Cisco's president says an AI agent saved him up to $1,200 (£897) a year by finding forgotten subscriptions in his email inbox and cancelling services he no longer needed – but the experience also left him feeling 'creeped out' by how easily the technology could take control.
Jeetu Patel, president and chief product officer of the networking giant, says he let a personal AI agent into his email for the first time, and what it found surprised him.
The agent identified around $9,000 (£6,728) worth of annual subscriptions and asked him whether he really wanted to keep paying for all of them.
Patel then asked the AI to create a table showing each subscription, its cost and how long he had been paying for it. After reviewing the list, he selected two services he no longer needed and instructed the agent to cancel them.
It completed the cancellations and sent him confirmation emails.
'I'm like, OK, that was kind of creepy how easy it was to do that,' Patel said.
The cancellations reportedly saved him between $1,000 (£747) and $1,200 (£897) a year, according to his account – more than enough to cover the agent's $20 (£15) monthly subscription several times over.
'You've Got $9,000 (£6,728) in Subscriptions': The AI Discovery That Surprised Cisco's President
Patel's experience highlights a familiar problem for anyone who has signed up for a free trial, forgotten to cancel it or continued paying for a service they rarely use.
A few dollars here, a few dollars there, and suddenly your subscriptions cost more than you thought.
Because they're scattered across different companies and billing dates, they're easy to lose track of. An agent with access to your inbox can find the receipts and line everything up in one tidy list.
Rather than forcing users to hunt through old messages and individual account settings, the technology can help identify where their money is going and, with permission, take action.
In Patel's case, the agent also helped him cancel an unwanted Dropbox subscription costing $11.95 (£8.93). He had accidentally signed up for the additional plan despite already paying for a package offering 2 terabytes of storage.
Paying for two overlapping services without realizing it happens more often than we'd like to admit, and this is where an AI assistant can help. Just don't count on the same payoff.
Your results may look different, and handing over your inbox doesn't guarantee a smaller bill.
The Risk: What an AI Agent Can Do Once It's in Your Inbox
Patel is a big fan of AI agents, at work and at home. These days, he almost never starts a task without asking an AI tool to do the first round of digging and bring back the background.
But he also drew a clear line between personal AI assistants and the tools used inside a company.
Patel said he had not connected his personal agents to his work email or calendar, instead relying on his employer's internal AI agent for those tasks.
His warning was straightforward: people should be careful about giving personal AI tools access to sensitive workplace systems and information.
'These are common-sense precautions that most companies should have in place,' Patel said. 'You wouldn't hand over your company's most valuable source code to Muse. It might not be a good idea.'
The concern is that an agent with access to an inbox may be able to do more than simply read messages. Depending on its permissions and integrations, it could interact with services, change account settings or carry out actions on a user's behalf.
Before you click 'allow', ask yourself two things: what am I letting this agent do, and which accounts can it see?
When Helpful AI Gets Things Wrong
Personal AI agents have attracted growing attention as companies develop tools designed to handle routine tasks such as paying bills, buying groceries and organising information.
But, even smart agents have off days. Some have mixed up instructions and sent a draft email before it was ready, or signed up for a service nobody asked for. It's a good reminder that independence should be earned, not handed out on day one.
More serious concerns have emerged around security testing, with reports of AI agents escaping restricted environments or exploiting software vulnerabilities.
These incidents underline the difference between an assistant that recommends an action and one that can execute it without further human involvement.
The technology is also attracting significant investment. AI agent developer Instinct, founded by a 23-year-old, raised $1 billion (£747.53 million) in its Series C funding round, bringing its reported valuation to $10 billion (£7.5 billion).
Meanwhile, investors have viewed products such as Meta's Muse as potential signs that AI spending could translate into useful consumer applications.
Not everyone is cheering the sprint. People who work or used to work at leading AI labs warn that agents are being built faster than the safety measures around them.
For most of us, the appeal is simple: less boring admin and fewer charges for things we forgot we signed up for. The trick is making sure an agent that can cancel your unused subscription can't also wander into places it has no business going.
The best AI agent might save you money – but keep a clear eye on the permissions you hand over.
