Pro AI feature
Promise Tracker.
Every commitment, in every meeting, captured automatically — and tracked to completion. With the exact quote, the speaker, the deadline, and a tap to play the moment it was made.
Send the Q4 budget breakdown
Due Friday"I'll get the breakdown over to you before the end of the week."
Loop in legal on the consent question
Due Mon"I'll send legal a quick note about the recording stuff and copy you."
Pull the Acme retention numbers
Kept · Wed"Kept" — message in Slack 9:42am with the numbers attached.
The problem you've been carrying in your head.
You finish a meeting. Someone said they'd send you the report by Friday. You said you'd loop in legal on the pricing question. Sarah committed to the marketing brief by next Wednesday.
You jot some of these down. You miss some. By Friday, the report hasn't arrived and you can't quite remember whether the deadline was Friday or Tuesday or "before EOW." By Wednesday, Sarah hasn't sent the brief and you're not sure whether to nudge her or let it slide.
Your brain is doing this work. Inefficiently. With error.
Promise Tracker is the feature that exists because that work shouldn't be your brain's job.
How Promise Tracker works.
- 1
Record the conversation.
Promise Tracker works on every recording — in-person meetings, calls captured via Mac Catalyst, conversations recorded on the fly. No special meeting setup, no bot, no calendar invite required.
- 2
Commitment detection.
After transcription, an on-device classifier scans every utterance for commitment-shape phrases — "I'll send", "we'll get back to you", "by next Wednesday", "let me handle that". Each detected commitment is anchored to the speaker, the timestamp, and the recording.
- 3
Entity resolution.
Each commitment is parsed into structured form:
{speaker, recipient, action, deadline}. Speakers are matched against your existing speaker library; recipients are inferred from conversational context; deadlines are normalized into actual dates. - 4
Trackable cards.
Every commitment becomes a card you can filter — owed to me, owed by me, open, kept, missed. Sort by deadline, by speaker, by project. Tap any card to jump to the exact moment in the recording where the promise was made.
- 5
Auto-detect follow-through.
When the promise is kept — Sarah sends the brief, the report lands — Bonfiyah detects it. Future recordings, calendar events, and (Phase 2) email integrations confirm completion automatically. No manual checkbox required.
- 6
Reliability scores over time.
Pro AI builds a promise reliability score for each recurring contact — what percentage of their commitments do they actually keep? Surfaced on every contact's People Memory page. Calibrate expectations on the people you work with most.
Who Promise Tracker is for.
Anywhere stakes are paid in commitments.
Managers
Run 1:1s with the receipts.
Stop being the one who has to remember every team commitment. Promise Tracker captures who said they'd ship what, by when. Walk into 1:1s and stand-ups with the prior week's promises pre-loaded.
Sales teams
Catch the next-step quote.
"We'll have budget approval by month-end" → tracked. "I'll loop in procurement" → tracked. Reference the prospect's exact words in your follow-up email two weeks later.
Attorneys
Witness commitments under oath.
Every commitment opposing counsel makes on the record gets a tracked card with timestamp + audio. When they walk it back three weeks later, you have the source quote with a chain-of-custody hash.
Founders + execs
Investor + board commitments.
Cap-table reviews, board meetings, investor coffees — the commitments people make in those rooms shape your runway. Promise Tracker surfaces them; you don't reconstruct from memory.
Privacy
Where the inference runs.
On-device. Promise Tracker uses an on-device transformer that ships in the app bundle. Your transcripts do not leave your iPhone for this feature — and the recordings themselves don't either, unless you explicitly enable iCloud sync.
The reliability score, the People Memory profile, the deadline normalization — all happen locally. Promise Tracker is built so the privacy posture isn't a feature trade-off.
FAQ
What counts as a commitment?
Any utterance with the linguistic shape of a commitment: future-tense action plus a stated or implied subject ("I'll send", "we'll have"), often with a deadline marker ("by Friday", "before the meeting"). Confidence scoring lets you tune sensitivity — high-confidence commitments are auto-tracked; low-confidence ones flagged for your review.
Does it work in conversations with multiple speakers?
Yes. Speaker diarization runs on every recording, and Promise Tracker resolves each commitment to the speaker who made it. Persistent speaker profiles mean "Sarah" in last week's recording is the same person in this week's.
Can I export tracked promises to a task tracker?
Yes — Apple Reminders, Things 3, and Todoist via direct integration; Markdown / CSV for everything else. The export carries the action, owner, deadline, source quote, and a deep link back to the audio moment.
What if I disagree with how a commitment was extracted?
Tap to edit the action text, change the owner, adjust the deadline, or delete the card entirely. Edits feed back into the on-device model so future extractions on the same device improve.
See Promise Tracker on a real recording
We'll send a 90-second screencap of Promise Tracker running over a real (anonymized) sales conversation — including the moment it caught a commitment the rep had missed.
Send the exampleOpens your email app. We respond manually until our newsletter system is fully set up. See our privacy policy.