CustomerGPT:
A.I. Objection Handling
CustomerGPT is a state-of-the-art educational simulation designed within the Articulate Storyline platform. The tool simulates AI-driven customer conversations, offering a unique opportunity for sales representatives to hone their objection-handling skills in situations tailored to their own products. Try the demo to experience it firsthand!
About CustomerGPT
How Does it Work?
At its core, CustomerGPT harnesses the power of the GPT-3.5 LLM (Large Language Model), couples it with intricate prompt engineering, and delivers responses in real-time, ensuring a seamless and dynamic learning experience.
Common Challenges
Past efforts to design interactive role-play simulations using generative AI have often faltered. When tasked with assuming a character's role, even the most advanced AI chatbots, like ChatGPT, will sometimes veer off track, get confused, or eventually even switch roles — imagine an AI customer suddenly acting like a seller in the middle of an exercise. Because of those issues, until now, generative AI chatbots have been too unreliable for use in role-based educational simulations.
The Solution
CustomerGPT addresses these issues by using different AI prompts for each part of the simulation, which ensures a consistent beginning, middle, and end of the conversation. For instance, a separate prompt is reserved for directing the chatbot to conclude the conversation, ensuring the dialogue has a clear closure. The AI-generated output from each prompt is integrated into subsequent prompts using variables, a technique known as "prompt chaining." This sequential and interconnected approach ensures that the AI maintains the trajectory of the conversation by understanding the context of everything up to that point, while also enabling the ability to reliably direct the conversation in a specific direction. Additionally, high quality Few Shot examples, integrated into all of the prompts, enhance the relevance and precision of the AI-generated outputs.
Real-time Interaction & Non-verbal Communication
Generated AI responses are relayed to Storyline in real-time using OpenAI's API streaming feature, enhancing the sense of immersion in a real conversation. Additionally, the expression of the Storyline character is dynamically controlled by artificial intelligence. This ensures that the nonverbal communication, reflecting customer responses like skepticism or interest, aligns with the AI-generated customer dialogue.
Measuring Results with CustomerGPT
CustomerGPT is designed with a robust measurement system to help L&D teams understand the progress and performance of learners.
Persuasion Power Score
Post-conversation, AI undertakes an analysis based on a prompt containing a benchmark rubric and model answers. The outcome is the "persuasion power" score - a metric assessing product knowledge, customer understanding, objection handling, and communication skills. Instructional designers can easily modify or add additional assessment criteria.
Database Integration
CustomerGPT has the ability to log dialogue, scores, and feedback in a database to provide a broad view of a learners' performance over time. The data can also be integrated with Natural Language Processing (NLP) tools to glean additional insights as well as integrated into Learning Record Stores (LRS) via xAPI statements.
