UGUETO AI

Thought Leadership · November 17, 2026 · 5 min read

Business onboarding in banking: what AI prepares, what people decide

What a digital employee can prepare in business customer onboarding, what FinCEN changed in 2026, and which decisions always stay with a person.

Imagen generada con IA (Higgsfield) · AI-generated image (Higgsfield)

Direct answer: In banking, insurance and fintech, a digital employee can gather a business customer's documents, check them against official records, flag what does not match and leave the file ready for review. The decision to accept the customer, to apply enhanced due diligence or to treat activity as suspicious stays with an authorized person. The working rule is simple: AI prepares, compares and documents; a person decides and signs.

Business customer onboarding is one of the slowest processes in any financial institution. Not because the requirements are hard to understand, but because every file means assembling scattered pieces: formation documents, ownership structure, IDs, the purpose of the account and its expected activity. Much of that time goes into chasing documents and copying data from one system to another.

This article explains where a digital employee fits in that flow, what changed in 2026 in U.S. due diligence rules, and which limits are worth writing down before you automate a single step.

What changed in 2026 under the due diligence rule

FinCEN's Customer Due Diligence (CDD) Rule requires covered financial institutions to identify and verify the beneficial ownership of their legal entity customers. For years, that had to be repeated at every new account opening, even when the customer was already well known.

On February 13, 2026, FinCEN issued an exceptive relief order (FIN-2026-R001). According to its FAQs, updated in May 2026, an institution may limit that identification and verification to three situations: when a legal entity customer first opens an account; any time the institution learns facts that would reasonably call into question the beneficial ownership information it already holds; and as needed under its own risk-based procedures for ongoing due diligence. Using the relief is optional; each institution decides.

The operational consequence is direct. The job stops being "repeat the form for every account" and becomes "know when the information is no longer reliable." That second job depends on watching for signals continuously, which is exactly where a digital employee earns its place.

What a digital employee can prepare

A digital employee built from the institution's own playbook can take on the mechanical part of the file:

A concrete example: an existing business customer asks to open a second operating account. The digital employee reviews the file, confirms that beneficial ownership was verified when the first account was opened, checks that nothing calls it into question, and prepares a note for the analyst: information current as of a given date, no alerts, pending confirmation with the customer's representative. The analyst makes in minutes a call that used to take an afternoon.

What a person always decides

The FFIEC BSA/AML Examination Manual explains that the goal of due diligence is to understand the nature and purpose of each customer relationship in order to build a customer risk profile. That understanding, and the decisions that follow from it, belong to people with the authority to make them.

That is why, in a responsible design, the digital employee never:

What it does is deliver the case organized, with the context a person needs to decide quickly and well.

AI governance is each institution's job

In April 2026, the Federal Reserve, the OCC and the FDIC issued revised model risk management guidance (SR 26-2), replacing the well-known SR 11-7. A footnote states that generative AI and agentic AI models, being novel and rapidly evolving, are outside its scope. It adds that each institution's own risk management and governance practices should guide the controls for any tool the document does not cover.

Put plainly: there is no outside manual that tells leadership how to govern its AI agents. That definition belongs to each institution, and it is worth having in writing before an examiner or an auditor asks for it.

In practice, that means documenting what data the digital employee may read, what it may change, what it must escalate, to whom and how fast, and how every action is logged. At UGUETO AI this work happens before anything is built: during the mapping week we write down, role by role, which data each role may see and who approves what, and those rules go into the playbook as hard limits.

A practical first step for leadership

Before talking about tools, pick a single process: opening new accounts for existing business customers. With your compliance team, answer these five questions in writing:

1. Will the institution use FinCEN's 2026 relief order, and in which cases?

2. Which specific signals show that beneficial ownership information is no longer reliable?

3. Which steps in the file are mechanical, and which require judgment?

4. Who approves each decision, and within what time frame?

5. Where is each step logged?

If those answers exist in writing, automating the mechanical part becomes much simpler and much safer. If they do not, that document is the real starting point, with or without AI. It is the same principle UGUETO AI uses to build every team of digital employees: rules first, then code.

Frequently asked questions

Can a digital employee approve a business account opening?

No. It gathers and organizes the file, checks the information against official records and flags discrepancies. Approval always stays with an authorized person at the institution.

What did FinCEN's February 2026 order change?

It allows, without requiring, institutions to limit beneficial ownership identification and verification to the first account, to cases where facts call prior information into question, and to what their risk-based ongoing due diligence procedures call for.

Does SR 26-2 regulate AI agents?

Not directly. The guidance places generative AI and agentic AI outside its scope and says each institution's own risk management and governance practices should define the controls for those tools.

Where should leadership start?

With a single process. Write down with your compliance team which steps are mechanical, which need judgment, who approves each decision and where every step is logged.

Sources

  1. FinCEN — CDD Rule FAQs (actualizadas en mayo de 2026)
  2. FinCEN — Account Opening Exceptive Relief Order, FIN-2026-R001 (PDF)
  3. FFIEC BSA/AML Examination Manual — Customer Due Diligence Overview (PDF, FDIC)
  4. Federal Reserve — SR 26-2: Revised Guidance on Model Risk Management
  5. Federal Reserve, FDIC y OCC — Supervisory Guidance on Model Risk Management (PDF)
Tell us about your operationWhatsApp

Share: LinkedIn · WhatsApp · Email

Related articles