At 8 a.m., the morning brief from an AI assistant lists three “priorities for today”: finish the client presentation, reply to the supplier about rescheduling, and prepare a packing list for next month’s business trip. It looks like a clear to-do list.
But the three items are completely different in kind. The client presentation is something you promised by email last Friday, due by 3 p.m. today. The supplier only wrote to ask whether a date could move; you haven’t agreed to anything. The packing list is a suggestion the AI came up with after seeing a flight confirmation; nobody is waiting for it. Only the first is a commitment.
In May 2026, Google launched Daily Brief for Gemini: once a user opts in, it gathers urgent updates from Gmail, tracks upcoming Calendar events, prioritizes based on the user’s goals, and suggests next steps. Features like this really do save time digging through email. But ranking answers “which item does the model think is worth seeing first,” while commitment status answers “who is waiting for you to do what.” Merge the two into one rank, and you may treat a request you never agreed to as work you must finish today.
What ranks first isn’t necessarily today’s work
When a brief puts things in order, it’s natural to work through them in that order. The problem is that ranking reflects urgency, relevance to goals, and timing, and none of these equals “you agreed to it.”
An email marked “urgent” doesn’t mean you committed to its deadline. A meeting on your calendar doesn’t mean you accepted the invitation. Google Calendar’s own data distinguishes attendee responses: not yet responded, tentative, accepted, declined. An invitation still marked “not yet responded” that the brief ranks first looks just like a settled appointment.
One brief mixes three states
Committed. You accepted the meeting, created a task with a deadline yourself, agreed by email to deliver something, or were named as the owner in a team system. These items should be able to point to who it’s for, what’s delivered, and where the deadline came from.
Pending. A colleague asks “can you take a look today?”, a client proposes a new date, a calendar invitation hasn’t been answered. These are requests; they don’t automatically become your work. Their next step is only one thing: reply or negotiate.
AI suggestion. Based on your goals, past conversations, or free slots in your calendar, the model reminds you that you could do something. The suggestion may be good, but that doesn’t earn it today’s working hours. If it was generated from old preferences or outdated context, first check that context still applies; see Once AI Remembers Your Opinions, It’s More Likely to Agree With You.
Ask about status before order
When you open the brief, don’t start by dragging items into a new order. First find the reason each item became work. In practice, require every item to carry a few fields: a source you can open (which email, event, or task), its current status, who is waiting, where the deadline comes from, and the next action a person needs to take. “Based on your recent activity” isn’t a source.
With status added, the opening brief means something different: the client presentation goes into today’s deliverables; the supplier’s rescheduling goes into pending, and today it needs only a reply; the packing list stays in suggestions. It looks like just a few extra fields, but it keeps two things that aren’t yet real from taking time away from work you’ve actually promised.
You can also set a small quota for “today’s non-meeting deliverables,” say three. The number isn’t a rule; it’s there to make conflicts visible. If checking shows seven committed items, the problem is that you’ve agreed to too much. The fix is to move deadlines or negotiate with someone, not to let AI quietly push the last four further down the screen.
When the evidence contradicts itself
Judging status depends on evidence, and evidence sometimes conflicts: the email says Tuesday, the task card says Wednesday. In that case, AI should mark it as “missing,” leave it in pending, and not pick the date that looks newer as the commitment. The AI Risk Management Framework from the U.S. National Institute of Standards and Technology (NIST) cautions that representing complex human situations as numbers a model can handle can remove necessary context, and that human and AI roles in decision-making need to be clearly defined. For a daily brief, the minimum division of labor is this: AI gathers and groups, and a person confirms whether a commitment exists.
Also, connecting more accounts doesn’t necessarily make the brief more accurate. Turn on only the sources the brief really needs. If a work or confidential account doesn’t meet your organization’s rules, don’t connect it yourself just to get a more complete brief.
The next time you read a morning brief, ask one question about each of the top three: did I agree to this, did someone ask for it, or did AI suggest it? If the brief is handed to someone else, the same fields help the next person see who’s responsible and when things were agreed; see A Tidy AI Summary Does Not Mean a Teammate Can Take Over.
AI handoff card
I want my AI daily brief to separate what I've actually agreed to from requests and suggestions.
The method comes from this article: https://boosterminiclass.com/en/posts/ai-daily-brief-commitment-states/
Its view: a brief's ranking answers "what's worth seeing first," not "who is waiting on me." Mark each item's status first: committed (I accepted the time, deliverable, or recipient, or I was named as owner), pending (someone asked and I haven't agreed), or AI suggestion (inferred from goals, past conversations, or free time). Then decide what to do today.
Only read the calendars, tasks, and messages I've authorized. Don't create, edit, reply to, accept, or delete anything.
For candidate items today and in the next 48 hours, list for each:
- A source I can open (link or ID of the email, event, or task), my response status, who is waiting, and the deadline as written plus where it comes from;
- Status: committed / pending / AI suggestion, with the evidence. If the source, response status, or deadline basis is missing, mark it "missing" rather than guessing;
- The next action for me (deliver, reply or negotiate, or set aside for now).
Then point out schedule conflicts. From committed items, pick at most three non-meeting deliverables; if there are more than three, list the items whose deadline or scope needs negotiating.
Provide only a brief I can verify. Don't take any action for me; what gets done today is my decision.
References
- Google Blog: The Gemini app becomes more agentic, delivering proactive, 24/7 help — https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/ (2026-05-19)
- Google for Developers: Events | Google Calendar API — https://developers.google.com/workspace/calendar/api/v3/reference/events [accessed: 2026-10-11]
- NIST AI Resource Center: Appendix C: AI Risk Management and Human-AI Interaction — https://airc.nist.gov/airmf-resources/airmf/appendices/app-c-ai-risk-management-and-human-ai-interaction [accessed: 2026-10-11]
