Skip to content

Daily work

Talk to Celia

Plain-English chat over your anonymized student data. Cohort questions, aggregations, and time-window queries — translated into structured queries against your tenant database.

Who uses this
Every authenticated team member.

What it is

A chat interface for ad-hoc questions. Celia translates your question into a structured query against your data and returns a result list. Results are clickable and deep-link to Slate.

Each answer is bounded by your daily Claude budget, shown in the footer of the chat. Plan large reasoning sessions accordingly.

Every page in CeliaConnect, including this one, shows a circular module icon next to the page title that matches the sidebar tile for that section — handy for keeping your bearings as you jump between Dashboard, Students, Talk to Celia, Flows, Schedules, and Settings.

If your workspace isn't ready yet

If your data dictionary is empty, your Slate connection isn't configured, or no Flow has run yet, Celia replies with a friendly card telling you what to do next instead of guessing. The card explains the gap in one sentence and gives you a one-click button to fix it:

  • "Open data dictionary" — when you have no semantic field mappings yet.
  • "Connect Slate" — when no Slate connection is on file.
  • "Create your first Flow" — when the connection and dictionary are in but no students have flowed in.

Follow the button, complete that one step, then come back and ask your question again.

What you can do here

  • Ask cohort questions — "Students who…", "Top 10 by…", "Count of…".
  • Use time windows — "in the last 30 days", "this week", "since fall 2026".
  • Click a result row to deep-link to that student in Slate.
  • Save common questions as Flow templates if you find yourself asking the same one daily.

Common workflows

Find this morning's outreach list

  1. Ask: "Top 10 students with critical engagement in the last 7 days."
  2. Click each result row to open the student in Slate.

Sanity-check a dimension

  1. Ask: "Count of students by readiness tier in the nursing cohort."
  2. Compare against last week's count to spot drift.

Build a list for a counselor

  1. Ask the question that defines the list.
  2. Copy the anonymous IDs from the result.
  3. Paste into a Slate query / mailing list keyed by ID.

What works and what doesn't

Works well

  • Cohort questions ("Students who…").
  • Aggregations ("Top N", "Count of…").
  • Time-windowed questions.

Doesn't work (intentionally)

  • Lookups by name — Celia never sees names. Asking "What's John Smith's score?" returns a friendly "search by Slate ID instead" message.
  • Free-text writing — Celia doesn't draft counselor notes. Different surface, different problem.
  • Cross-institution comparisons — your tenant data is physically isolated.

Common gotchas

"I don't know that field"

Add a semantic label for it in /settings/data-dictionary. Celia's prompt only knows the fields you've mapped.

Daily budget hit

Each chat answer costs Claude tokens. The footer shows your daily ceiling and remaining budget. If you hit zero, the next day resets it. Contact support to lift it.