Human in the loopGuard

Task detection and escalation

Reminders users ask for in plain words, followed up until they're done.

When a user writes “remind me to call the pharmacy at 6pm”, Humane extracts the task and its time without a separate API call. Open tasks climb a five-level ladder, firing reminder and escalation events until someone completes or snoozes them.

Tasks · user_42

remind me to call the pharmacy at 6pm

engine.task.auto_created

Call the pharmacy · remind_at 18:00 · escalation_level 0

  1. L1

    18:00

    Due now · first nudge

  2. L2

    +5 min

    Overdue · firmer reminder

  3. L3

    +20 min

    Persistent · flagged for a person

  4. L4

    +50 min

    Critical · escalation event

Illustration of task detection and escalation
How it works

From a sentence to a tracked task

The task detector runs on every process() call with its own time parser. Detected tasks come back on the response under auto_created_tasks. A sweep runs every minute and moves overdue tasks up the ladder until they're done.

  • L1 at the reminder time, then L2 after 5 minutes, L3 after 20, L4 after 50
  • proactive.task_reminder and proactive.task_escalation webhooks along the way
  • Snoozing pushes the time back and resets the level
  • Repeated top-level firings stop after a set limit instead of nagging forever

Proactive triggers

  • mood_dropmood < 0.30cooldown 1h
  • energy_lowenergy < 0.25cooldown 2h
  • trust_milestonetrust > 0.70, 10+ turnscooldown 24h
  • user_inactivesilent for 48hcooldown 24h
  • engagement_dropquiet 24h, mood < 0.5cooldown 12h

→ proactive.mood_drop + proactive.triggered · suggested_action included

Illustration of proactive triggers and their defaults
In code

See what's due, mark it done

python
r = client.process("user_42", "remind me to call the pharmacy at 6pm")
for t in r.raw.get("auto_created_tasks", []):
    print(t["title"], t["remind_at"])     # Call the pharmacy  2026-...T18:00

for t in client.list_tasks(due="escalating"):    # already at L2 or above
    print(t["escalation_level"], t["title"])

client.snooze_task("tsk_01HQ...", minutes=30)
client.complete_task("tsk_01HQ...")

Task questions

Something else? Ask us.

Every AI should know who it's talking to.

The safety and human-oversight layer for emotional AI: user state across sessions, risk alerts, safety gates and human takeover, with every decision explained.