BankingNewsAI Daily Brief  · 

Apple starts rolling out Siri AI across its platforms

🏦 3 Banking AI🤖 3 General AI

Banking AI

Financial institutions & fintech technology

3 stories

CMG launches Omni Income with Freddie Mac calculator API

For a mortgage operations executive at a regional bank

CMG’s live use of Freddie Mac’s calculator API puts more pressure on income review before underwriting. Your bank must decide whether fragmented calculations and repeat document work are costing eligible loans both cycle time and representation and warranty relief.

CMG Financial launched Omni Income, an in-house AI-enabled income analysis platform that uses the Freddie Mac Income Calculator API in its live workflow. CMG says it is the first lender to put the API into production. The platform extracts income data from borrower documents and presents the data, calculation and source documents together for review and verification. For eligible loans submitted through Freddie Mac’s Loan Product Advisor, income calculated through the API may qualify for representation and warranty relief. In beta tests, complex self-employed income cases took single-digit minutes, versus about an hour manually or roughly 40 minutes with a conventional extraction tool.

→ Action

Mortgage operations: Map income-calculation handoffs and identify eligible Loan Product Advisor loans where duplicated calculations delay representation and warranty relief.

Read article →  from Yahoo Finance

PayNearMe acquires AI company Marr Labs

For the payments platform head at a bank

PayNearMe is putting agentic AI inside payment servicing, not just support. Your bank must decide whether its payment rules and compliance controls can constrain an agent that acts in payment flows.

PayNearMe acquired Marr Labs’ technology assets and hired key employees to add agentic AI for voice, messaging and workflow automation in regulated settings. Marr Labs built AI technology that supported millions of customer interactions each month at financial services companies. PayNearMe processes more than $50 billion in payments each year, and more than 20,000 businesses use its PayXM platform. PayNearMe plans to integrate agents directly into PayXM rather than offer them as a separate tool. Marr Labs says payment agents need access to data, business rules and workflows while operating within compliance requirements.

→ Action

Payments product: Check AI agent access to payment data, business rules and workflows for compliance controls.

Read article →  from Finextra

TS Imagine ships AI platform for capital markets

For the capital-markets COO at a US bank

TSIQ pushes AI from research into trade, risk and portfolio workflows. Your control framework must test each recommendation, source trail and action permission before an agent can affect a regulated decision.

TS Imagine launched TSIQ, an AI platform for capital-markets workflows. The company says TSIQ can explain, recommend and act with human oversight and control at each stage. Its proprietary TSIQ ontology links data, products, financial concepts, relationships and business logic to produce auditable answers. Clients can apply TS Imagine’s data foundation to their own data in a secure environment, with outputs tied to transparent logic and verifiable sources. TSIQ will be available across TS Imagine’s platform for trading, risk and portfolio management, wealth management and capital servicing.

→ Action

Model risk: Test TSIQ outputs against approved data sources, trading policies and risk limits before allowing workflow actions.

Read article →  from Finextra

General AI

Large language models & AI infrastructure

3 stories

Google introduces Gemini 3.8 Live and 3.8 Live Extended Thinking

For the contact-center head at a regional bank

Voice automation can now keep talking while it calls tools and reads visual context, which makes weak authentication and handoff controls harder to hide. Your bank must set clear limits on what an agent can do before a customer or employee approves it.

Google launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking for voice agents and conversational use in the Gemini app, Google Workspace and Search. Gemini 3.8 Live handles visual inputs in near real time and can switch between 97 languages during a conversation. It can execute tools and API calls in the background while it continues talking with the user. Extended Thinking reasons and speaks at the same time, using verbal acknowledgments and progress narration for multi-step background tasks. Google says Extended Thinking scored 35.1% on Sierra’s τ-Voice-banking benchmark and 68.6% on τ-Voice.

→ Action

Contact-center controls: Test customer authentication, human handoffs and approval gates before a voice agent can call tools or complete account actions.

Read article →  from Google

Salesforce and NVIDIA announce Koa reasoning model

For the Salesforce CRM head at a regional bank

Koa gives your bank a Salesforce-hosted option for multistep CRM agents, but Salesforce’s benchmark is not your control test. Decide whether its case-routing, service and sales actions meet your approval, error and audit requirements before widening agent access.

Salesforce and NVIDIA announced Koa, a CRM reasoning model for Agentforce built on NVIDIA Nemotron. Salesforce post-trained Nemotron 3 Super on a proprietary synthetic dataset modeled on nearly three decades of CRM deployments; no customer data was used to train Koa. In Salesforce’s CRM benchmark, which includes opportunity updates, case routing and follow-up scheduling, Koa matched or exceeded leading model performance with three times fewer errors. Salesforce controls Koa’s model weights and runs post-training and inference within its own trust boundary. Koa is in use inside Salesforce and is moving into customer pilots with Baxter Credit Union, 1-800Accountant, Engine, Formula 1, UChicago Medicine and Xero.

→ Action

Model risk: Test Koa on case routing, opportunity updates and follow-up scheduling before pilot use.

Read article →  from Salesforce

Apple starts rolling out Siri AI across its platforms

For the fraud head at a regional bank

Do not treat a Siri-assisted payment as proof of clean consent when Siri can act on screen, camera and cross-app context. Fraud teams need to test how a coached customer can reach a payment action before they spot the scam.

Apple began rolling out Siri AI, its Apple Intelligence assistant, across iOS 27, iPadOS 27, macOS 27, and other platforms. Siri AI can use personal context across messages, emails, and photos, answer questions about screen content, and retrieve current web information. Siri mode in the iPhone Camera app can help users split a bill and pay friends with Apple Cash, or add a custom pass to Wallet. A dedicated Siri app uses iCloud to sync conversation history across Apple products, while Call Context can surface confirmation codes during calls to businesses. Siri AI is rolling out as an English beta, with five additional languages due in October.

→ Action

Fraud operations: Run scam drills where a caller coaches customers to use Siri on onscreen content or to split a bill with Apple Cash.

Read article →  from Apple

Get this in your inbox every morning

Free · No spam · Unsubscribe anytime