Tested with the 3AEA™ Framework. The split shown is an example, not a result.
Then measure whether it paid off.
Count your starting point before you turn anything on.
Use the setup for 30 days.
Check in and compare what you expected with what actually changed.
Decide what happens next.
ExpandChangeStop
Starting pointAfter 30 days
Hours on one job. Example only.
Start with zero delegation. Add only the access and authority that produce measurable value.
The question
Five checks
Did it pay off?
How much of your business can you safely hand to AI?
AI can save you time. The question is what you give it in return. An AI may need access to your calendar to schedule an appointment. It does not necessarily need authority to cancel one.
Approval is not the same as protection.
Every job below is tested with the 3AEA™ Framework: Access, Authority, Approval, Exposure, Alternative. ?Five questions behind every suggestion: what AI must see, what it may do, when you must say yes, what could go wrong, and whether a simpler way works.
Have a check-in code??The check-in code is the long string of letters and numbers this page gave you when you pressed "Save my starting point" at the end of your first visit. It holds your starting hours, rates and counts. Paste it here about 30 days later to see what the change actually saved. If this is your first visit, you do not have one yet. Start with your goal below.
Check in: did it pay off?
Paste the check-in code you saved last time ?The code this page gave you when you saved your starting point., or use the starting point saved on this device.
Your goal
Your jobs
Pick the jobs that eat your week. For each one, see how much of your time is tied up in it, which steps AI can take on, which steps stay with you, and what not to share with the AI setup you choose. Then pick a setup that gives AI the least access that still gets the job done.
For each job, enter how many weeks a month, how many days a week and how many hours a day it takes, what an hour of your time is worth, and what hurts about it today.
Where your time goes, and where AI can help
HighMediumLow shows how much of each job AI can likely take on. ?High means AI handles most of this job well. Low means a simpler fix usually wins. These are starting guesses. These levels are rules of thumb written by Suneeta Modekurty. They have not been measured on real businesses yet. Your own hours and rate decide the dollars.
Every recommendation is tested using the 3AEA™ Framework: Access · Authority · Approval · Exposure · Alternative.
Now let's make sure it actually pays off
A recommendation is only a guess until you measure it. The next step sets what you will count before you turn anything on, so AI only gets credit for what it actually changes.
Before you start, track these three things
For each job, pick the option you will start with, say how many hours you expect to get back, and start counting. If you don't know a number today, leave it blank and start counting this week. That count becomes your starting point.
Before you sign up for any AI tool
Some AI tools ask for a card when you sign up. Meta's Muse reportedly places a temporary charge of about $1 to confirm your age. That check shows the card works. It does not show who is using it, or how old they are.
Turn on alerts for every transaction on your business card, including $1 charges. Many banks and card apps let you set the alert amount as low as you like. A small charge you don't recognize can mean someone opened an account with your card.
Decide who in your business may open an AI account with the business card. Then you'll know which alerts to expect.
How your 30-day check will add it up
Cash the change returned = costs you eliminated + profit on added sales you can attribute − what the new setup costs
Time the change returned = hours you got back − your time checking its work
Time is not cash. Hours you get back become money only through what you do with them, so the page reports them separately.
Come back in 30 days. We'll compare what you expected the change to save with what it actually saved, and tell you whether to expand, change or stop.
What it's called, and what it actually checks
AI tools name their safety features in reassuring words. The feature is usually real. The gap is between what the name suggests and what it actually checks.
What it's called
What it actually checks
What to do
"Approval" Muse asks before it buys, sends or posts.
Only the final step. Not what the tool reads, what it remembers, or what your chats teach Meta's AI.
Set the default to "Always ask," connect accounts so it can only look, and turn off training in Data controls.
"Always allow" One of the buttons on every approval screen.
Nothing. It creates standing permission for that kind of action on that app until you change it.
Never tap it on anything that sends, posts or spends. Review your allowed actions once a month.
"Age verification" A card check at sign-up, reportedly a temporary charge of about $1.
That a card number works. Not who is using it, or how old they are.
Turn on alerts for every charge on your business card, and decide who may open AI accounts with it.
"Your card stays private" Muse pays with a one-time card number from Link by Stripe.
That the merchant never sees your real card number. Not how much gets spent over a month.
Set your own spending limits in your ad account and with your bank, rather than asking the agent to stay under a number.
We are not saying these features are bad. We are saying each one checks less than its name suggests. New lines get added as we look at more tools.
How each job is judged: the 3AEA™ framework
A₁ Access
What must the AI see?
A₂ Authority
What is the AI allowed to do?
A₃ Approval
When must a human authorize its actions?
E Exposure
What data, business information, or consequences become exposed?
A Alternative
Is there a simpler way to get the same result with less access, authority or exposure?
3AEA™ is the method underneath this page, by Suneeta Modekurty. It runs as a loop: Find → Decide → Baseline → Track → Adjust → Remeasure.
Supporting a group of businesses?
3AEA™ can also show anonymized patterns in where businesses are saving time, where AI requires unnecessary access or authority, and which lower-exposure alternatives are working.
METRIS™
Discuss an AI decision →