What are the unique needs of B2B Customer Experience teams?
What should I look for when considering an AI solution for my CX team
Last updated
What should I look for when considering an AI solution for my CX team
Last updated
When customer experience leaders start considering what they need for their team it is often unclear what to look for. This topic covers the key needs we commonly see across CX teams.
Enterprise teams seek AI that is deterministic. What makes an AI deterministic
Repeatable results: The same question must result in the same/similar answer.
Factual answers: Answers must be based on a fact that is present in data sources
Traceability: It must be possible to trace an answer to the source of data
Enterprise CX teams need the ability to control the AI and train it to behave in ways that create value
Human-In-Loop: Team members need to be able to make edits to answers, these changes must be remembered and must be used to improve the answer for future questions.
Override Facts: Often facts change, for instance, a new release of a product may support an integration that was not previously supported. Users must be able to edit the AI to know this new and corrected answer
Persona-based data sources: Support roles work with customers already using the product and troubleshooting a specific scenario. While sales engineers work with prospects and customer success managers work with existing customers. They have different data needs and AIs must be able to recognize and act accordingly
Fine-grained control: Even after you have configured AI with the persona-based data, it is possible that certain pages/URLs/docs cannot be used for a specific answer. Users must be able to override AI responses to make these changes
Most enterprise data is created by humans. The problem with this model is documentation is often limited and covers only the most common aspects of the product.
Low-Density Data: Documentation and knowledge bases have high-density data and are available for AI, however conversation streams like ZenDesk, SFDC ticketing history, Slack, and other data sources reduce the quality of AI responses
Adjacent Products: Dependencies like knowledge of adjacent products, libraries, and domains are not covered in customer docs and lead to frequent "Unable to Answer" scenarios
Domain Topics: While companies document their own products they are unable to document the domain topics. This is often required to help improve the quality of AI responses.
