Your AI agent needs its own identity
Why an AI agent needs its own phone, e-mail and WhatsApp, like a secretary, and what is still missing for it to act with your approval
Grok Bot now has an e-mail. According to the announcement the bot can use that email to sign up for something, talk to a company on your behalf, or even schedule something with a person.
Seems like not much? Exactly. The agent will be able to have an autonomy you can barely imagine. It will not need to sign into your account to do something for you. It will have its own details to communicate with other people.
There is no point in giving your account to the agent
Having the agent logged into your account is the most used way today. Your e-mail, your WhatsApp, your social networks. Everything it does will be tied to your user, and your routine will get mixed with its tasks.
The solution is to give it its own details so it can interact with other people. You keep your accounts and it keeps its own. That is how it gets its autonomy.
It is like a secretary
You do not give your phone to your secretary. She has her own phone and e-mail, and you know that the person talking to you is her.
With the agent it will be the same. Give it a phone number, an e-mail and WhatsApp. It will contact the company, and when it has an answer that needs your decision, it will bring that answer to you.
My agent’s name is Claudia Romano
She has a first name and a last name. Claudia Romano. She has a phone number, an e-mail in her name, and a WhatsApp with a profile photo. People think they are talking to my secretary.
The name helps create that bond with the other person. She has her profile, her photo, and the replies will come in her name. Whenever she needs my approval, she brings it to me.
Claudia is already in some WhatsApp groups with suppliers. She talks, carries things out, and looks up information to help with the demands between the two sides. A secretary working for an accounting firm.
Where you approve is the limit
With the agent working in your name, you can gain hours in your routine. Asking for a refund is one example. You approve it and it will handle the process with the company.
Anything that is in your name or involves your money, it will ask for your approval.
Card and bank account
The agent will also be able to have a card and a bank account. Whenever it needs to spend, the agent will ask for your approval.
There is already a Stripe product called Link. It does this job of asking for approval before spending. Giving agents the ability to pay.
In it you say how much the agent can spend, and when it creates the payment it will ask you for permission. Then the agent receives a card that can be used only once, or a token. It will never have access to the card number. The only thing left is for them to create an account for the agent.
Today my main problem is that the bot still cannot create boletos for me. A boleto is a Brazilian bank payment slip. When it can do that, one more area of my life will be taken care of.
Repositories will have access too
The same goes for repositories. Today we can use the agent on your account. But in the same way it will do everything in your name.
It will have its own account, and you will grant what it is allowed to run.
SaaS products need to open this door
For the bot to work on its own, SaaS products need to let it have an account. Permission grants so it can carry out actions, and APIs so it can get to the data.
Until they have a screen that allows the agent to be used, it will have to log into your account and click where you would click.