GenAI Studio should enable organizations to customize general-purpose tools like OpenAI for use in the contact center with just a few clicks.
The news keeps rolling in from Enterprise Connect. This week Five9 announced a solution it hopes will help organizations customize common generative AI models for the contact center. Last week, Five9 conducted a pre-briefing with three Five9 executives, including Matt McGinnis, vice president of product, industry, and solution marketing; Sarika Prasad, senior manager of product marketing for AI and automation; and Jonathan Rosenberg, CTO and Head of AI. During the briefing, it introduced a new tool meant to help contact centers customize GenAI models to better meet their specific requirements.
McGinnis set the stage for the discussion. “[GenAI Studio is] a bit of a forward-leaning preview for us,” he stated. “It’ll be available in the second half of this year. It starts to set the tone for what this means for AI and how we will manage the different AI features and capabilities on the intelligent CX platform.”
The need to create bespoke AI features for CX is crucial. While there are many off-the-shelf LLMs, general purpose won’t meet the needs of contact centers long term. This is a highly customized area of a business where minor variances in recommendations or analyses can have a big difference in agent insights, which in turn significantly impacts how customers perceive a brand. Customers are still experimenting with AI today, so Five9’s future-leaning comment is appropriate, as it’s more important to get the AI right than get it out now.
Customizing General-Purpose AI Tools for the Contact Center
Five9 says that Prasad told me they have talked to their customers and heard feedback on Five9’s post-call AI summaries.
“We have heard our customers say they are thrilled with the summarization feature,” she told me. “They are looking for more precise customer summaries. For example, they wonder, ‘Can I get three-line summaries? Can you insert our call-attached variables into the summaries?’ And then they have requests like, ‘Can you translate a Spanish call into English and provide that summarization?’”
Rosenberg added: “The big thing with generative AI and all these general AI applications is that they are the only contextual data, and that’s the stuff that you stick into the prompt. In the GenAI world, that’s how you use the tool. With ChatGPT, every consumer in the world can use it and how we use it for a chatbot or an Agent Assist is largely to do with the prompt.”
Prompts are critical to GenAI—they contain the contextual data to instruct the model on what results are expected. Rosenberg explained: “For example, if you wanted to use GenAI to summarize a call, you need to provide instructions to summarize the call: ‘Please summarize the following call between a contact center agent and a customer.’ In this case, I need to give it the data to do that job—the transcript of the conversation.”
Five9’s GenAI Studio features include:
- Choice of Generative AI models: Five9 says they’re offering a range of choices—from third-party to proprietary generative AI—that will enable customers to utilize the model that makes the most sense for their business.
- Integration of various contextual data forms: GenAI Studio can answer questions, summarize calls, and guide agents using personalized, business-specific information.
- The ability to test customized models: To gain more control over outputs, GenAI Studio will provide testing ability using actual call transcripts to understand performance and make improvements.
- Access to a repository of sample prompts: GenAI Studio will provide access to prompts so companies can get off to a good start, speeding up prompt development and reducing the need for prompt engineering expertise.
The company also says that GenAI Studio will power all Five9 AI apps, starting with Agent Assist AI Summaries. Customers should gain a good deal of control over the summaries—being able to format them as needed, add customer sentiment, and condense or add audience-based context.
Final Thoughts
Five9’s approach is very practical and will be beneficial. Users of ChatGPT often struggle with understanding how to structure prompts and then further refine them to get the results they want. In a contact center, this could lead to frustration as a “sort of right” answer won’t provide the results a user is looking for.
A prompt library and customization options can help organizations get off to a running start. This will allow them to harness the power of GenAI faster and more accurately.
