We at XpertRule don't like to be buzzkills but we also pride ourselves on being realists. 

So in the spirit of realism, it’s time to pour a little cold water on the fairytale that so many AI vendors are spinning right now.

The three-act story goes something like this: Contact centre adopts Gen AI; Customer satisfaction and retention soars; CX leader lives happily ever after. 

It may sound like a contact centre dream come true but the truth is that there are serious issues with Gen AI that need addressing before it can become a reality. 

The good news? We know how to solve them. But first, let’s map out the problems.  

Gen-eric AI: why some language models are too large for the contact centre

Contact centre conversations are out-of-domain, frequently stuttered or irregular in nature, and often very specific to your policies and services.

Meanwhile, large language models (LLMs) are trained on large corpuses of text - text that is not necessarily specific to your business or your industry.

There are many scenarios in which a huge diversity of data — for example, ChatGPT drawing on the entirety of the internet to generate quick responses — is useful. Contact centre interactions are arguably not one of them: the larger the model, the more inference it needs to make, the more potential for biases, and the less specialized and accurate your responses to customers. 

The risk is that using an LLM for frontline service doesn’t improve your customer experience at all, because while it improves pure efficiency in terms of handling call volumes, it garbles everything else: tone, accuracy, quality.  

You wouldn’t buy a Ferrari and drive it like a Toyota. So why would you adopt an LLM and use it like a glorified chatbot? 

What happens when it hallucinates? 

Then there’s the issue of safety. Between inaccuracies, compliance violations, “hallucinations”, and biased and opaque outputs, there are a whole host of risks associated with bringing Generative AI into your contact centre.

It’s often said that the solution here is to have agents on hand to review and make edits to AI-generated responses, offering themselves as human shields against the much-publicized “hallucinations” of LLMs.

But even this is harder than it sounds. See, ChatGPT isn’t just inaccurate. It’s been shown to be dishonest about being inaccurate, willingly misleading users with made up facts. Alarm bells should be ringing here for any CX leader interested in safety and compliance. 

So in a highly regulated corporate environment, technology as powerful as Generative AI needs a safety net; a layer of good sense and human expertise to buffer its risks, while optimising its considerable capabilities. 

Decision Intelligence is perfectly placed to be that layer. Here’s why. 

Integrate Gen AI into your contact centre with Decision Intelligence.

The goal for any CX leader should be to maximize the efficiency gains of LLMs in the contact centre while ensuring they’re safe, human-centred and on-brand.

But how do you infuse your language model with the unique tone, information and policies of your brand and products, while guarding for any potential inaccuracies? That sounds like a lot of responsibility to simply lay at the feet of your agents (even your very best ones). 

Here’s something you can lay it at the feet of instead. Put simply, viabl.ai, our Decision Intelligence platform, is good — very good — at capturing, embedding and scaling the human knowledge and logic that LLMs lack. Domain experts can use its intuitive dashboard to design, build, manage and optimise knowledge-based workflows and APIs which can be integrated with AI and used to support your agents.

These workflows can incorporate LLMs — and the specific tasks they are incredibly powerful at completing, like processing heaps of unstructured data — into a wider tech ecosystem that can automate all of the necessary checks and balances required of Generative AI.

Generative AI alone is not smarter or better at customer service than a human worker. But Generative AI combined with the human expertise that Decision Intelligence embeds? Well, that’s a different story.  

Talk to us about how we can support your customer experience and contact centre transformation.