CRAC

Most AI transformations are designed looking at the customer. CRAC looks inward: it answers administrative, legal, and benefits questions from the bank's own employees, on WhatsApp, any day, any time.

Before, resolving something as simple as how many vacation days were left could mean a call, an email, and a form. What changed isn't the answer: it's that the answer now arrives when the person needs it, not when the office opens.

The human doesn't disappear, they move

The assistant generates the answer and people shift to supervising and improving it. When a query goes beyond what it can resolve, it hands off to someone. That combination of automatic generation, integration with the systems, and human judgment where it's needed is, to me, the right way to apply this technology to an internal process.

Why this case matters more than it looks

An internal tool adopted by the vast majority of people without anyone forcing it proves something no presentation can prove: that the technology solved a problem people actually had. That result is what led the bank to institutionalize the approach instead of treating it as an isolated experiment.

That's how adoption gets built. Not with a central mandate, but with a case the person next to you saw work.