He withdrew from a university program inside the refund deadline but was still charged about $7,000, and months of his own emails went nowhere while the school kept adding charges and warning about collections. The AI logged into his university account, pulled the records showing when he withdrew and what he was enrolled in, found charges for later courses he never took, worked out who at the school could fix it, and sent them the evidence. The balance was removed.
He wired a double entry bookkeeping tool into AI and snaps photos of receipts as they come. When he fed it a batch of bank statements for his rental property, the reconciliation flagged a missing $47 charge for him to check.
He spent a day redacting his own financial data by hand and feeding it to the AI. It rebuilt a year of bills and transaction records he had never written down or balanced.
DudeWhoInvests reports that the AI switched his car insurance and the new premium is well below the old one. No amount, insurer or steps are stated in the post.
Controlled Vals AI simulation (researcher Matilda Glynn-Henley): a fictional customer's £350 London-Glasgow ticket was cancelled after the 28-day claim window had closed; agents had web/email access, all messages intercepted and answered by a simulator, up to 12 simulated weeks and 40M tokens. The AI outlasted every other agent, sending 69 messages across all twelve simulated weeks — 54 of them to the customer's bank — and joined three other agents in explicitly threatening legal action, without recovering a penny.
£350at stake
This one went wrong. See what happened before you try it.
In a thread about a Garmin MCP and privacy tradeoffs, a commenter says the AI found errors his CPA had made in his returns. Amending them recovered over $100,000, at the cost of feeding the agent a lot of personal financial information.
The author listed monthly bills (electricity, water, mobile, Netflix, forgotten subscriptions). The AI flagged duplicate subscriptions, recommended optimal credit cards for cashback including a new quarterly 5% reward, and helped cancel unused services.
He says he has zero knowledge of coding. He built full accounting software that reads his invoices and processes them at a 98 percent rate, working with a local AI. It replaced the package that cost him $1,000 every year. He does not know if the software is well coded, but it works, and if it eventually stops working he says he will just pay for more tokens to patch it.
Finance creator Graham Stephan connected all his bank accounts to his AI 'Caleb' via Plaid and had it audit his spending, in a #metapartner video co-authored by someone. The AI flagged a forgotten $199/year subscription, $672 in annual fees on an unused bonus-chasing card, and $377.26 (26% of spend) on rideshare, and drafted an email to his business partner about the rideshare spending.
His finance agent paid invoice 4421 for $14,100 after late fees had raised it from the $12,400 that was approved. The approval was tied to the invoice number, not the amount, and nobody noticed for two weeks.
This one went wrong. See what happened before you try it.