The Importance of Assessing Staffing Quality in the Age of AI

U.S. regulators are more receptive to banks using AI in their AML programs, and they are still penalizing banks for having too few qualified people. A documented staffing assessment is how compliance teams can reconcile regulatory expectations.

On October 8, 2026, the Office of the Comptroller of the Currency (OCC) fined American Express National Bank USD 350 million for anti-money laundering failures. According to the OCC, about USD 13 billion in suspected money laundering was not properly monitored or reported over nearly 11 years. Among the deficiencies, the regulator cited “inadequate staffing, insufficient expertise, weak employee training” and gaps in internal audit. The Federal Reserve issued a separate cease-and-desist order.

One message from two directions

The American Express case is not isolated. In June 2026, the OCC issued a cease-and-desist order against United Texas Bank in Dallas. The order required the board to ensure “sufficient staffing levels, and the performance of periodic re-evaluations of resource and staffing needs.” It also required a qualified BSA officer with “full autonomy, independence, and responsibility,” and resources dedicated to the bank’s alert and case inventory.

At the same time, policy is moving closer to embracing technology. On April 7, 2026, FinCEN proposed a rule to refocus AML/CFT programs on effectiveness. When weighing enforcement, FinCEN’s Director will consider whether a bank is “employing innovative tools such as artificial intelligence” that demonstrate the program works. The proposal also keeps a U.S.-based compliance officer and ongoing, role-based training as core requirements.

Read together, the signal is clear. Bank supervisors want institutions to use AI, but they will not accept it as a reason for thin, untrained or unsupervised teams. Adequate staffing remains one of the core pillars of a BSA/AML program, and examiners measure it against the institution’s size, complexity and risk profile, not against a benchmark headcount.

Why AI raises the bar 

AI is a capacity lever, not a substitute for staff. Used effectively, it frees analysts from repetitive, cumbersome work. It does not remove the need for judgment, review and accountability. Illustrative estimates for a mid-sized program show the scale of what each lever can free up:

  • Model tuning. Above- and below-the-line testing to cut false positives can free the equivalent of 8 to 12 full-time analysts.
  • Alert triage automation. Rules-based closure of low-risk, documented alert types can free 5 to 7 FTE.
  • Data and narrative assist. Auto-assembled case files and AI-drafted SAR narratives, with analyst review, can free 4 to 6 FTE.
  • Workflow analytics. Real-time dashboards on alert aging, capacity and quality help redeploy the people already on the team.

Each of these tools also creates new work. Every automation change is a model change that must be validated before going live and documented. Someone has to review AI-drafted narratives, test the tuning and explain the logic to a regulatory examiner. As routine alerts are automated, the cases left for humans are the complex ones, which need more experienced investigators, not fewer.

That is why headcount should not be cut on the promise of savings. A prudent rule is to hold staffing levels until measured savings have been sustained for at least two quarters.

How a staffing assessment works

A credible staffing assessment ties resources to measurable demand, using a documented method an examiner can trace end to end. It usually follows five steps:

  1. Measure demand. Count the work: alerts, cases, SARs, enhanced due diligence reviews, screening hits and lookbacks.
  2. Model capacity. Apply productivity standards, average handle times and availability factors. The core formula is simple: required FTE = (annual work units × average handle time) ÷ net productive hours per analyst.
  3. Identify the gap. Compare required FTE with approved and filled FTE, function by function.
  4. Close the gap. Combine hiring, upskilling, automation and pre-contracted managed-service capacity for spikes.
  5. Govern and report. Re-validate quarterly, set escalation triggers and report to the board.

Key employee retention belongs in the model too. Losing an experienced investigator can cost roughly nine months before a replacement reaches full productivity, so turnover is a staffing control, not just an HR metric.

Assessing quality, not just headcount

The American Express order cited expertise and training alongside staffing levels. When reviewing a staffing assessment, compliance leaders and boards should evaluate:

  • Regulatory coverage. Every applicable law, regulation, contractual duty and internal policy has an assigned owner.
  • Risk alignment. Staffing follows the risk assessment, so the highest-risk products, customers, jurisdictions and vendors get the most attention.
  • Workload capacity. The team can complete monitoring, testing, reporting, investigations and remediation within expected timelines.
  • Skill fit. The team has the right expertise in AML, sanctions, fraud, privacy and cybersecurity, and now in validating and supervising AI tools.
  • Key-person risk. Critical activities do not depend on one person or a small group of specialists.
  • Independence and escalation. The compliance function can challenge the business and escalate concerns without conflicts of interest.
  • Evidence quality. The model supports timely documentation, issue tracking and audit-ready reporting.

The last question is the one that matters most to a board: which staffing, training, process or technology change would reduce the most compliance risk for the cost?

The bottom line

AI will change how AML teams work, but it does not change what regulators expect: having sufficient highly qualified people, with clear authority, to run a program that works. 

Integro Advisers helps financial institutions and multinational companies design risk-based AML/CFT programs, including independent staffing assessments and the governance of AI tools in compliance.

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

Introducing Integro Talks, an engaging podcast series presented by Integro Advisers.  Join us as we delve into the dynamic realm of compliance, risk management, and due diligence, delivering expert insights and analysis to empower you in making well-informed business decisions.

Find us on:

Find us on:

Podcast