All use cases

Prevent / Care

Prediction why-isnt-it-showing deflect and escalate

"My ovulation prediction isn't showing" is the message a cycle app hears most — most are the same fixable reasons, but a handful are "look at MY cycle specifically," and treating those two the same is where it goes wrong.

Next best thingClose the loopBehavioral triggerRisk triggerChatIn-appWellness (B2C apps)Health & Telehealth

What it does

The single most common support message in a cycle-prediction app is some version of "why isn't my ovulation prediction showing?" Most of those have the same general answers (not enough logged data, an irregular cycle the model needs more of, a setting toggled off), and they flood the human queue today. But a meaningful slice are profile-specific: "my cycle is this long, I logged these things, check why MINE isn't predicting." Answering that generically is both useless and unsafe, because it needs someone to actually inspect her history. The agent resolves the common version instantly and recognizes the profile-specific one as a hand-off, deflecting the volume while routing the cases that genuinely need a human to look.

How it works

  1. 1
    Trigger. a user asks in chat why a prediction (ovulation window, period date) is not showing or looks wrong.
  2. 2
    Decision. the agent classifies the ask — is this a general "why can this fail to display" question it can answer from the knowledge base, or a profile-specific request to inspect *this* cycle and *these* logs? Specific dates, lengths, and "check mine" phrasing tip it toward escalation.
  3. 3
    Action. general cases get the plain reasons and the self-serve fix (log more, check the setting, what an irregular cycle needs); profile-specific cases route to a human who can open her history, with the conversation and her stated details attached.
  4. 4
    Follow-through. the agent confirms back to the user what is happening ("I've passed your details to someone who can look at your cycle directly"), and records which branch fired so the deflect/escalate split stays measurable.
Example
UserWhy isn't my ovulation prediction showing this month?
AgentUsually it's one of a few things: the app needs a bit more logged history, a very irregular cycle takes more data before it can predict confidently, or the ovulation insight is switched off in settings. Want me to check that setting with you and show what to log next?
UserNo, my cycle's been really long lately, I logged my temps and an LH test, so check why MINE specifically isn't predicting.
AgentThat's the right call to have looked at directly — I can't inspect your cycle history myself. I've sent your details and this chat to a team member who can open your data and follow up with you here.

Configuration

How the agent is wired for this use case.

Triggera prediction-related question in chat ("why isn't my prediction showing / why is it wrong"), arriving from the in-app messaging channel
Tools & actions
  • Messaging channel · receive the question and reply in-thread
  • Knowledge base · answer general "why a prediction may not display" cases and the self-serve fixes
  • App backend (settings / logging state) · check generic, non-profile factors such as a toggled-off insight or insufficient logged data
  • Support queue · route profile-specific requests to a human with the chat and the user's stated cycle details attached
  • Memory store · record which branch fired (deflect vs escalate) to keep the split measurable
Autonomythe general-deflect branch runs unattended with judge-gated replies; the moment a request becomes profile-specific (inspect *my* history, *my* dates) the agent stops self-serving and hands to a human — it never guesses about an individual's cycle data
Channelschat · in-app
Escalationany profile-specific "check my prediction / my cycle" request, or distress in the thread, routes to a human who can inspect history, with full context attached

What you need

The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.

Signals

prediction-related chat messages, prediction-display state for the user

Data

knowledge base of generic prediction-display reasons and fixes, app settings and logging-completeness state, the user's stated details in-thread, persistent conversation memory

Guardrails

a hard line between generic guidance and profile-specific interpretation — the agent never reasons about an individual's cycle data, it routes it; judge gating on every reply; distress or health red flags in the thread escalate; audit trail on the deflect/escalate decision

Metrics it moves

  • ticket-deflectionup, as the repetitive generic version of the question resolves in chat instead of queuing for a human
  • safe-escalation-rateup, because profile-specific requests reliably reach someone who can inspect history rather than getting a useless generic answer
  • contact-ratedown on the routine version while the cases that need a human still get one

Related use cases

Prevent / Care

Cycle-anomaly health flag with safe escalation

the proactive sibling that flags a logged anomaly before she asks

Prevent / Care

Femtech community moderation

the same intent-classification judgement applied to community posts

Prevent / Care

Clinical red-flag detection and safe escalation

the shared detect-and-route safety posture for health-sensitive cases

See it on your own customer journey

Bring one drop-off, one churn cliff, or one silent segment. We will show you what a proactive agent with memory and judgment does with it.

Book a demo