Stop checking the same number five times a day. Because the checking never stops until you make the machine do it for you.
| MCP protocol | Anthropic open standard, released November 2024 |
|---|---|
| ChatGPT Tasks | Recurring scheduling feature, launched January 2025 |
| Claude Cowork cadence | Hourly, daily, weekly, weekdays, or on demand |
| Free plan limit | 3 active scheduled tasks on ChatGPT Free or Go |
1. Name the one number you obsess over
Pick a single number that actually changes a decision for you, not everything you could track. It might be the USD to INR rate if you send money home, a flight price for a trip you have not booked yet, or the rent for a specific area you are watching. One number, not a dashboard.
2. Set the exact trigger, not a vague wish
Decide the specific value that would make you act, before you set anything up. Not ‘let me know if it’s good’ but ‘tell me if USD to INR crosses 84.50’ or ‘tell me if the flight drops below 32000 rupees’. A number an AI can compare against is a number it can actually check for you.
3. Give the AI a way to actually check it
Most modern AI assistants can now browse the web or connect to a live data source through a plugin, often called MCP, which just means the AI is allowed to run a small tool instead of only talking. Ask it plainly: ‘Can you check the current USD to INR rate for me right now?’ and confirm it gives you a real, current number, not a guess.
4. Schedule it to run without you asking again
Most AI tools now let you set up a recurring check, daily or a few times a day, the same way you’d set a phone alarm. Tell it in plain words: ‘Check this every morning at 9am’ and let the scheduling feature handle the rest.
5. Tell it to go quiet unless the trigger fires
This is the step people skip, and it’s the one that saves your attention. Say clearly: ‘Only message me if the number crosses my target. Otherwise say nothing.’ A tool that pings you every day is just a new notification to ignore, one that stays silent until it matters is the one you’ll actually trust.
6. Test it once by hand before you trust it
Before you walk away, ask it to run the check manually one time and show you the result. Confirm the number is current and the trigger logic is right, for example by temporarily setting the target close to today’s actual value so it should fire, then reset it to your real number.
A worked example
A Dubai based professional sends money to family in India every month and used to check the exchange rate obsessively on his banking app.
Before
Opens the currency app 4 to 5 times a day, sees numbers like 83.90, 84.10, 84.05, feels no wiser each time, and eventually transfers on a random day out of tiredness rather than a good rate.
After
Sets a target of 84.50 with the AI, gets silence for six days, then on day seven receives: ‘USD to INR is at 84.62 this morning, above your 84.50 target. This might be a good day to transfer.’ Makes the transfer that day, has not opened the currency app since.
Three ways to bend it
Job hunters can track a role, not a rate
If you are waiting for a specific role to reopen at a company you want, set the trigger to a job title and company name instead of a number. Ask the AI to check the careers page daily and message you only when that exact title appears.
House hunters can track a price, not a page
If you are watching rent or property prices in one area, give the AI the listing site and your maximum budget. Ask it to check daily and alert you only when a unit appears at or below your number, so you stop refreshing the listings yourself.
Investors can track a level, not a chart
If you follow one stock or fund, give the AI the ticker and the price level that would make you act, buy or sell. Ask it to check once a day and stay silent until that level is touched, instead of you watching the chart out of habit.
When it goes wrong
It messages me every single day even when nothing changed
Your instruction was probably ‘check and update me’ instead of ‘check and only message me if X happens’. Go back and restate the silence condition explicitly, AI tools follow the instruction you actually gave, not the one you meant.
It gave me a number that felt off or outdated
Ask it directly where the number came from and when it was last updated. If it cannot tell you, it may be guessing instead of checking a live source, ask it to use its browsing or tool feature specifically rather than answering from memory.
The scheduled check just stopped running after a few days
Some free tools pause recurring tasks after a while or need you to reconfirm them. Open the tool once a week for the first month to check the schedule is still active, until you trust it enough to leave it alone.
When not to use this
This is not worth setting up for a number that changes only once a month, or a decision you are not actually close to making. If you are not really planning to transfer money, book a flight, or make an offer soon, you are just building a toy that pings you for no reason. Set this up for the one decision that is actually live for you this month, not every number you are mildly curious about.
What you can now do
You stop being the one who checks, and become the one who gets told.
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
- Write down the single number and the exact trigger value on paper before opening any AI tool, so you are not designing the rule on the fly.
- Ask the assistant directly where its number came from and when it was last updated, and require it to use a browsing or connected tool rather than answering from memory.
- Set the recurring check with an explicit stay silent unless condition, then test it once by temporarily moving your target near today’s actual value to confirm it fires.
- Reopen the tool once a week for the first month to confirm the scheduled check is still active, since some free tiers pause recurring tasks quietly.
- Skip this setup for any number you only check out of curiosity rather than a decision you are actually close to making this month.
Questions people actually ask
How do I get an AI to check a price and only message me when it hits my number?
Give it three things: the exact number or event that matters, a recurring schedule like once a day, and a clear rule to stay silent otherwise. For example, tell it to check a flight price every morning and only message you if it drops below your target, saying nothing on every other day.
What is MCP and do I need to understand it to set this up?
MCP, short for Model Context Protocol, is an open standard from Anthropic that lets an AI assistant connect to live tools and data sources instead of only generating text from memory. You do not need to understand how it works, you just need to confirm your assistant can browse or use a connected tool before you trust any number it gives you.
Can ChatGPT or Claude actually run a check every day on their own?
Yes, both now support recurring scheduled tasks you set up once in plain language, such as a daily or weekly check, and the assistant runs it automatically afterward. Limits and exact options vary by plan and by provider, so confirm the schedule saved correctly and re-check it occasionally rather than assuming it will run forever unattended.
Why does my AI alert message me every day even when the number has not changed?
This almost always means the instruction you gave was something like check and update me rather than check and only message me if the target is hit. Go back and restate the silence condition explicitly in plain words, since these tools follow the literal instruction given, not the intent behind a vaguer one.
How do I know if the number an AI gives me is real and current, not a guess?
Ask directly where the number came from and when it was last updated. If the assistant cannot point to a live source or timestamp, it is likely answering from memory rather than checking anything, so ask it to explicitly use its browsing or connected tool feature and confirm the result again before you rely on it.
Sources
- Introducing the Model Context ProtocolAnthropic
- What is the Model Context Protocol (MCP)?Model Context Protocol
- ChatGPT now lets you schedule reminders and recurring tasksTechCrunch
- Schedule recurring tasks in Claude CoworkClaude Help Center
What happens next
Expect scheduling and live tool access to keep expanding across assistants, from simple web checks toward direct connections into calendars, banking apps and job boards through standards like MCP. The open question worth watching is reliability over weeks rather than days, since a scheduled check that quietly stops running is the main way this kind of setup currently fails people.
