A Credit Union Saved $1.6M by Making AI Listen

The Deployment

Agentforce, Salesforce’s agentic AI platform, has crossed $1 billion in annual recurring revenue, with revenue up 205% year over year. But the more instructive story sits inside one specific deployment: PenFed Credit Union now runs its entire operation (call center, mobile, web, and branches) on Salesforce, with 76 AI agents working alongside human staff. PenFed CEO James Schenck joined Salesforce’s earnings call personally to walk through the results, a rare move that signals how central this partnership has become to the credit union’s operations.

Why This Matters for BFSI and Collections

The specific mechanics of PenFed’s deployment are what make this relevant to collections operations, not just the headline numbers. An agent called Agent Wingman listens to calls in real time, transcribes them, and prepares an updated record, but it waits for the human agent to approve the update before it is finalized. That single design choice (AI assists and drafts, a human confirms) is the difference between an automation risk and an operational advantage in a regulated environment. The results were concrete: nearly $1.6 million in savings this year, call handle time down 10%, after-call work down 50%, and held calls down 40%.

What the Numbers Do Not Say Out Loud

It would be easy to read this as evidence that AI should simply take over more of the call. Schenck’s own framing argues the opposite: “It’s got to be right each and every time,” he said, adding that he wants employees doing knowledge work and building relationship trust, not manually entering what happened on a call. The savings here did not come from replacing the human agent’s judgment; they came from removing the mechanical burden of documentation around it. That is a narrower, more specific claim than “AI cut costs by automating the call,” and it is the one actually supported by the data. The after-call work reduction (50%) is nearly five times larger than the handle-time reduction (10%), which tells you where the real inefficiency was sitting all along: not in the conversation itself, but in what happened after it ended.

The Practical Read

For collections and BFSI operations evaluating where AI actually pays off, PenFed’s deployment offers a cleaner blueprint than the broader Agentforce growth numbers do. The highest-value place to start is not necessarily full call automation; it is the surrounding documentation and record-keeping work that consumes agent time without adding customer value. That is also lowest-risk to automate because a human still confirms the outcome before it is committed. A voicebot or copilot that drafts the case update, flags the next action, and waits for sign-off gives collections teams measurable efficiency gains without handing over the parts of the conversation (negotiation, judgment, empathy) that still require a person. The lesson from PenFed is not that AI should decide. It is that AI should carry the paperwork so people can do the parts of the job that actually need them.

[Read the full report]

Related Post

Oriserve’s Generative Voice AI Platform is Driving Strategic Transformation in BFSI Revenue Operations

 Oriserve (ORI), a bootstrapped startup with a team of over 100 professionals based in Mumbai and Delhi, is revolutionising enterprise communications as a next-generation voice-based Generative AI platform tailored for Banking, Financial Services and Insurance (BFSI). With over 1.2 billion conversations orchestrated globally, ORI is establishing a formidable presence in India and the Middle East […]

Banks Use Only 30% of Their Own Data for AI

The Bottleneck Only about 5.5% of organizations are seeing real value from their AI investments, and the gap is not usually the model; it is the data feeding it. In regulated industries, the most useful training data (transaction histories, complaint transcripts, identity checks) is also the data compliance teams are most nervous about exposing. Gartner […]

McKinsey’s 2026 AI Trust Survey Discovery

The Maturity Gap Behind the Headlines McKinsey’s 2026 AI Trust Maturity Survey, based on responses from roughly 500 organizations collected between December 2025 and January 2026, found the average responsible AI maturity score rose from 2.0 to 2.3 year over year. But that improvement masks an uneven picture: only about 30% of organizations reached a […]