Loans that fall through halfway, adjustments that wait weeks for IT, and operations no one can fully see: three bottlenecks of digital banking.
Facundo needs a loan to fix his car and applied through the bank’s app, which offered it to him pre-approved as soon as he opened the home screen. He enters the amount, chooses the instalments, and everything goes well until a message appears: “to continue, please contact your branch.” He calls. Someone answers who can’t see the application he had already started, so he explains it again. That person tells him the system for that type of loan belongs to another team, and that he’ll be transferred. Facundo hangs up, and with him, in effect, the operation hangs up too: no one on the other end knows where it was left, and he has no intention of trying again that week.
No one decided it should be like this. The pre-approved loan, onboarding a new customer, negotiating a debt: these are processes that almost every bank designed as a series of separate screens, each owned by a different team, expecting the customer to hold the thread from one end to the other. The customer shouldn’t have to hold anything. All it takes is one break for the entire operation to be lost, along with the customer and the loan.
Journeys that start and never finish
The problem isn’t any particular channel; it’s the seam between channels. A pre-approved loan starts in the app, continues in a call centre, sometimes ends in a physical branch, and at every seam there’s a chance for the process to fall apart: a piece of data that doesn’t travel from one system to another, a person who lacks the context of what was already discussed, a wait that makes the customer give up. The same happens with onboarding a new customer and collecting an overdue debt: they are long journeys with many steps, and the traditional bank sustains them on the hope that no one gets tired along the way.
Every change, a request to IT
And when something in that journey needs to change —a question that confuses people, an unnecessary step, a new product line to add— the business can’t touch it. It has to ask IT (Information Technology), join a request queue, and wait for the corresponding work cycle. A change that is minor on paper —rewording a sentence, adding a validation— can take weeks to reach production, because it competes with everything else IT has to prioritise. Meanwhile, the problem a customer like Facundo ran into this week will be hit by whoever calls next week.
What isn’t measured can’t be managed
On top of that, almost no one at the bank can see, as it happens, where each conversation is breaking down. You find out a loan was lost once the customer has already left, not while they were leaving. There is no traffic light warning that a conversation is getting complicated, nor a simple way for someone to step in on time. The diagnosis always comes afterwards, with the customer gone.
With Singular, Facundo’s call never happens, because there’s no need to transfer him to anyone. Pre-approved loan, customer onboarding and collections journeys are handled end to end within the same conversation: the platform keeps the context from start to finish, so there is no seam between systems where the conversation can get lost. If the bank needs to adjust a question or add a new step, it does so in Studio, without writing a single line of code and without waiting for IT’s next work cycle: the business edits its own agents at the pace the business needs. And while the conversation is happening, Comandante monitors it live, with a health traffic light showing which journeys are working well and which are getting complicated, so that someone can step in manually before the customer hangs up, not after.
The difference isn’t that Singular has more features: it’s that Facundo’s loan is resolved in a single conversation, without anyone at the bank having to hold the thread for him, and without a one-sentence change depending on an IT queue. That is what separates a journey that gets completed from one that is lost halfway.
In the next article in the series: why an agent that makes up a figure is, in regulated banking, much more than a system error —and how Singular prevents it by design.

