply / escalate
A digest on a schedule, alerts in real time
The same graph fed by two inputs: chat messages run the alert branch immediately, while an hourly schedule runs the summary branch. Neither triggers the other's work.
discord+schedule→one graph, two scopes
Development work
These need the code-writing capability, which is off until you turn it on and runs in a sandbox that opens a pull request rather than pushing to your branch. For work bigger than a single recipe, Project Autopilot takes a goal and a repository and plans the phases itself.
Plan → implement → test → fix
The full development loop: plan the work, build it, run the tests, and go back and fix what failed — up to a limit you set on attempts, time and spend.
issue→plan→implement→test↺fix
Code review pipeline
Reads a pull request's diff, reviews it against the standards you describe, and posts the findings where your team reads them.
PR webhook→read diff→review→comment
Reproduce → fix → verify
Turns a bug report into a failing test, fixes the code until that test passes, and refuses to open the pull request if it does not.
bug report→failing test→fix→gate
Refactor behind a test gate
Restructures code and holds the change to one rule: the suite that passed before has to pass after. Nothing ships if it does not.
target→refactor→suite→only if green
Docs from the code
Regenerates documentation from what the code actually does now, and opens a pull request with the difference.
schedule→read code→write docs→PR
All 49 are in the template gallery
Choose one, answer three or four questions in your own words, press go.