Why AI Agents Could Break the Subscription Economy
Source: Mike Todasco; Substack · 2026-10-12
Columbia House was founded in 1955 by Columbia Records as a way to sell items that may have fallen off the Billboard charts, directly to consumers. (A novel concept at that time.)
The hook? Get 12 records (later CDs) for a penny.
The catch? A bunch of arcane rules and deadlines, where if you didn’t comply, you’d be paying more than full price for other recordings. This is known as Negative Option Billing.
Even after selling Brian’s information to every catalog in America, the only reason that these record clubs could survive is that my diligence and rule-following were the exception. Most people would forget a step somewhere, and suddenly they’re forced to buy the newest Bloodhound Gang CD for $30. My CD collection was subsidized by everyone else. I was, to introduce another economics term, a “free-rider.”
The concept of Negative Option Billing remains prevalent today. A website that says, “Use this for 7 days for free, and then we’ll charge you.” Subscribe and save services. And of course, the king of this racket, gym memberships.
But these business models will have to change. The advent of AI agents will make it effortless to keep on top of bills, due dates, and shady subscription charges. Everyone will have their own college student... or better yet, a world-class expert, with infinite time, ready to manage the confusion and drudgery of Negative Option Billing. The Columbia Houses of the world have no hope in this future.
I even did it myself. I have AT&T Fiber that has crept up to over $100/month. A couple of clicks with Muse, and it negotiated a retention deal that will save me over $400 over the next year with no commitments or loss of service. Said another way, Muse saved me $400, and I did nothing. But AT&T just lost $400 in pure operating profit they were probably banking on, due to general consumer laziness and inertia. Now what happens when every AT&T customer has an agent? How will this change the way companies operate?
Dark patterns are the little steps that a website uses to complicate the cancellation process (that they’d never use when signing up). Amazon agreed to pay a $2.5B settlement to the FTC last year over how they made canceling Prime deceptively difficult. Despite the litigation, these patterns still proliferate. But your agent, with infinite patience to read every line of the T&Cs and the ability to navigate to the correct path, will not have these same issues.
The Amazon.com marketplace makes more profit from advertising than it does from actually selling goods. So it is little surprise that Amazon recently blocked Meta’s Muse from accessing their site. Sure, they’re giving up a bunch of low-margin sales... but the problem for them is that AI agents don’t click on ads. And to keep its margins, Amazon needs you to click on ads, not just have agents find the cheapest stuff.
As Axios recently reported, companies in the Negative Option Billing business have already seen their share prices drop since the release of Muse. Agents make it easier to end that gym membership (Planet Fitness), cancel that recurring order of puppy chow that is now cheaper at Walmart (Chewy), or find your car insurance rates lower than yours (Allstate).
AIs only work well when given enough context. It doesn’t know how to help you if it doesn’t know what you need help with. You have to give these AI Agents your information so they can serve you best.
But that also means you might be handing over your email, which has notes from exes, password reset links, W2s, legal docs, and let’s just admit it, all of our digital lives in there. Meta says they’re doing the right things with privacy for Muse by setting up a separate virtual computer for each user, limiting their access, etc. And they may be... but this is Meta. And without Apple and Google, who already have all your info, still not being serious players with AI Agents, you have to make a decision if you want to play this game. Do you trust Mark Zuckerberg or some four-month-old Silicon Valley startup more?
Remember when I said Columbia House had no hope in this future? Well, last month they announced it would shut down the business for good. It wasn’t killed by AI; streaming did that years ago. But the timing is poetic. Days after the original negative-option empire shut its doors, Meta released an agent whose favorite party trick is canceling things.
Back in college, I became an expert in subscription service terms and timing to beat the system. Now everyone gets that expert, 24/7, and it never forgets a deadline. That’s great for us... initially. Remember, I only got 12 CDs for a penny because others forgot to cancel. When nobody forgets, nobody gets it for a penny. Expect fewer free trials, more annual prepay, and more “agents not welcome” signs.
Because when we’re all free riders, there’s no one left to ride on.
Any weird or scammy charges you are seeing on my bills?
Any free money in my email that I may not have claimed? Look for class-action settlements, etc.
List every bank fee, annual fee, and add-on I’ve been charged this year and negotiate to have them reversed.
Remind me a week before any special-offer trial ends, and negotiate with the vendor to continue the trial pricing.
Manage my points, miles, and travel credits across all programs and flag anything about to expire.
Go through my email and see if there are coupons or offers that are actually worthwhile (i.e., not typical offers).
Monitor my purchases (flights, etc.) and notify me whenever the price drops within a set window, and request the difference as a credit.
List every subscription I’m paying for and flag ones I haven’t used in 90 days.
See if my family or I have unclaimed property.
Audit my overall expenses based on what you know about me. Find opportunities to save money.
Lines of inquiry this paper opens 10
Research framings built by reading the notes related to this paper — the questions it feeds into.
How effectively can sellers exploit AI shopping agents through description changes? How can AI agents earn consumer trust with financial decisions? How can merchant-funded agents remain loyal to consumer interests in digital commerce?- Do AI agents actually eliminate profits from negative option billing?
- How are publishers adapting to agents that skip human-dependent revenue?
- How quickly can subscription businesses shift to annual prepay and licensing models?
- Can sellers distinguish agent-driven cancellations from human ones to maintain pricing?
- Which actors currently control the largest shopping agent platforms?
- Why would Amazon block Meta's Muse if agents threaten human-dependent margins?
- How much ad revenue does Amazon lose when agents skip sponsored listings?
- What share of Amazon's total profit comes from marketplace advertising?