1. A person decides
Approving a loan, moving money or filing a return is always done by a named, accountable person.
BankIA / Responsible AI
Responsible AI in financeAI helps. The money and the decision belong to people.
In finance, a fast but wrong answer is costly. These are the principles we use to design every artificial intelligence feature on the BankIA platforms, and the real status of each one.
- AI proposes
- Code calculates
- Evidence for every figure
- Everything is logged
Six non-negotiable rules
2. Code calculates
Installments, interest, ratios and taxes are calculated with exact arithmetic and verifiable rules. AI does not calculate figures.
3. Every figure has evidence
What the AI reads from a document comes with its source page and confidence level.
4. No black boxes
Suggestions explain why they appear and can be dismissed.
5. Minimum access
AI features only see what the user can see and, at first, only in read-only mode.
6. Everything is logged
What the AI proposed, what the person decided and when, so it can be audited.
What exists today and what is coming
We distinguish between what works today and what is in the plan.
TodayWhat it does today
- Explainable suggestions. Thinking of you proposes actions on budget, savings and due dates, with the reason for each.
- Custom category rules. The client defines how to classify their expenses.
- Rule-based reminders. Overdue-installment and due-date notices that are never repeated.
- Help center assistant. Answers usage questions based on the help texts.
Coming soonWhat’s coming
- Document reading for credit analyses. In DatIA Recaudo, with evidence for each data point.
- Classification of accounting documents. In ContIA Intelligence, to prepare drafts.
- Risk and unusual-transaction alerts. Signals for human review.
- Conversational assistants and summaries. With read-only permissions and human confirmation.
Today we do not use language models to decide or to calculate. What appears under “Coming soon” does not exist yet.
What we will do with data when using AI
Before launching generative features, we will define and publish these terms.
No training on your data
We commit not to use client data to train third-party models without express authorization. This will be formalized in a contract.
Separate data
Each institution in its own space; the AI only accesses the space of whoever is querying it.
Identified providers
We will disclose which model providers are involved and in which region the data is processed.
Design commitments; they are formalized in contracts and in the legal review before each feature launches.
What AI will not do
- Does not approve or reject loans. Prepares information; the committee decides.
- Does not move money. No AI feature will execute transfers without confirmation.
- Does not make up figures. If it can’t find the data, it says so.
- Does not replace the accountant or the analyst. Takes repetitive work off their hands; professional responsibility remains theirs.
Want to know how we would apply this at your institution?
Let’s talk through your processes and define where AI helps and where it doesn’t.