Your next personal shopper could be AI — but if we don’t secure agent identities, digital commerce may drown in fraud and deception.
In agentic AI’s near future, we’ll all be generals commanding armies of agents on commerce missions across the digital landscape.
Today we use those agents primarily as research assistants, giving them a task to perform independently, whether it’s gathering information or monitoring for the latest data on targeted topics. In short order, though, the agents will start taking on more prominent objectives that involve moving money for their users, purchasing anything from a carton of eggs to an airline ticket.
But money movement crosses a critical line. It’s the trigger for Know Your Customer and Know Your Business requirements designed to stop fraud and money laundering. The same will be true for Know Your Agent (KYA).
The agents will represent businesses and people, so the goal will be to give those proxies specific identities that show where they came from and what they’re allowed to do. Certainly, there’s a need to verify the agent’s creator, which could be a person or business, but it’s also important to determine if it’s acting on behalf of someone else and if that person is a bad actor.
There are many different possible agentic connections, and they all need verification. Like any new technology, the possibilities are as plentiful as the pitfalls.
The need for KYA in digital commerce
A big part of agentic technology is capturing what the user would do and building that into the agent. A person, for instance, could create an agent for grocery shopping and set parameters for how much to spend, how often to buy household staples, what recipes to scan for ingredients and timelines for when food should be delivered.
The same could apply to travel planning. The user provides information about the destination, accommodations and budget, and the agent finds everything — including transportation, entertainment options and lodging — within the specifications.
But no matter the transaction, the interactions will happen through APIs, and users don’t get a chance to inspect the agent’s code. So how do they know the agent isn’t representing, or has been compromised, by a bad actor?
KYA will involve an authentication process designed to prevent fraud. Let’s say the agent goes to the grocery store site to buy items on behalf of the user. The agent would have to prove it’s legitimate by showing what is essentially a passport, which is a unique identifier that confirms it has a certain set of permissions, its developer and user have been verified, and its coding hasn’t changed since the last time an organization saw it.
KYA will help determine if there’s a bad actor behind the agent, if that person or business was allowed to create it and if it has permission to make a particular purchase.
The advent of agent directories
Agent creation will go both ways. Much like people or businesses will make agents to work for them, merchants can also create proxy commerce agents for human customers.
It works for the merchants because it’s good for business and builds loyalty. But it further complicates verification because it involves the merchant and the user.
People would need to know the agent they engage with is truly representing the store. The user would also need to create an account and go through verification. Those credentials would travel with the agent and apply at the time of purchase.
So how will users know they can trust an agent? There will be repositories of well-known businesses and their agents, whether for airlines or major stores, so people know they’re using verified technology. The repositories would have to overcome technical challenges, such as making sure the agents in the directory remain verified and no one uses them for fraud or business impersonation.
The first step toward trusted repositories would be a standards authority that establishes the guidelines for agent verification and security. Organizations across the globe could participate, adding verified agents to the repository and forming an overarching directory umbrella that points users to the best source of information.
The age of agentic
At this point, the advancement of agents is limited to only those businesses that have the technology to engage with them. Most businesses don’t have that right now.
The tech has to exist on both sides of the commerce equation. Businesses will need servers or, at the very least, fulfillment agents that know how to take agentic direction.
But it won’t be long before it’s a balanced equation. Businesses will be motivated to adopt the technology because they’ll see the value in an increased customer base.
When that happens, fraud will follow. People and businesses have identities that are constantly stolen and misused, and the same will happen with agents.
If we’re not careful about how we safeguard and design agentic identities, we’ll be in the same boat as we are with fraudulent people and businesses. But it will be worse because all we’re dealing with are bits of code.
The armies of agents are coming. It’s our responsibility to greet their arrival with secure, intelligent technology.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?
