Numbers you can trust, and an assistant that ranks your teams.

Several figures were being read from tables that had stopped being the source of truth. They now come from the completion and progress records themselves, so what you see reconciles with what happened.

• Quiz pass rates read the real progress record — not a completion flag that was always true.
• “Learners Enrolled” counts distinct people rather than summing enrollments, so a worker on three courses is one learner.
• Time-on-task, recency and the daily-activity timeline are all sourced from live completion events.
• Empty and low-reach states now say plainly that there is not enough data yet, instead of implying a zero.

Ask who is behind and get an ordered answer across your whole organisation, or by team.

• Org-wide per-person readiness table with ranking, so “who needs attention first” has one answer.
• Team-level readiness ranking — ask which team is behind.
• Conversation memory, entity resolution and freshness checks, so follow-up questions understand who you meant.
• Answers refuse to substitute a name it is not sure about, and the overdue list no longer repeats people.
• Rebuilt reply rendering — tables, quotes and structure survive intact — with an action bar on every answer.

A self-service account page, and organisation settings scoped to the people who own them.

• New My Account page for your own profile and password.
• Organisation settings are now restricted to org administrators.
• Trainers can see the AI credit balance without being given billing access.