Skip to main content
Print this. Keep it on your desk. One page. Two halves.
Login: AllCare → Care Partner workspace Slack: #ai-concierge-launch Pillar Champions: Germin, Melinda P0 escalation: page Ramy + Sakr direct

​
Your two queues

​
Top half: the 3-minute Pre-Dispatch review

For every task in Pre-Dispatch, run this. Three minutes or less.

​
Step 1. 30-second sanity check

  1. Real request? Or noise?
  2. Right patient? Look at confidence tier:
    • VERIFIED → fine, move on
    • PROVISIONAL → eyeball it
    • NOT_FOUND → manual resolution
  3. Right urgency? STAT matches “now / ASAP / critical” only.

​
Step 2. 60-second classification check

  • Pillar correct? (Pharmacy / Clinic / Routing / External)
  • Task type correct?
  • Subtype correct?
  • Parameters complete? (medication, dosage, provider, etc.)

​
Step 3. 60-second decision: pick ONE

Time budget for an edit: 60-90 sec. Longer than that = Reject + Reroute.

​
Bottom half: cheat sheet + edge cases

​
Decision cheat sheet

​
Pre-Dispatch SLA (target dwell time)

​
Open Tasks (your follow-up work)

The work you do today, sourced from AllCare instead of Freshdesk:
  1. Watch dispatched tasks. Status updates flow back automatically when pillar marks done.
  2. Follow up if a task stalls past its SLA.
  3. Communicate back to the requester: confirmation, status, completion, blockers.
  4. Close the loop in AllCare. The task isn’t done until you close it.

​
When to escalate to a Pillar Champion

  • Can’t tell real vs noise
  • PROVISIONAL patient, can’t confirm in 2 min
  • More than 3 edits needed = Reject + Reroute instead
  • Same wrong classification 3+ times in an hour (prompt fix, not single edit)
  • PHI exposure smell
  • Patient safety smell
  • You need a break

​
Report Bug (anything else not right)

Click Report Bug anywhere in AllCare. The form auto-fills the task URL, then asks:
  • What happened wrong?
  • What is the expected behavior?
Add patient context if relevant. Submit fires to #concierge-bugs (team channel) and Linear (auto-created issue). Reviewed daily. Use it any time something is not right. Bugs, comments, feedback.
For wrong AI dispatch decisions, use Reject + Reroute or Edit + Approve. Those auto-log corrections to the AI training data. Report Bug is for everything else.

​
What you are NOT responsible for

  • AI quality (Sakr + Ramy + 4 PM review)
  • Pillar workflow correctness (pillar owns it after dispatch)
  • Outcome chaining (Tasks Pool handles it)
  • Following up on stuck downstream tasks (the pillar coordinator does this)

​
Daily 4 PM review (week 1)

15 min with Sakr + Ramy. Bring:
  • Edit count + most common edit pattern
  • Reject count + reasons
  • Patterns you noticed
  • Anything that frustrated you
v1, May 3, 2026. Update with Day 1 learnings.