Skip to main content
Version: 8.3

Onboarding agent

The Onboarding agent is part of Creatio's autonomous banking agent portfolio for Bank.AI. Unlike AI assistants that suggest actions for a human to carry out, autonomous agents execute end-to-end workflows on their own. They coordinate decisions, actions, and handoffs across systems while keeping humans in the loop where oversight is required.

Applying for a banking product has traditionally meant visiting a branch, submitting paperwork, and waiting for staff to review each document. The Onboarding agent replaces this with a self-service digital experience. The agent can:

  • Guide customers through a structured online application at their own pace.
  • Check each uploaded document right away and return immediate feedback.
  • Generate an AI summary of the case before the banker reviews it.
  • Route cases that need human judgment to a banker for a final decision.

For the bank, this means fewer manual touchpoints, faster processing, and consistent quality across every case.

note

The Onboarding agent is installed separately. To add it to your Bank.AI configuration, contact your Creatio success manager.

Self-service application portal

Customers submit their application through a guided portal in Bank.AI. The portal walks them through each stage in order:

  • Business details: Legal and operational information about the business.
  • Applicant information: Personal details about the primary applicant.
  • Banking product: The product the customer is applying for.
  • Supporting documents: Files required to verify the application.
  • Beneficial ownership data: Ownership and control structure of the business.

Each stage opens only after the previous one is complete. This keeps data structured and consistent as it flows into Creatio. Bank staff do not need to clean it up before processing. Customers can save their progress and return at any time.

AI-powered document verification

Document review is one of the slowest steps in banking product onboarding. The Onboarding agent checks each document right after upload and assigns it a verification status. Customers get feedback at once instead of waiting for a banker.

The agent evaluates each document against two criteria:

  • Document guidelines. Each document type has a defined format, a set of required fields, and structural requirements. The agent checks whether the document meets them. It also extracts key data fields and uses them to populate the record and related forms.
  • Applicant data. The agent cross-references the extracted data against the applicant record. It checks whether the name on an ID card matches the applicant, whether an EIN letter has consistent dates, or whether an Articles of Organization covers all required fields.

When the agent finds a discrepancy, it flags the issue with a plain-language note for the banker. Discrepancies include mismatched dates, placeholder values, or signs of a fake credential.

Each document gets one of the following statuses:

  • Active: The document passed all checks.
  • Manual verification: The document has issues the agent cannot resolve on its own. A banker reviews it.
  • Declined: The document failed verification. The agent records the reason.

Customers can re-upload a corrected document at any time to continue.

Application summary and red flags

Before a banker reviews a case, the Onboarding agent generates a structured AI summary. It consolidates data from all stages: business details, applicant information, beneficial ownership structure, and the outcome of each document review. For each document, the summary includes the agent's findings and any red flags.

The agent ranks red flags by severity. Critical issues appear at the top so the banker can act without reading every detail. Examples include:

  • A novelty ID — a document that looks like a government-issued ID but is not genuine, such as a prop or sample credential.
  • An identity mismatch between a document and the applicant record.
  • A document submitted for the wrong participant.

Lower-priority points appear below. These support due diligence but do not block the case on their own. Examples include a newly formed business in a higher-risk industry or a placeholder contact email.

The summary helps bankers decide faster. Instead of opening each document by hand, the banker receives a clear assessment with the agent's reasoning already laid out.

Validation and decisioning

After the customer submits, the agent runs a validation check against the bank's acceptance criteria. The decisioning mode depends on the product setup:

  • Automatic. The agent approves or declines without banker input when the outcome is clear.
  • Manual. The agent routes the case to a banker for a final decision.

The agent also routes a case to a banker when it detects ambiguity or when a document is still under manual review.

The validation has two outcomes:

  • Approved. All required documents are verified and the data is consistent. The case moves to the agreement and signature stage. The portal delivers the closing documents to the customer.
  • Not approved. A required document is declined and the setup does not allow automatic re-routing. The case does not move forward. The banker reviews the agent's rationale and decides whether to decline or ask the customer for corrected documents. If the customer resubmits, they upload a replacement through the portal and the agent checks it at once.

This keeps bankers focused on cases that need their judgment. Routine cases move through the pipeline without manual steps at every stage.


See also

Account servicing agent

Bank.AI overview