Stop forgetting the promises you made in email. Because broken promises quietly cost you trust, and this catches them before anyone else has to remind you.
| Connection protocol | MCP (Model Context Protocol) |
|---|---|
| Scan window | Sent folder, last 7 days |
| Recommended cadence | Weekly, Friday morning |
| Trust erosion group | 87% of employees studied, 2025 workplace trust research |
1. Connect AI to your Sent folder
Most AI assistants can connect to real tools through a method called MCP, which just means giving the AI something it can actually use, not just talk to. Connect it to your email’s Sent folder specifically, nothing else. This keeps the setup small and safe.
2. Ask it to hunt for promises
Give it a clear prompt: ‘Scan my sent emails from the last 7 days. Find every promise I made, phrases like I’ll send, let me check and get back to you, or will share by. List each one with who it is for and the deadline.’ Be specific about the phrases so it does not miss casual promises.
3. Make it output a simple table
Ask for one column for the promise, one for the person, one for the deadline, one for status. A table is easy to scan in ten seconds, which is the whole point.
4. Run it every Friday morning
Set a recurring reminder on your calendar or phone to run the check once a week, right before your week resets. Friday morning works well because you can still send a quick note if something slipped.
5. Cross off what’s done, carry the rest
Each week, mark finished promises done and let AI carry the open ones into the next list. Nothing gets lost between weeks, and you always know exactly what you still owe people.
A worked example
A marketing manager sends dozens of emails a week and casually promises things like ‘I’ll send the deck by Friday’ without ever writing them down anywhere else.
Before
Scattered across 20 sent emails: ‘I’ll send the deck by Friday,’ ‘let me check with the vendor and get back to you,’ ‘will share the revised budget early next week.’ None of it written down anywhere.
After
Promise: Send Q3 deck to Priya — Deadline: Fri Aug 21 — Status: Open. Promise: Confirm vendor pricing with Rahul — Deadline: mentioned ‘today,’ Aug 18 — Status: OVERDUE. Promise: Share revised budget with finance team — Deadline: ‘early next week,’ needs a firm date — Status: Open.
Three ways to bend it
For salespeople
Scan sent emails and CRM notes for follow up promises made to prospects, like ‘I will send pricing by Monday,’ so no lead goes cold because you forgot to reply.
For managers
Point it at your sent Slack or Teams messages instead of email, since that is where most manager promises to their team actually happen.
For freelancers and consultants
Run it against client email threads specifically, since a missed deliverable promise with a client costs more trust than one with a colleague.
When it goes wrong
It only picks up obvious promises like ‘I’ll send X by Friday’ and misses vaguer ones.
Tell it to flag anything without a clear date as ‘needs a deadline’ instead of skipping it, so vague promises still show up on the list.
It flags polite phrases as promises, like ‘happy to help’ or ‘sure, no problem.’
Give it two or three real examples of what counts as a promise and what does not in your prompt, and it gets sharper fast.
I’m not comfortable giving AI access to my whole inbox.
Restrict it to the Sent folder only, searched by keyword, for the last 7 days. You are not handing over your whole inbox, just a narrow, repeatable search.
When not to use this
If your job runs on verbal commitments made in meetings or hallway conversations rather than email, this technique will miss most of what you actually owe people, use a notes app checklist instead. And if your workplace has rules against connecting AI tools to your email, check with IT before you set this up, not after.
What you can now do
You know exactly what you promised and to whom, before it becomes an apology email.
Part of Making Life Easier With AI, a weekly walk-through of one small thing AI can take off your plate.
Take this further
Act as a blunt hiring manager who has read ten thousand resumes, not a career coach. I will paste my full resume and the job description I want. Rewrite the whole resume for that role, section by section, in this order: summary, experience, skills, education. Rules: every experience line leads with impact, not duty. Use bracketed placeholders like [8 percent] for any number I did not give you, and list at the end every placeholder I need to replace with a real figure. Keep it to one page of text. Plain formatting only, no tables or columns, so screening software can parse it. After the rewrite, tell me the three weakest claims that need evidence before I send this anywhere. My resume: [paste resume]. The role: [paste job description].
Takeaways
- Set up an MCP connector scoped only to your Sent folder, not your whole inbox, before you run any prompt.
- Write the prompt with explicit phrases like ‘I’ll send,’ ‘let me check,’ and ‘will share by’ so casual promises don’t slip through.
- Put a recurring Friday morning reminder on your calendar to run the promise check every week without fail.
- Tell the AI to flag anything without a clear date as ‘needs a deadline’ instead of dropping it from the list.
- Give the AI two or three real examples of promises versus polite phrases so it stops flagging things like ‘happy to help’ as a commitment.
Questions people actually ask
How do I stop forgetting promises I made in emails?
Connect an AI assistant to just your Sent folder using MCP, then ask it to scan the last seven days for promise phrases like ‘I’ll send’ or ‘will share by.’ Have it output a simple table listing each promise, who it’s for, and the deadline, then check that table every Friday morning before your week resets.
What is MCP and how does it let AI read my email?
MCP, or Model Context Protocol, is an open standard that lets AI assistants connect to real tools and data sources instead of just chatting. Anthropic introduced it so applications like Claude can link to a Sent folder, calendar, or database through one consistent method, rather than needing a custom integration built for every single tool.
Is it safe to give an AI access to my whole inbox?
You don’t need to. Scope the connector to your Sent folder only, searched by keyword, over a fixed window like the last seven days. That narrow, repeatable setup limits what the AI can see, and most connector tools ask for approval before taking any action, so nothing gets read or sent without your say so.
How often should I check for broken email promises?
Once a week works well for most people. Running the check on Friday morning gives you a clear picture of what’s still open before the week resets, and enough time to send a quick follow up note if something slipped through. Mark finished promises done and let the list carry forward whatever is still open.
Can I use this for Slack or client work instead of email?
Yes. Point the same prompt at Slack or Teams messages if most of your commitments happen there instead of in email, which tends to fit managers better than salespeople. For client work, run it against email threads with that client specifically, since a missed deliverable promise costs more trust with a client than with a colleague.
Sources
- Introducing the Model Context ProtocolAnthropic
- What is the Model Context Protocol (MCP)?Model Context Protocol
- Gmail ConnectorAnthropic
What happens next
Expect more native email and messaging connectors to ship with the same read only, scoped by default permissions Claude’s Gmail connector already uses, which should make setups like this easier without a third party MCP server. Watch for AI assistants adding built in weekly digest or reminder features that could automate the Friday check itself, and for workplace IT policies that specifically address AI access to Sent folders and chat logs.
