Two subscriptions sat on two different accounts and billed at different times of the month. That timing kept them outside the view of the financial apps already in use. After the accounts were connected to Candor, Grok Bot found the overlap immediately, opened a support conversation outside Candor, and pursued the refund. The verified result was +$1,149.69.
Updated September 5, 2026
How it works with your agent
They missed two subscriptions because the charges appeared on different accounts at different points in the month, so neither a single-account view nor a narrow review window made the full pattern obvious.
This is a common shape of financial waste: no individual transaction looks broken. The evidence only becomes useful when one operator can compare all relevant accounts, retain what it saw, and investigate across billing cycles.
Once connected to Candor, Grok Bot could inspect normalized transaction records from both accounts in one financial workspace and recognized the pattern immediately.
Candor supplied current records and deterministic financial calculations. The agent supplied the judgment: it compared the charges, understood the timing, and decided the overlap deserved investigation. That separation matters because Candor does not manufacture recommendations or ask a model to guess financial facts.
Grok Bot did, using customer-support access the user had granted outside Candor. Candor remained read-only throughout and never contacted the merchant or moved money.
The support request took persistence, but it ultimately succeeded. The realized outcome was a +$1,149.69 refund, not an estimate of possible savings and not an annualized projection.
It proves that a general-purpose agent becomes materially more useful when it can examine complete, current financial records across accounts while the data layer remains read-only.
It does not prove that every subscription review will produce a four-figure refund. It does show why cross-account context, exact records, and a clear action boundary can outperform a dashboard that waits for a person to notice each charge.
This case is published with identifying details removed. The amount reflects the completed refund received by the customer.
Keep reading
One message to the agent you already use. You see the price before anything private, approve the access and accounts yourself, and your agent is working within minutes.