THE TEARDOWN
Zero point one percent against five percent. The only thing holding that gap open is that nobody gets around to closing it.
The Structural Signal
Torsten Slok published a note on Sunday, September 27, 2026. The title asked whether an agentic bank run is coming.
Slok is chief economist at Apollo Global Management. His argument is short.
If every household ran an agent to optimize idle cash, banks lose their cheapest funding. Banks could lose a large share of the cheap deposits, he wrote.
The rate gap carries the whole case. A US checking account pays about 0.1% on average.
Fintech accounts pay 3.3% to 5.0%. He named Revolut, SoFi, Varo, LendingClub and Wealthfront.
The agent he points at exists already. Meta launched Muse on September 8.
It can view a user's financial accounts through a data connection. That is the missing piece, not the model.
Now the disclosure that belongs up front. Apollo runs an $849 billion credit arm lending to businesses in competition with banks.
It has said it aims to manage $1.2 trillion in private credit by 2029. A note arguing bank funding is fragile is not a neutral document.
The Mechanical Breakdown
A retail deposit franchise has six parts. Most people see one.
One: the bank takes a deposit and pays close to nothing for it.
Two: it lends or invests that money at a higher rate. The gap is net interest margin, and it is the business.
Three: the margin depends on deposit beta. That is how much of a rate rise a bank must pass through to keep the money.
Low beta means cheap funding. Every basis point not passed through drops to the bottom line.
Four: beta stays low because of friction. Find the rate, open the account, move the direct deposit, redirect the payments.
Then remember to start any of it. That last step is where most households stop.
Five: the rulebook rewards the stickiness. Liquidity rules assign stable retail deposits a low assumed runoff rate.
Six: the bank prices its loan book against that assumption. Mortgages and small business credit are funded by money nobody moved.
Now place the agent in that chain. It touches step four and nothing else.
It does not change rates. It does not touch the loan book or the rulebook.
It removes the forgetting. That is the entire mechanism.
Here is where a model replaces human judgment. The decision not to move money was never a decision.
Inertia is the absence of one. An agent makes absence impossible.
Wall Street Is Betting on Dollar Debasement — Here's Why
A major Wall Street trade is back.
It's built around fears that your dollars could buy LESS.
It's called the "debasement trade."
And with America's national debt now past $40 trillion, investors have been moving into scarce assets like gold amid concerns over debt, inflation and the dollar.
Why should retirees care?
Because when inflation erodes purchasing power, the dollars you've saved don't stretch as far.
Your IRA.
Your 401(k).
Your cash savings.
If your retirement depends heavily on dollars, what those dollars can actually buy matters.
Gold surged in August as the debasement trade came roaring back.
We've put together a free guide revealing 3 steps you can consider taking as concerns over America's debt and the dollar intensify.
Wall Street is already paying attention. Are you?
Legacy vs. Autonomous
Legacy rate shopping is occasional and uncoordinated. A household compares rates when something prompts it, then acts weeks later or never.
That lag keeps beta low. Banks have a century of data on how slowly people move.
The agentic version is continuous and synchronized. The same models read the same rate feeds for millions of households at once.
Direction is not the novelty. Simultaneity is.
Deposit migration has always happened. It happened gradually because people are not coordinated, and shared software supplies exactly that coordination.
Now where the machine case breaks, in four places.
Insurance limits are the first. Coverage caps mean an agent chasing yield must split balances or accept uninsured exposure.
Liquidity need is the second. The agent does not know the household owes a contractor on Tuesday.
The third is the largest and it is almost never stated. Many high-yield fintech accounts are sweep programs.
The balance gets routed onward to partner banks. Money leaves one bank and arrives at another, repriced rather than removed.
So the system-level drain is smaller than the headline. The margin compression is real either way.
The fourth is evidence. Slok gave no estimate of size or timing, and agent-driven transfers do not yet show in deposit data.
The DOJ Says Your Cash Isn't Legally Yours
In federal court, the Department of Justice argued that cash isn't property — meaning Washington could freeze or seize your savings without warning. When governments go broke, they come for the people. Smart Americans are moving wealth out of the system now. Our free kit reveals 3 strategies to protect your retirement before the trap snaps shut.
Capital Flow Implications
Four pools move.
The first is net interest margin at deposit-funded banks. Non-interest-bearing balances are the cheapest money in finance.
The second is the branch network. Its remaining economic function is friction, and friction is what the agent deletes.
The third is the rate-shopping industry. Comparison sites and deposit brokers sell a service an agent performs free and continuously.
The fourth is private credit, and it is the interesting one. If bank funding costs rise, bank lending contracts.
Somebody fills that gap. Apollo has said it intends to be that somebody at $1.2 trillion.
Follow it to the end. The firm warning about deposit fragility is positioned to fund what banks cannot.
That does not make the mechanism wrong. It does mean the note is a trade thesis as well as an analysis.
Note the regulatory counterweight from four days earlier. The Fed proposed an anti-evasion rule on September 24 aimed at affiliate yield on stablecoin balances.
Same spread, different instrument. Authorities are already contesting who may pay a household for idle cash.
Verdict
Inertia was the asset. It sits on no balance sheet and it has no defender.
What compresses is net interest margin on retail deposits. The premium for being where money rests by default goes with it.
What survives is the account itself. Swept balances still end up inside insured banks, just not the same ones.
Watch deposit beta in fourth-quarter results. That is where this stops being a theory and becomes a number.
Watch the permission model harder. Nothing moves until a household lets an agent initiate a transfer, and almost none have.
Territory: +machines on the inertia, +incumbents on the account.
Retail deposit margin compresses, while the balance often lands back in an insured bank at a higher price.
