Reasons, not just a label
Every verdict carries the criteria it matched, the passages it relied on, the risks it found, and what it still needs to know. You can argue with it.
Bid intelligence
RevX10 watches thirteen procurement portals, folds what it finds into a single record, and grades each one against your own rulebook — so the shortlist that reaches your team is the part worth arguing about.
13
Portals wired in
7 crawled server-side, 6 via the browser extension
13
Rulebook criteria
Every tender scored against all of them
5
Triage verdicts
GO · KIV · NO_GO · PENDING · LEARNING
How it runs
Each portal has its own crawler — a JSON API where one exists, a headed browser where Cloudflare demands it, the Chrome extension where a human login is unavoidable.
Thirteen different payload shapes collapse into one tender record. Reference numbers and source URLs are fingerprinted, so a re-crawl updates rather than duplicates.
Each new tender is read against the rulebook: 13 criteria, evidence quoted from both the tender and the policy, and a GO/NO-GO split with a confidence figure attached.
Verdicts are a starting point, not a verdict. Opening a tender marks it in review; a human confirms or overrides, and that decision feeds back as training signal.
What it handles
Every verdict carries the criteria it matched, the passages it relied on, the risks it found, and what it still needs to know. You can argue with it.
Runs are fingerprinted on content, so a nightly re-crawl re-scores only what actually changed and quietly skips the rest.
Portals leave expired notices up for months. Anything past its closing date is filtered before it reaches your queue.
Every record carries a thread. Ask why it scored the way it did, or what a clause implies, without leaving the row.
Start here
Crawls run on a schedule, scoring happens as records land, and the queue is waiting when your team opens the dashboard.