Aurora is the one assistant across the whole ecosystem. It answers from your workspace: the projects, schedules, invoices, and call sheets your team already keeps here. Grounded answers, cited sources, and a human holding the publish key.
Powered by Brio, the conversational runtime that gives Aurora its chat, voice, and SMS channels. Brio runs the conversation; your org's records supply the truth.
Every item below carries its real state. Live means shipping, preview means a deliberately limited surface you can touch today, and coming means built for a product that has not opened yet.
The field app. Live now.
Point the phone camera at a receipt or an invoice. Aurora reads the vendor, the total, and the date into a draft, suggests a cost code from your own spending history, and waits for you to confirm. Nothing writes itself into the books.
The Aurora tab ships in the field app today as a guided assistant. Ask where your next shift is and it walks you to the right surface; it will not fake a number it cannot see. The conversational runtime arrives with Brio.
Where the organization lives.
Your standards, SOPs, and event guides become the corpus Aurora answers from, synced per event. The answer at the gate matches the binder in the production office because it is the same document.
The producer's console. In active development, and Aurora is already wired into it.
Ask a question and get an answer built only from your org's indexed documents, with each claim cited by source number and a confidence grade on the whole answer. When no source matches, it says so instead of guessing.
Streaming chat with conversation history, kept inside your tenant. Claude Sonnet for the day-to-day, Claude Opus when a question deserves the long think.
A rail of next actions computed from real org data: blocked tasks, approvals waiting, open incidents. It never applies anything on its own. Every card deep-links to the surface that clears it, and you do the clearing.
Client-ready proposal drafts from the record's own metadata plus your instructions. You accept or discard; a draft never publishes itself.
AP invoices, COIs, and W-9s read into structured fields with a confidence score on every extraction. Anything under the bar routes to a human review queue instead of the ledger.
A transcript in, a tight recap and concrete action items out, saved on the meeting note. Field agents keep a derived column fresh from a record's own values whenever the record changes, against an allowlist of tables you control.
Where the world is experienced.
Aurora answers the phone. Box office questions, parking, lost and found, credential lookups, over voice, SMS, or web chat, around the clock, with a human paged the moment a conversation needs one. This is the Brio channel layer end to end.
Aurora reads through the same row-level security your team does. Your tenant is separated in the database itself, and another org's data is unreachable by construction.
Grounded answers come only from your indexed corpus, cite each claim by source number, and grade how well the sources support the answer. No matching sources means it says so instead of guessing.
Drafts stay drafts until someone accepts them, and suggestions never apply themselves. Extractions below 90 percent confidence route to a review queue.
We do not train models. Aurora runs on Anthropic Claude, and your conversations are logged inside your own tenant where you can query and export them.
Every call is rate-limited before it reaches the model and usage is metered per org, so a runaway script cannot run up your bill.
Streaming, thinking, citations, and confidence share one design language across every product, so you always know when Aurora is generating, what it consulted, and how sure it is.
The conversational runtime lands surface by surface, and waitlist members hear about each one before anyone else. One email per opening. No drip campaign.
Start free, scan a receipt tonight, and let the rest of Aurora meet you where the work is.