Agentforce Voice Never Sleeps: The Florida Prepaid Case Study

Every contact center transformation project starts with the same question: How many calls can we automate? Florida Prepaid never asked that question.

Florida Prepaid manages college savings plans that families open when their children are young and maintain for years. More than 200,000 accounts. Every phone call is not just a transaction; it is a moment that either builds or damages trust in a relationship the family will carry forward for decades. This awareness is what set Florida Prepaid’s Agentforce Voice implementation apart from others.

Not Starting With the Wrong Question

At the beginning of the project, the team asked itself: Which calls require a human, and which do not?

This distinction is critical. Most contact center projects are built around maximum deflection; Florida Prepaid’s was built around redesigning the service model. Artificial intelligence for routine questions such as enrollment, eligibility, and plan usage, and human capacity for matters that require real judgment.

Chief Information and Technology Officer Ashley Falls summarizes this approach as follows: “We did not view Agentforce Voice as a deflection tool. We saw it as a way to scale responsibly.”

From 10% to 75%

On the first day, only 10% of calls were routed to Agentforce Voice. The scope was intentionally narrow: only publicly available knowledge articles, no account data, and no identity verification.

After weeks of monitoring, performance improvements, and growing organizational confidence, the rate first increased to 30%, then to 50%. Today, 75% of calls received during business hours are handled by Agentforce Voice. Calls managed by artificial intelligence last an average of 1.6 minutes, while conversations handled by human agents are completed in an average of 10.9 minutes.

Starting narrow was not a limitation. It was the only way to prove value and build the trust needed to expand.

The Midnight Advantage

The after-hours picture needs to be considered separately. Previously, a customer calling at night had only one option: call back the next day. Now, 100% of incoming calls are answered by Agentforce Voice. Routine questions are answered immediately, while a Salesforce case is created for the following day for more complex matters.

After-hours operations are the easiest area to capture in a contact center project. There is no handoff scenario, no risk of sensitive conversations, and no agent displacement. It is the starting point where the agent owns the channel independently, use cases are clear, and value is immediately visible.

Where Did the Agents Go?

Agentforce became so effective in the chat channel that approximately 35% of agents were reassigned to proactive, revenue-focused tasks. The fact that average handling time increased by approximately two minutes for conversations requiring escalation is not a negative signal; it shows that agents are able to spend more time on matters that genuinely require attention.

Who Closes the Loop?

When Agentforce Voice went live, Florida Prepaid’s contact center team did not step back; on the contrary, they stepped forward. Agents review transcripts, flag gaps, and feed improvements back into the configuration. NeuraFlash developed custom tools for call sequencing and tailored deflection reports.

The result: an agent that improves with every cycle under the guidance of the people who understand customer relationships best. The technology scales, while the judgment shaping it remains human.

Florida Prepaid’s story is not an artificial intelligence success story that scaled rapidly, but a story of trust that grew deliberately and slowly. There was no goal of maximum deflection and no claim of overnight transformation.

Instead, there were clear boundaries, the right data, escalation where human judgment was required, and a continuously learning loop. A customer calling at 2 a.m. no longer listens to a recording. They receive an answer. But there is still a human hand behind that answer; not on every call, but in making the system a little better every day.