Most financial interfaces are good at displaying transactions. They are much less effective at showing continuing permissions: subscriptions, automatic renewals, recurring donations, delegated purchasing powers and other decisions that persist long after the moment in which they were made.
Gentle Bank is an early design-research concept about that gap. It does not begin with a new financial product. It begins with a question: how could financial systems preserve a person’s present agency when an old decision continues to act on their behalf?
A subscription is consent extended through time
When someone starts a subscription, the visible action may take only a few seconds. Its consequences can continue for months or years. The payment remains technically authorised, but the person’s attention, circumstances and preferences may have changed.
This does not mean that every recurring payment should require constant confirmation. Repeated prompts can become coercive in their own way. The design problem is subtler: a system should help a person notice when the conditions under which consent was given have materially changed.
Reminders without financial paternalism
A gentle financial agent would not shame a person for spending money or quietly optimise their life according to an external idea of responsibility. It would provide context and return the decision.
Your subscription renews in three days
49 zł on 22 September. The service has not been used in the last 30 days.
The point is not to make the agent the adult in the room. The point is to restore visibility at the moment when an earlier decision becomes active again.
When the conditions change, consent may need to return
A price rise is an especially clear case. A person may have agreed to 39 zł per month, not to an indefinitely adjustable relationship. Gentle Bank could allow each person to set their own threshold—for example: ask me again if the price increases by more than 10%.
The same logic can apply to a change in billing frequency, a new automatic-renewal period, a different recipient or a newly introduced cancellation condition. Re-consent would not be a universal interruption. It would be triggered by changes the person has previously defined as meaningful.
From card access to scoped authority
AI agents create a related problem. Giving an agent broad access to a payment method collapses several distinct permissions into one: what it may buy, how much it may spend, for how long, from whom and whether it may create future obligations.
A gentler architecture would issue a scoped, revocable and purpose-bound mandate. A mandate could contain:
- PurposeWhat the money may be used for
- AmountA per-purchase or total limit
- DurationWhen the authority expires
- ExclusionsWhat the agent must never do
- Change conditionsWhen fresh consent is required
- RevocationHow the person can stop it immediately
The agent is not given financial power in the abstract. It is given a bounded human intention that it may temporarily enact.
Agents should disclose whose interests they serve
Delegated payments become more complex when one agent represents a seller and another represents a buyer. Capability disclosure is not enough. A commercial agent should also make its position legible: who it represents, whether it has a sales incentive, whether it may request a purchase and whether it can execute one.
A personal agent could then interpret an offer in the context of the user’s own rules: compare alternatives, detect that a purchase creates a recurring mandate, ask about a price-change threshold and refuse to treat a seller’s recommendation as neutral advice.
Agents should disclose not only what they can do, but whose interests they serve.
What this concept does not yet claim
Gentle Bank is a conceptual design and research direction, not a bank, payment service or consumer financial product. It does not yet establish:
- technical security or fraud resistance;
- compliance with banking, payment-services or consumer-protection law;
- the reliability of subscription detection or usage estimates;
- how responsibility should be allocated when an agent acts incorrectly;
- whether these interventions improve financial agency in practice.
Those questions require collaboration with payments specialists, security researchers, legal experts, financial institutions and—most importantly—people who would live with the system.
A research agenda, not an answer
The next step is not to build a bank. It is to test the underlying interaction: when does a reminder restore agency, and when does it become pressure? Which changes are significant enough to justify re-consent? Can a mandate remain understandable without becoming a miniature legal document? How should a system communicate uncertainty about whether a subscription was actually used?
Gentle Bank offers a direction for investigating these questions: financial technology that treats consent as something living—visible, limited, revisable and still belonging to the person who gave it.