The pitch claims ~80% gross margin because it counts phone minutes and ignores the model. Drive the assumptions yourself. The moat is the yellow sliver.
Meela is $20. Medical Guardian is $32–47 with a pendant.
Per minute of live voice. $0.05–0.10 cached & tuned; $0.18–0.46 naive.
Shared 24/7 human escalation. Set to $0 to see the “no human” version.
Dashboard, storage, SMS, CS. Per parent, per month.
Caps discretionary chat only — never the daily check-in.
Untick to pay Twilio retail. Watch how little moves.
Cost to serve the average parent
One bar, to scale. The pitch is built on the yellow sliver.
—
Cost to serve, by how much she actually talks
Everyone gets the same 150-minute daily check-in. What varies — wildly — is the chat. The lonelier the parent, the more the product costs and the more she loves it.
—
| Segment | % of base | Check-in min | Chat min | Inference | Telephony | Desk | Infra | COGS | Margin |
|---|
Owning Fibernetics saves a couple of dollars a parent per month. That is worth having and it is not a strategy. The AI is the business’s cost structure, and it is rented — from OpenAI, Google or Anthropic, whoever you pick. Any deck that claims 80% margin is counting the minutes and ignoring the model.
The spec promises unlimited, endlessly patient tech help and that Iris never hangs up first. Both are lovely. Both are uncapped commitments on a metered input — the customer who loves the product most is the one you lose the most money on. Pull the fair-use slider and watch the tail come back.
It never touches the daily check-in — that stays unlimited and untouched, because it’s the ritual and the proof of life. It only bounds open-ended chat, where almost no one legitimate ever reaches the ceiling. You keep the promise that matters and you lose the one that bankrupts you.