You already know the problem. The school sends a newsletter on a Friday afternoon. The dentist sends an appointment confirmation that mentions a separate follow-up. The summer camp sends a registration receipt that buries the first day inside paragraph three. Each of those emails is, in theory, the moment you should add a date to the family calendar. In practice, you do not. You read it on your phone in the pickup line, tell yourself you will type it in later, and by Sunday it has been pushed off-screen by sixty other emails.
That intake step - the one where you read an email and translate it into a calendar entry - is where the system breaks. You have tried Google Calendar. You have tried Cozi. You have probably typed picture day into a shared calendar at least once. The calendar is not the problem. The act of getting things into it is. Email-to-calendar exists to remove that step. This is how it actually works.
The four-step pipeline
When people ask how email-to-calendar works, the honest answer fits in a paragraph, but it is easier to see in four steps.
- You forward an email to a personal Sense address you get when you sign up. It looks like a normal email address. You can send to it from any inbox.
- An AI reads the prose. Subject line, body, signature, and any PDF attachments. It looks for dates, times, and the words around them.
- It classifies what it finds. Each date gets sorted into one of three buckets: an event with a start and end time, a deadline that becomes a reminder, or a save-the-date with no firm time yet.
- The calendar updates. The new entries appear on your shared family calendar within about a minute. You get a short notification that says what was found.
That is the entire flow. The reason it feels different from "just type it in" is that the part that fails you - finding the picture-day sentence on page four of a six-page newsletter at 4:47pm on a Friday - is the part the AI is doing.
What the AI is actually doing when it reads
You can think of it like this. The AI is reading every sentence the way a person would, but without ever skimming. It is looking at three things at once: the date or time itself, the noun next to it, and the verb that connects them.
Take a sentence like "Permission slips are due Monday morning." The date is Monday morning. The noun is permission slips. The verb is "are due." That combination tells the AI this is a deadline, not a calendar event. It writes a reminder for Monday morning, not a 9am meeting called "Permission slips."
Take another: "Crazy Hat Day is Friday." Same structure, different verbs. "Is" with a noun like Crazy Hat Day signals an all-day event. So Friday gets a full-day entry with the right title, not a reminder.
Take a harder one: "Save the date for the spring concert, more details to follow." The date might be specific or vague. The phrase "save the date" tells the AI to make a tentative event with a note that the time is not yet confirmed. When the follow-up email arrives, the AI will recognize it as the same event and update it instead of creating a duplicate.
None of this is magic. It is pattern recognition over millions of sentences written exactly the way schools, coaches, dentists, and camp directors write them. The reason it works well on family logistics in particular is that these emails follow predictable shapes. A field trip permission slip reads like a field trip permission slip. The dentist's appointment confirmation reads like every other dentist's appointment confirmation. The AI has seen this prose before. It knows which date is the appointment and which one is the cancellation deadline.
What it does not do, and why that matters
This is the part you actually care about, and it deserves to be said plainly.
Sense does not connect to your inbox. It does not log into Gmail. It does not have a token that lets it read your other email. There is no OAuth checkbox where you hand over read access to everything you have ever received. The model is push, not pull. You send Sense an email. That is the only way anything gets in.
That is intentional. The trust ask of "let an AI read your whole inbox" is a different trust ask than "let an AI read this one email I just forwarded." We do not think the first one is fair to ask for. So the architecture matches: Sense reads exactly what you forward, and nothing else. If you forward the school newsletter and not the dentist confirmation, the AI sees the newsletter and not the confirmation. If you stop forwarding, the flow stops.
After the AI reads a forwarded email, the events and reminders go onto your family calendar, and the original email is summarized and stored in your private Sense account so you can scan it later in one place. The content is not used to train AI models, and it is not shared with any other Sense family. If you delete a forwarded email from Sense, the extracted events and the stored copy go with it.
You should also know what the AI does poorly. Dates inside flat images with no caption or alt text are invisible to it - if a school posts a graphic with the book fair dates baked into a JPG and nothing else, you will still need to scan that one yourself. Handwritten notes pasted into an email as a photo are hit-or-miss. And a phrase like "next Tuesday" that you forward six weeks after the email was sent will sometimes land on the wrong Tuesday. The AI is good. It is not infallible. Sunday-night calendar review is still a five-minute habit worth keeping.
The trust model, said honestly
If you have ever said the words "I don't trust AI with my family's stuff," that is a reasonable thing to have said, and we are not going to argue you out of it. We will tell you what the model is, and you can decide.
The model is: you control the input, we read only what you send, the output lives on your calendar where you can edit or delete anything, and the original content is not used to train anyone's AI. Email-to-calendar is not "AI watching your inbox." It is closer to "a very fast assistant who only sees the emails you hand them." If that assistant did a bad job, you would notice on the calendar in less than a minute, and you would stop forwarding. There is no lock-in to the flow.
The reason this matters more than the technical details is that the alternative - reading every school email yourself, on your phone, in the school pickup line, and remembering to translate it into a calendar entry that night - is the system that has been failing. Trying email-to-calendar on one newsletter does not require you to trust a category. It requires you to trust one email.
Forward one email and see what lands on the calendar
The fastest way to know whether email-to-calendar is something you want in your family is to try it on one email you already have in your inbox.
Common questions
Does Sense read my whole inbox?
No. Sense does not connect to your inbox at all. You forward individual emails to a personal Sense address. The only emails the AI ever sees are the ones you specifically push to it. If you stop forwarding, the flow stops.
How does it know an event is an event and a deadline is a deadline?
It reads the sentence around each date. "Picture day is Tuesday at 9am" reads as an event. "Permission slips are due Monday" reads as a reminder. The AI is classifying each date by the verbs and nouns around it, the same way you would if you were transcribing the email by hand.
How fast does the calendar update?
Usually within about a minute of you hitting forward. A long PDF attachment with a full season's worth of sports games can take a little longer because there is more text to parse, but a single school newsletter or appointment confirmation is essentially instant.
What happens to the email after the AI reads it?
The events and reminders go onto your shared family calendar. The original email is summarized and stored in your private Sense account so you can scan it later in one place. Forwarded emails are not used to train AI models and they are not shared with any other Sense family. If you delete the email from Sense, the extracted entries go with it.
What if the AI gets something wrong?
You edit or delete the entry on the calendar the same way you would any other event. There is no penalty and nothing locks in. Most parents skim the calendar on a Sunday night, fix the one event the AI got slightly off, and leave the rest. The win is not perfect extraction. The win is that the events made it onto the calendar at all.
If you want a closer look at the product side of this, including attachments and family-member assignment, see How Sense AI reads your emails. For a real test of the same pipeline on a six-page school newsletter, see Can AI actually read a school newsletter?