6 Use Cases for REGAL AI Call Summaries

Three months ago we shipped what we thought was a simple “set it and forget it” AI feature: REGAL AI Call Summaries. REGAL AI Call Summaries are – as they sound – AI-generated summaries of call transcripts. We’ve now gathered feedback from customers on the different ways they are using the feature and what more they’d like to do with it. Some of the answers might surprise and inspire you.

Where We Started

One of the challenges of running a successful Sales or Support team is transferring knowledge from one agent to another about a customer when so much of the context is locked up in previous call conversations. To customers, this problem manifests as having to repeat themselves every time they interact with your brand. 

The original motivation behind REGAL AI Call Summaries was to give agents and managers a quick way of understanding what happened on a call they weren’t a part of: 

• Agents could use the summaries to review a prospect's or customer’s recent calls before hopping on the next, so that they’d have more context. (Goal: Better Agent-Customer conversations.)

‍

• Managers could use the summaries to quickly review calls – rather than reading full transcripts or listening to recordings – giving them more oversight over their call center in less time. (Goal: Manager hours saved reviewing calls.)

‍

Prompt Engineering

We wanted to produce a single call summary that could satisfy both of these use cases. The summary needed to give sufficient context yet get to the point as succinctly as possible. To achieve this, we landed on the following prompt: “You are SummarizationGPT at a call center that makes sales calls. What was the result of this call?” 

We tested several other iterations of prompts against a sample of historical transcripts, including e.g., “Summarize the call” (resulted in more of a “teaser” than a “reveal”), “Synthesize the most important points of the call” (didn’t produce meaningfully different results), and “Critique the call” (focused on coaching points rather than call outcome).

To demonstrate the key differences between these prompts and the results they produced, here’s a corollary of each of the prompts applied to summarizing the movie “The Sixth Sense.” 

Prompt 1: Summarize {“The Sixth Sense”}

• Output: The Sixth Sense" is a supernatural thriller about a boy who can see and communicate with ghosts, seeking help from a child psychologist to deal with his abilities. As their interactions progress, unexpected twists and revelations come to light, leading to a surprising and emotional conclusion.
• 🚫 Accurate, but after reading this summary, you’d still have to watch the movie to know what happened.

‍Prompt 2: Critique {“The Sixth Sense”}

• Output: The Sixth Sense" is a chilling mystery thriller that centers around a young boy who can see and speak to the dead. The suspenseful atmosphere is maintained throughout the movie, with unexpected twists and a jaw-dropping revelation in the final act. M. Night Shyamalan's direction keeps viewers on the edge of their seats as they uncover the truth alongside the characters. The film's haunting score and compelling performances, particularly from Haley Joel Osment and Bruce Willis, make it a must-watch for fans of suspenseful cinema.
• 🚫 Contains lots of descriptive words that judge the quality of the movie and its impact on the audience (“suspenseful”, “jaw-dropping”, “haunting score”, “compelling performance”), but again, you’d still have to watch movie to know what happened.

‍Prompt 3: What was the result of {“The Sixth Sense”}?

• Output (🚨Spoiler Alert🚨): In the movie "The Sixth Sense," the result is a shocking twist ending that reveals that Malcolm Crowe, the child psychologist played by Bruce Willis, has been dead the whole time. Throughout the film, he has been trying to help Cole Sear, a young boy who can see and communicate with ghosts, but it is ultimately revealed that Crowe himself is a ghost. The movie ends with Crowe realizing his true nature and finding closure with his unresolved issues, allowing him to move on to the afterlife. The twist ending is a major revelation that changes the audience's understanding of the story and characters.
• ✅ This summary doesn’t just “tease” the surprise ending; it reveals it in the first sentence, and then provides more context. The boy sees dead people and the psychologist helping him has been dead the whole time. The fact that this is the only prompt that required a “Spoiler Alert” explains why we selected it.

‍

6 Customer Use Cases

REGAL AI Call Summaries have now been live for three months. In that time, from surveys and speaking to customers, we learned of six different ways customers are making use of the feature to improve their business outcomes:

1. “My agents find them motivating” – for this brand, agents expressed that customer calls can be nerve wracking, and after completing a call they tend to dwell on all the things they did wrong. But when they saw the call summary of their own call that focused on the key points and outcome, they realized it went better than they remembered. 
2. “It saves my agents note taking time” – this brand used to require that agents submit notes when summarizing a call so the next agent had proper context, and now they no longer need to, saving ~30 seconds of agent post-call work on every call.
3. “They make it easier for my managers to QA calls” – this brand likes the summaries because it allows them to review many more calls in less time, and only dive into the transcripts/recordings that warrant a deeper review.
4. “It’s useful for updating my CRM” – this brand ingests the REGAL AI Call Summary into their CRM so that their business teams that are not in Regal can also benefit from knowing what happened on calls in Regal.
5. “We pump them into a shared Slack channel for VoC” – this brand uses Regal’s Journey Builder to trigger a Slack notification to a shared Slack channel each time a summary is produced so their product managers can keep a pulse on customer conversations.
6. “I like sharing the summary with my customers” – this brand sends a follow up text to their customers after a call with the summary of what was discussed. This was the most unexpected and novel use case to us, and seemed like it could generate unexpected customer delight.

‍

What’s Next

Given the varied ways customers are using REGAL AI Call Summaries, one of the things we want to explore is producing different types of summaries for different intents. For example, perhaps the “Customer-facing AI Summary” should use a different prompt and be sure to strip out anything negative. For the “Agent Note Taking AI Summary”, perhaps customers want to configure their own prompt for each type of call – e.g., an acquisition call might warrant different notes than a retention call or support call. We also want to experiment with different formats – such as bullet points or highlighting certain terms – to improve readability. 

Finally, as a complement to the unstructured summaries, we want to start extracting more structured AI Insights such as “Intent(s)”, “Pain point(s)”, “Objection(s)”, “Outcome(s)” etc. to enable more automated use cases. We already support “Customer Sentiment” and “Agent Sentiment” as well as Call Trackers that extract certain phrases brands want to track, but inferring those key pieces of structured information without customers having to know what to configure is even better.