Process a document request

What it's for

Take a document request from when it enters Plataforma de Recepción Inteligente to when it is closed as an archived task or, if the case calls for it, escalated to a project in Gestión de ayudas.

Before you start

• Role: operator with access to Plataforma de Recepción Inteligente.
• Your organization has a collection cycle open or is about to create one.
• There is at least one active AI agent for document classification.
• Estimated time: the cycle described covers anywhere from minutes to weeks depending on the real case.

Who is involved

• Cycle administrator: defines what documents your organization expects, from whom, and by when.
• Recipients: consortium partners, counterparts, internal teams that upload documents from their link.
• Operator technician: reviews the daily inbox of documents classified by the AI and validates or corrects.

Step 1: The request arrives

A document request enters Plataforma de Recepción Inteligente through one of these channels:

• Collection cycle: the administrator has previously defined a cycle (for example, “initial counterpart documentation 2026”) with expected types and recipients. The system automatically creates a document request for each recipient, with their unique upload link.
• Portal upload: the recipient receives the link by email, opens it, and uploads the documents without needing a Futuh account.
• Manual incoming file: in some cases, the operator themselves uploads a document that arrived through another channel (a manual email, an in-person delivery) to the corresponding cycle.

What it does NOT process today

The current platform does not process incoming email inboxes as an automatic trigger. That is in the backlog. [VERIFY: ticket pendiente Algalia §3.1.]

Step 2: AI classification

As soon as a document enters, an AI agent from the module analyzes it:

• It reads the file name, its MIME type, and, if it is a PDF, the extracted content.
• It compares against the expected types in the cycle (a national ID, an activity certificate, a financial report, a responsible declaration).
• It emits a classification with a confidence score between 0 and 1.

If the confidence score exceeds the configured confidence threshold (default 0.7), the document is automatically validated. If not, it is placed on hold for human review.

For the detail of the AI mechanism, see Plataforma de Recepción Inteligente.

Step 3: Human validation

1. Open Plataforma de Recepción Inteligente → Received documents inbox.
2. Filter by status Pending review to see what the AI did not classify with high confidence.
3. For each document, review the proposed classification and the confidence score.
4. Confirm or correct the classification, or reject the document if it does not apply to the cycle.

The operator can confirm large batches at a glance when the AI gets it right with high confidence. The review only stops on the doubtful cases.

Step 4: Closure as a task

Once a document is validated, it is associated with a task in the cycle. The task is closed when all expected documents from a recipient have arrived and been validated.

This means each recipient has their own task in the cycle, with a visible progress percentage. The cycle coordinator sees the progress of the entire consortium or file in a single view.

In the majority of canonical cases (documentation collection from partners, internal grant justification, counterpart documentation), the flow ends here: cycle complete, all documents archived as attachments on the Futuh records, escalated reminders sent to late recipients, file closed.

Step 5 (optional): Escalation to a project

There are cases where a document request does not stay as a simple task but instead triggers a full project. Canonical examples:

• Documentation arrives for a private call that was not in Radar, but which deserves to be submitted.
• A partner entity asks your organization to lead a new consortium.
• A public authority asks your organization to provide an expert report or formal opinion.

In these cases, the operator technician manually escalates the task to a project in Gestión de ayudas. The system preserves the bidirectional link: the project knows it came from Plataforma de Recepción Inteligente, and the validated documents remain accessible from the project.

From that point, the flow continues inside Gestión de ayudas with the formulation and submission cycle.

Escalation is a human decision

The AI classifies and proposes. Escalation to a project is not triggered automatically: it is always decided by a team member. The reason is that a project commits resources and dedication for months, and that decision is not delegated to a confidence threshold.

Step 6: Automatic reminders to recipients

While the cycle is open, a daily cron checks recipients who are late and sends escalated reminders according to the configured policy:

• Tier 1: friendly tone, polite reminder.
• Tier 2: assertive, mentions the deadline.
• Tier 3: urgent, notice that the deadline has passed.

Each reminder is recorded in the chatter of the corresponding request, marked automatically by the system to distinguish it from reminders the operator sends manually.

How this flow fits with the rest

Module	What it contributes to the flow	Detail
[[Plataforma_de_Recepcion_Inteligente]]	Receives documents, classifies with AI, manages the collection cycle	Plataforma de Recepción Inteligente
[[Silvio]]	Queries pending requests and alerts about delays	Silvio
[[Gestion_de_ayudas]]	Receives the project when a request escalates	Gestión de ayudas
[[Federacion_Futuh]]	Provides the canonical document type catalog for the sector	Federación Futuh

Learn more

• Plataforma de Recepción Inteligente for the module detail.
• Capture and formulate a call for proposals if the request escalates to a project and you need the formulation flow.
• Glossary for the canonical terms: document request, task, cycle, AI agent.