BankingNewsAI Daily Brief ·
US Senate negotiators consider requiring AI companies to assess risks
Banking AI
Financial institutions & fintech technology
Figure partners with Sierra on AI loan agents
For the chief lending officer at a regional bank
Require prelaunch proof that automated voice and SMS follow-up treats comparable borrowers consistently and sends edge cases to people.
Figure partnered with Sierra to deploy an AI agent for stalled U.S. home equity applications. The agent works in Figure’s loan origination system, contacts applicants by voice and SMS over multiple days, and guides credit checks, ID verification and bank-account linking before handing cases to loan officers. In an early pilot, applicants who used the agent moved through friction stages at rates 30% to 52% higher, while engaged borrowers funded 67% more loan volume. Figure says combining the agent with loan officers lifted funded-loan conversion 143% versus loan officers alone and will roll out the integration to network partners in coming months.
→ Action
Fair lending: Test AI follow-up scripts and escalation rules for disparate treatment and outcomes before using them on stalled HELOC applicants.
Read article → from Stocktitan
Navy Federal uses AI to fight scams
For the fraud chief at a regional bank
Your unauthorized-fraud stack will not catch customers who authorize scam payments. Decide whether you have payment-context data, scam-account intelligence and authority to stop a transfer when the customer insists.
Navy Federal Credit Union uses Cube AI to identify accounts that scammers direct victims to pay. Cube’s AI bots engage scammers and send the recipient account or handle to financial institutions after scammers share payment instructions. Navy Federal says fraud attempts against members fell about 25% from 2024 to 2025, while scams became a larger focus. The $204 billion-asset credit union has prevented about $125 million in wire-scam losses over the past 19 months. It now refuses transfers when enough fraud or scam red flags exist, after ending its practice of requiring warned members to sign affidavits before proceeding.
→ Action
Fraud operations: Test scam-payment controls against customer-authorized wires, including recipient-account intelligence, escalation rules and transfer-stop authority.
Read article → from Banking Dive
General AI
Large language models & AI infrastructure
US Senate negotiators consider AI risk requirement
For the chief risk officer at a regional bank
Do not wait for a federal rule to force discipline on AI firms. Require written risk assurances, incident notice and remedies now, because provider accountability will become part of your own model-risk case.
US Senate negotiators are considering a requirement for AI firms to mitigate known major risks.
→ Action
Third-party risk: Add AI-vendor clauses for documented risk controls, incident notification and remedies to current contract templates.
Read article → from Reuters
Salesforce introduces Trusted Enterprise AI Harness
For the CIO at a Salesforce-using bank
Do not treat CRM agents as a Salesforce-only project. The test is whether its Control Plane can inventory, permission and govern Salesforce and third-party AI together; a separate control stack will create gaps.
Salesforce introduced its Enterprise AI Harness, a composable architecture for AI context, agency, action, governance, security and models. It includes an AI Control Plane that gives businesses one place to see, manage and control agents and AI across the enterprise. Customers can use all six capabilities or select only those they need, with Salesforce technology, existing technology, or third-party models, agents and systems. Salesforce says existing customers can use their current technologies as a foundation and upgrade as new capabilities become available.
→ Action
AI governance: Map CRM-connected agents and third-party models to required inventory, access permissions and control ownership.
Read article → from Salesforce
Get this in your inbox every morning
Free · No spam · Unsubscribe anytime