How Coveo Relevance Generative Answering is addressing a top AI concern
If the biggest question for customer experience leaders about generative artificial intelligence (AI) is, “Can we really trust this technology?” Coveo is now live with an extension to its platform that provides a definitive and positive answer.
As it announced last week amid the Dreamforce conference in San Francisco, Coveo is now live with Coveo Relevance Generative Answering across its own customer self-serve experiences, including in-product help, documentation and community. The solution allows enterprises to combine search with generative answering based on Large Language Model (LLM) technology, along with secure indexing and AI relevance capabilities.
According to Bonnie Chase, Coveo’s Senior Director of Service Product Marketing who led a session about the company’s approach to GenAI at Dreamforce, most business leaders recognize these are still early days for the technology. However, they are trying to strike a balance between leveraging its capabilities while also ensuring they aren’t exposed to damaging errors or other risks.
“What people want to know is, once an answer to a question is generated through AI, how do you know that it’s not hallucinating? How do you know that it’s telling you the truth?” she told 360 Magazine. “A lot of that’s grounded in making sure that you have a really good content strategy, but you also need to have validation mechanisms.”
Coveo Relevance Generative Answering, for example, gives users citations to the sources – such as a knowledge article – alongside the answers to back up the information it provides.
The need to take a ‘total experience’ approach with gen AI
Chase said the majority of companies who participated in Coveo Relevance Generative Answering’s beta program have started with internal use cases, but some have been quick to deploy it for customer-facing tasks as well. New Zealand-based cloud accounting software provider Xero, for instance, went live with Coveo Relevance Generative Answering in little more than a week to provide greater self-service capabilities to the small businesses who visit its corporate website.
At Dreamforce, Chase led a session that tried to get business leaders to take a more holistic approach to the way they incorporate innovations like GenAI into their operations.
“What I want people to remember is that the technology is there to support an experience that drives outcomes,” she said, pointing to tools like chatbots that were quickly deployed but have often led to disjointed experiences. “Everyone’s kind of scrambling to figure it out and make something happen. I caution against just getting the technology out.”
Instead, Chase recommended having an internal discussion about GenAI early on to determine where it’s best suited to provide answers. They should also think about who within and outside an organization should get access to different kinds of answers. In fact, she said Coveo Relevance Generative Answering offers privacy settings and other permissions that can provide the safeguards enterprises expect. Coveo also has a zero retention policy, she added, which means you control how long your index data is kept, and the company is not using customer data to train its GenAItools.
How customer feedback leads to the best gen AI solutions
While many GenAI products have been developed behind the scenes with little transparency, Chase noted that Coveo gathered an advisory group for Coveo Relevance Generative Answering made of more than 25 customers and some 20 design partners, including Zoom, Xero, VMWare, Synopsys and Informatica to name a few. This provided invaluable feedback for the solution’s GA release later this year, she said, such as the need to provide analytics on how GenAI can impact the bottom line through call deflection.
Working directly with customers has also helped Coveo understand where it needs to offer further education and manage expectations. While Coveo Relevance Generative Answering may provide answers in multiple languages, for instance, Chase said they probably shouldn’t rely on it as a translation tool. There have also been a number of questions on how to ensure good knowledge management using AI.
“What I’ve been recommending has been the typical knowledge management best practices: making sure that your content is easy to read, and that you’re chunking the paragraphs versus just dumping a lot of information into one place,” she said.
GenAI may not be a silver bullet for every business need, Chase added, but when organizations deploy it they will be looking for tools and platforms that provide confidence in how it will impact customer and employee experiences.
“We feel strongly that we’re the best solution because from over the last decade we have been developing AI and working to index enterprise content, with secure connectors to hundreds of different systems, so we have that foundation,” she said. “Now it’s just adding the generative process into it. This is really about making these solutions more robust and able to offer the creative answers people want and need.”
