Given the excitement created by ChatGPT, you must be thinking how Generative AI can benefit your enterprise.
ChatGPT uses GPT (Generative Pre-trained Transformers) from OpenAI to provide you with a powerful conversational tool that can answer your questions in human-like English. While ChatGPT has been trained with a ginormous public dataset, it has no knowledge of your reference data for Customer Support. What you need is a Generative AI platform that is purpose- built for Enterprise Customer Support.
Generative AI Technology
Generative AI is part of Deep Learning technology involving large neural networks, called large Transformers, where the neural network is able to generate information such as summary from a text, answer to a question from a document or generate an image from a description. GPT is the Generative AI technology from OpenAI. Other companies such as Google and Anthropic also have very large Generative AI models such as PaLM and Claudia.
While Generative AI has been around for a while, in the last 4 years due to advancement in hardware such as GPUs and TPUs, it is now possible to generate the summary or answer to a question with very high accuracy, since you can have a neural network now with billions and billions of nodes. For example, the GPT 3.5 model from OpenAI is trained on about 45 Terabytes of textual data and has 175 billion parameters to perform natural language tasks rivaling human performance.
How can Generative AI be used for Customer Support?
GPT technology (and other Generative AI technologies) can take a question and a document (plain text document) to generate an answer in natural English, if the question can be answered by the content of the document. Otherwise, it returns “no answer”. This can be used to provided sophisticated self-service for customers, along with solutions to boost the productivity of your support staff.
Gaps in Generative AI for the Enterprise
Consider a high tech or industrial enterprise with complex products. The Reference Documentation probably includes 1000s to millions of documents -- in PDF, HTML, Microsoft Word/ Powerpoint/ Excel, Google Docs/ Slides/ Sheets, Relational Database Tables, and XML documents. You probably have a lot of images, videos, and even drawings. Some of these documents can be 100s or even 1000s of pages, present in silos (web, Sharepoint, Google, Confluence, Salesforce, ServiceNow and more). Not only that, often times the solution to a support query is scattered across multiple documents.
Given your reference data, in order to use Generative AI like GPT (or ChatGPT) to generate an answer for a support issue, you need to first generate a Shortlist of possible 10 – 20 Answer Segments (sections of reference documents which may have the answer) . Without this Shortlist, no Generative AI platform will have any knowledge of your business and thus, will not be able to generate any answer.
Generating the Shortlist in real-time for a given support ticket or issue is complex, particularly given that you need high accuracy at a low cost to make it a viable solution.
Quark.ai, a Generative AI platform for Customer Support
Quark.ai’s Autonomous Support Platform is a Turnkey Pre-trained Generative AI solution for bringing efficiency to Customer Support through automation. The platform is purpose- built for Customer Support, and therefore pre-trained with an extensive dataset for this space, enabling Quark.ai to onboard a customer rapidly in weeks.
To start with, we ingest all your reference data – structured data in relational databases, XML or CSV, unstructured data in PDF, HTML, Word, PPT, Google docs and more, and data from real-time applications like SAP.
Next, when a Support Query is input into this platform, Quark.ai processes it as follows –
It interprets the Support Query, which could be a support ticket, keywords or a natural language sentence, to determine the issue being discussed.