Conversational AI (mostly seen in the form of speech and chat bots) has become integral to many contact centres. That’s despite the fact that they’ve only been able to perform basic tasks – in many instances mirroring bad practises from traditional IVRs. It’s traditionally been based on rigid scripting, limiting its conversational capabilities on narrowly pre-configured scenarios, usually causing high customer effort and consequent frustration.
However, with emergence of generative AI technology like ChatGPT and Google Bard, we are seeing a major breakthrough in this area. Generative AI will very quickly revolutionise customer care bots’ capabilities because it is really able to communicate in human-like fashion. ChatGPT can handle complex issues that require deep understanding of company’s business and customer’s situation. It can also recognise the context of conversations, which enables it to provide personalised answers in relevant and helpful manner – similar to the way humans do.
Naturally, contact centre directors are excited about these capabilities, but they’re also being realistic. This revolution won’t see all customer services handled by generative AI alone. And that’s simply because that’s not possible.
Here’s why: Despite its human-like conversational capabilities, a generative AI’s knowledge is general – it doesn’t know anything about your company, its offerings, policies, and procedures – beyond what it can scrape from the web. On its own, the app doesn’t have access to customer touchpoints and communication channels like phone, Web or social messaging. It cannot access your CRM or the other backend systems necessary to support different processes and it cannot handover interactions to live agents when a customer’s inquiry goes beyond its knowledge and capabilities.
But you can use it if you take the right approach. For use in customer services, generative AI needs to be properly embedded into the contact centre infrastructure that provides the integration and orchestration layer. This means it can be trained in your company’s specifics and be enabled to reach to other systems and live agents.
This means leveraging the investments you may have made in conversational AI – the integration of these two types of AI is a formula for success. Generative AI provides advanced conversational abilities while conversational AI can take control of the bot with a framework for training and management. This will ensure that bot has relevant knowledge about the company and can also prevent it from falling off the track or even ‘hallucinating’ during interactions with customers – we’ve all seen the creepy examples of ChatGPT going ‘rogue’.
Conversational AI also provides necessary back-end integrations and means for collaboration with live agents. Going further, generative AI can even improve your conversational AI systems, speeding up bots’ training by automatically generating intent prompts, lexicons and even entire bot flows. Generative and conversational AI perfectly fit together by complementing and supporting one another.
Generative AI can also revolutionise agent assistance. If properly integrated into your systems, it can present agents with highly relevant information related to any interaction, displaying relevant articles from knowledge management, guiding agents through complex workflows, popping up relevant tools, filling out CRM fields, and summarising conversations to be inserted into ticket notes.
Get all this right, and you can expect solutions like this to significantly improve First Contact Resolution (FCR) and Customer Experience (CX) while decreasing Average Handling Time (AHT) and After Contact Work (ACW). Indeed, the technology is so advanced that it has the potential to make difference between companies who will become leaders in both CX and EX – and those who will be lagging.
But to be clear: ChatGPT is no silver bullet that can solve every customer service woe. The processes behind it need to be carefully designed with a mindful ecosystem of measurements that will ensure proper control, management and a constant improvement cycle.