How Cresta Creates Experts on Day One?
#005
Mobile Phone - Set to Landscape (Horizontal) view for better visibility
GOLD NUGGETS
Sources - They are clickable if you need more information
What's Cresta
Cresta Platform Engine
Let's check the Expertise Engine
Cresta Platform Engine
Sales Agents Problems
Consider the workflow of a contact center agent communicating with a customer. A conversation can have many different endings. The agent must rely on her expertise to optimize the outcome. She continuously evaluates the possible behaviors, expressions, questions, or offerings that would help close a deal. Naturally, experts produce better outcomes than average performers. Cresta captures the expertise of your best agents and distributes it across your entire team. Cresta’s AI observes expert agent behavior, identifies high-leverage actions, and weighs the rewards of each action based on prior outcomes. Actions with the highest values are converted into suggestions and shared with the rest of the team at the right moment. It’s as if every agent is your best agent.
We’re forgetful. It’s human nature. In 1855, Hermann Ebbinghaus illustrated our forgetfulness with the Ebbinghaus forgetting curve, with some studies suggesting we forget 50% of new information within an hour of learning it, and 70% within 24 hours.
Problem
When a sales representative joins a new company, they spend about 6 months learning the ropes and gaining experience until becoming completely effective in their roles. Traditionally, the agents are coached by a sales leader who would listen to conversations and jump in at the right moment to guide the representative. The sales leader may tell the agent to take actions like “Ask for a sale”, “Discover customer’s needs”, “Greet the customer” etc. In the context of live chat, it’s valuable to have a real-time coach to guide responses.
Cresta’s platform helps to solve this training problem by creating individualized responses for each agent. At appropriate moments in the conversation, sales agents can use these suggestions to answer customer inquiries more quickly and more effectively. Our models analyze the ongoing sales conversation to understand the situation, and then suggest the next best response to say according to the next best action to take.
The idea is to have a model that will generate the response by itself. Explain GPT-2 here which gives the response by itself.
Mention that the probably use GPT-3 Now.
However, the problem comes in the next step. what answer to choose from?
To improve the probability of agents making a successful sale, we also want our models to generate a response according to the best next action in that situation. Certainly, a model that generates text that is all over the place won’t suffice.
we want to generate goal-directed responses following a call-flow that maximizes the business metric of sales conversion.
Solution
Action Directed GPT-2
How does the chat look now with Actions
Results Compared
Explain BLEU Score
Case Study Improvement
Zayd Enam - Co-Founder at Cresta
Or is it...
2021
Embracing Tech Readers
Congratulations!
For those that really liked the case study...
How did you like the case study?
You want to share the case study with others?
You want to receive news about new case studies?
You want to talk with me about AI use-cases or content creation?
Cresta
By Tomas Pinjušić
Cresta
- 43