Traditional enrollment can be messy and inconsistent. It often relies on agents to manually decide who gets enrolled after a call—an extra step that’s easy to skip or get wrong. When that happens, good callers don’t get enrolled, bad actors sometimes slip through, and teams are left with unreliable data and extra work. The result is a process that slows things down, increases risk, and puts unnecessary pressure on agents.
That’s why we’re introducing Smart Enrollment, a new feature for Pindrop® Passport that makes it easier for Pindrop customers to enroll and authenticate callers. Instead of relying on agents to manually trigger enrollment, Smart Enrollment handles it automatically based on how callers interact with Interactive Voice Response (IVR) systems and agents. That means a smoother experience, less friction for teams, and faster, more cost-effective interactions.
How Passport enrollment works
Passport uses a set of customer-defined policies to recommend whether a caller should be enrolled after a call. These rules look at what happened during the interaction and determine if that caller should be enrolled or not.
Today, part of that decision often depends on agents. After a call, agents can leave feedback to say whether the caller should be enrolled. In practice, that step is easy to miss. Agents may skip it, or the feedback may not always reflect what actually happened on the call.
When that input is inconsistent, enrollment becomes unreliable. Legitimate callers might not get enrolled when they should, while in some cases, agents end up enrolling bad actors.
Smart Enrollment automates agent action
Smart enrollment takes the guesswork out of enrollment by automating the process. It uses AI to analyze the near-end call transcript to predict whether a caller should be enrolled.
For example, if a caller provides their accurate authentication information, the IVR confirms it, and they’re able to access their account, that’s a strong signal they’ve been successfully authenticated—so they can be enrolled.
On the flip side, if a caller only provides a name, fails to confirm anything else, and gets rejected by the IVR, that’s a clear signal they shouldn’t be enrolled.
That’s the core idea: by understanding how the IVR (and agents) behave during a call, Smart Enrollment can make more accurate, real-time enrollment decisions.
The Smart Enrollment process
Here’s how it works, step by step:
Calls come in and get transcribed using Automatic Speech Recognition (ASR). Smart Enrollment only focuses on the near end, the IVR or agent, since that’s all it needs.
Next, the solution redacts certain sensitive information from the transcript.
The redacted text is then fed into our Smart Enrollment model. This isn’t an LLM; it’s a smaller, faster, purpose-built model trained specifically for this job. Each customer gets their own dedicated model version, which helps with both accuracy and security.
Finally, the model outputs scores for the IVR and agent stages (if both are available). These scores show how confident the solution is in enrolling the caller and feed into your specific authentication policies to make the final call.
Models are trained for your unique needs
We train each customer’s model using a semi-supervised approach.
First, our solution collects a batch of near-end transcripts (certain identifying information already removed). Then it runs those through a set of LLMs, which provide a preliminary enrollment score.
Those scores aren’t perfect. But when the solution combines outputs from multiple LLMs, our technology gets a solid enough signal to train our smaller, proprietary models.
So our tech uses LLMs to quickly label the data, enabling the training of fast, practical models built for this specific task.
The result: strong performance. Smart Enrollment can enroll nearly 70% of callers while keeping false acceptances under 1%, according to our internal testing.
Better accuracy, less agent effort, stronger enrollments
Smart Enrollment helps take enrollment off agents’ plates and creates a smoother customer experience.
Impact will vary. If you’re enrolling everyone today, you can expect enrollment rates to drop as the model filters out risky callers that your agents previously weren’t catching. On the other hand, if your first-time enrollment rate is low, you could see substantial improvements.
This feature is currently opt-in only. Reach out to your sales rep or your account manager for more information.
