# Credit is not slow to decide: it is slow to assemble > An analyst does not lose time deciding. They lose it waiting for the missing document. What changes when the file arrives complete and validated. Credit ## Credit is not slow to decide: it is slow to assemble An analyst does not lose time deciding. They lose it waiting for the missing document. What changes when the file arrives complete and validated. **Luis Aguilar** September 15, 2026 3 min read IN SHORT - The bottleneck in credit analysis is rarely the decision. It is gathering and validating the paperwork before anyone can decide. - Every round trip over an illegible or missing document adds days, and the applicant leaves with whoever answers first. - An AI agent gathers the paperwork through conversation, validates it against the institution's records and assembles the file before it reaches the analyst. - Whatever does not match is flagged before the review, not after, so the analyst decides instead of chasing documents. Ask at your institution how long a loan application takes and you will get a number. Ask where that time goes and the conversation gets uncomfortable, because it is almost never spent deciding. It is spent waiting. A blurry proof of income. An identity that does not match the records. A document the applicant swears they sent and nobody can find. The analyst, the most expensive person in the process, ends up playing courier. ### Where the real cost sits Two things get confused inside the same indicator, and it is worth separating them. **Decision time** is what an analyst takes to assess a case that already has everything in front of them. It is usually measured in minutes and is rarely the problem. **File assembly time** is what passes between someone applying and the case being ready to assess. That is where the round trips live, and where the applicant is lost, because meanwhile they are already talking to another institution. That second stretch almost never has an owner. It does not appear in committee, nobody reports it, and so nobody attacks it. Where the days go in a credit file - 01 Application Trabajo The customer states what they need. Minutes. - 02 Waiting for documents Espera The applicant looks for papers, photographs them, sends them when they can. - 03 Legibility check Trabajo Someone opens each file and decides whether it is readable. - 04 Round trip over what is missing Espera The document is requested again. The clock starts over. - 05 Validation against records Trabajo What the document says is compared with what the systems say. - 06 Queue before review Espera The file waits its turn with the analyst. - 07 Assessment Trabajo The decision itself. Usually the shortest step. The steps marked as work are measured; the ones marked as waiting are not. That is why the reported time never matches what the applicant experiences. 3 de 7 etapas son espera, no trabajo. ### What an agent does before the case reaches the analyst An AI agent talks to the applicant on the channel they already use, usually WhatsApp. It asks for documents one at a time, checks each is legible and compares what the document says with what the institution’s records say. When something does not match, it asks again right there, inside the same conversation. When everything is in, it assembles the file and hands it to the analyst with the mismatches already flagged. Flow canvas Start Branch Contact · 5068… Contact · 5066… Else AI Agent CrediAmigo Digital assistant · live The path is defined on a canvas, without writing code: which document is requested, against which record it is validated and at what point the case is flagged and passed to a person. The rules belong to the institution. [See the flow builder →](https://www.shimliapp.com/en/funciones) The distinction matters for a committee: the agent approves nothing. It gathers, validates and orders. The decision stays where it was, with everything on the table. Three questions for your next committee - Of the days an application takes, how many go into assembling the file and how many into assessing it? If nobody has that split, that is the finding. - What share of applications drop off before reaching the analyst, and why? The cause is usually a document, not a policy. - Which checks does a person do by hand today that could be run against a record? That list is the scope of the first automation. [See the full credit case →](https://www.shimliapp.com/en/casos/credito) ### Questions on this topic Does the agent decide whether a loan is approved? No. It gathers the paperwork, validates it against the institution's records and assembles the file. The approval decision stays with the institution, within its policies and with whatever human review it defines. What happens if a document arrives illegible or does not match the systems? The case is flagged before it reaches the analyst, with the reason stated. The agent asks the applicant for that specific document again in the same conversation, without anyone having to open a ticket. Useful to someone on your team? Share it with whoever decides on this process. [WhatsApp](https://wa.me/?text=Credit%20is%20not%20slow%20to%20decide%3A%20it%20is%20slow%20to%20assemble%20https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Fincomplete-credit-file) [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Fincomplete-credit-file) [Email](mailto:?subject=Credit%20is%20not%20slow%20to%20decide%3A%20it%20is%20slow%20to%20assemble&body=Credit%20is%20not%20slow%20to%20decide%3A%20it%20is%20slow%20to%20assemble%20https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Fincomplete-credit-file) Copy link ### Keep reading [See all →](https://www.shimliapp.com/en/blog) - Platform #### FINNOSUMMIT 2026: one in three sessions was about AI Instead of summarising impressions, I counted the sessions: 22 of the 67 had AI in the title. That says a lot about where banking in the region stands. September 25, 2026 - Support #### Your next banking channel is already installed: WhatsApp AI agents are doing to apps what apps did to websites. What that means for an institution running real volume, and what separates a chatbot from an agent. September 25, 2026 - Platform #### The next AI advantage is: trust What the Volcano Innovation Summit 2026 left behind: when capability gets cheap, the advantage becomes trust. And trust has two sides. September 16, 2026 ### Put an agent to work on this process. Create the account and start today. Or we take 30 minutes and define which metric you'll move. [Create account](https://shimli.app/user/auth/create-account) Book a demo [See the full credit case →](https://www.shimliapp.com/en/casos/credito)