Agentic commerce is bigger than the "Buy" button

Your car insurance is up for renewal. Your prescription needs a refill. You still haven't booked the flight to see your parents.

None of this is particularly difficult. It just requires an evening you don't have.

Now imagine telling your personal agent what you need, what information it may share, and what it can do without checking back. It compares coverage, prepares the application, requests the refill, and watches for a flight that meets your instructions.

That's the version of agentic commerce worth getting excited about. It reaches into the work of everyday life, including the things we put off because dealing with them is a pain.

For the companies receiving those requests, it raises an immediate question: can we legally accept this action from an agent, and can we prove the person authorized it?

Get that right and you have a new way to serve the customer. Get it wrong and you may have an application they didn't authorize, information they never meant to share, a commitment they will dispute, or even a regulatory or compliance action taken against you.

Your next customer may send an agent

The personal agent is becoming a reality.

Instinct is an assistant that connects to your applications and devices. It can handle tasks like arranging a ride to the airport or booking a handyman. You can call or text it.

Meta introduced Muse last week. It is already one of the most popular apps in the App Store and should soon have hundreds of millions of users. It’s a personal agent with its own browser and computer environment that can work across a person's apps. The user decides how much access to grant it.

These products point toward a different relationship with the internet. You describe the outcome, and software does more of the visiting, comparing, and submitting for you.

The traffic shift is already visible. Cloudflare reported that non-human traffic crossed 50% in 2026, overtaking human traffic in its reporting.

The assumption that a person is sitting behind every online visit is becoming harder to build a business around. One agent researching a task can make many requests while its owner makes none.

We're hearing insurance companies ask how they can let a consumer apply without ever coming to their website by allowing the consumer's agent to submit their information instead.

Think about what that changes. Companies will soon acquire customers who never navigate to their homepage, create a password, fill out a form, or sign any terms.

The fraud risk and the revenue opportunity arrive together

It's easy to see why a fraud team would worry about this scenario. An agent can submit stolen identity information or act beyond the permission its owner intended to give, all at massive scale.

It's just as easy to see why a growth team would pay attention. Adobe's Q2 2026 AI traffic research found that AI-driven traffic to retail sites grew 393% year over year in the first quarter. In March, AI-referred retail visits converted 42% better and generated 37% more revenue per visit than non-AI traffic. These results come from referrals; having agents autonomously complete the purchase would generate even more revenue.

When agents can actually finish the work, an airline could receive a booking when a fare meets a customer's preapproved conditions. An insurer could receive a completed application overnight, ready for its required checks.

Your customer sleeps. Their agent keeps working. You wake up to a new order or an application that might otherwise have stayed on someone's to-do list.

That outcome depends on the customer's permission and the rules of the transaction. But it makes the commercial choice tangible: how much legitimate business could you accept if you had a reliable way to distinguish an authorized request from a fraudulent attack?

Start with whether you can accept the action

Simply asking, "Can we trust the bot?" skips over an important question. Is there a real person behind the agent, and did they approve? Is the agent allowed to submit this information or make the commitment? What must be approved by the human themself?

U.S. law already contemplates electronic agents. The ESIGN Act says a contract cannot be denied legal effect solely because an agent was involved if action is legally attributable to the person behind the agent.

That is not blanket permission for an agent to do anything a person can do. The requirements of the particular activity, including applicable consent, eligibility, and authorization rules, still matter.

For the business, the practical task is establishing who the person is, what they authorized, and whether the requested action satisfies those requirements. If the customer later says, "I never applied for that credit card," you need more than the agent's word.

This is why we're building Proof's identity infrastructure around high-assurance digital identity, verifiable authorization records, and a network that supports ongoing fraud checks. That evidence can help businesses make and defend an acceptance decision. It cannot make an otherwise prohibited action permissible.

Think about everything people need to get done

Buying insurance. Filing taxes. Ordering prescriptions. Placing a bet where permitted. Applying for a loan, opening an account, booking a flight, or applying for government benefits.

Some involve an immediate payment. Others establish eligibility, secure access to a service, or create a financial obligation. Each has its own rules, and each involves something of value to the person asking for it.

That is the scope of commerce we should be thinking about. Agentic commerce goes far beyond payments. A shopping cart only captures a small part of a person's economic life.

Businesses that can accept legitimate, authorized actions through personal agents will soon expand their channels for new business. For consumers, the payoff is getting the things done without having to personally operate every website involved.

However, most agents stop short today. On the other side of doing the work, it could create a financial liability or result in the person appearing in front of a judge. “Who authorized this?” has become the predominant question in the age of AI. Answering this question will unlock trillions of dollars in economic activity for society.

Your next customer may never visit your site. Their agent still needs a way to do business with you. Explore Proof's approach to agentic commerce.