Cortex is a research workspace for teams that have to forward what they find. Every claim is footnoted to the passage it came from, every good prompt becomes a playbook, and every answer shows its cost before it is written.
Why do buyers hesitate to forward AI answers to legal?
Seven of nine teams said an answer is only forwardable if the reviewer can open the document behind it1. Legal wants to know which version of a policy was read2, and whether it was still current when the answer was written3.
Cortex reads your documents in passages, not pages. It shows you which ones it used, which it set aside, and why — so when an answer is wrong you can see whether the problem was the question, the source or the search.
Save a good request as a playbook with inputs, approved sources and checks. Anyone on the team can run it; a run that fails its checks is marked, never quietly shared.
Write a launch brief for product aimed at audience. State the single promise in one sentence, then the three strongest pieces of proof…
Set a ceiling per answer and a limit per month. Cortex estimates each reply before writing it and stops at the line you drew, without touching anything already written.
$184 of the $200 limit is spent with three days left. At this pace the workspace reaches the limit around October 8.
About $0.012 for this reply. $0.030 left of this thread’s $0.08 limit.