Hands up anyone who can honestly say they haven’t asked Chat GPT something akin to the following by now:
- The meaning of life (preferably in the lyrics of Snoop Dogg)
- How to make a baby and then explain this uncringefully to my own offspring
- To write a highly-paid job application in the style of an already-qualified person
In the last nine months, Chat GPT has gone from a weird phrase no one’s heard of to a household name. That’s one hell of a gestational period for this LLM, which has broken records with its user interest (and weird requests). But before we start actually exploring some of these AI acronyms instead of casually throwing them around, we’ll state that we’re about to take a look at what the birth of generative AI (last one) means for the contact centre. What’s more, we’ll share how we can leverage generative AI, and more specifically LLMs like Chat GPT, to our advantage…
About Those Acronyms…
Chat GPT: Stands for ‘Chat Generative Pre-Trained Transformer’ and was released in November 2022 by the non-profit research agency OpenAI. They trained this Large Language Model (LLM) to understand complex questions and structure eerily articulate responses, that basically sound like a clever and empathetic creature called a human must have written them. The platform broke records by attracting one million users within five days (Statista).
LLM: A Large Language Model is a technology designed to generate human-like responses. Whereas your basic language model might get a robotic gist (or not at all vis a vis a chatbot saying ‘I’m sorry I didn’t quite understand that), an LLM should have all the vocab, nuance and complexity to sound intuitive, reasonable and capable. The complexity of any LLM largely depends on the size of its data set and Chat GPT’s is the entire world wide web, or 300 billion words. Given that there are one million words in the English language…this LLM is massive. MLM anyone? No ok we’ll stop.
Generative AI: Is kind of an umbrella term for artificial intelligence that can generate content including text, images and audio. It’s actually not new, and first came about in the 1960s in the form of chatbots. The way generative AI works is to use algorithms to learn from sets of ‘training data’ and subsequently create new examples of the same information. Which is why kids think ChatGPT is unmined gold when it comes to plagiarising GCSE essays. (Finextra.com)
What are the benefits of Chat GPT in the contact centre?
- Good chat
The glaringly obvious sell of Chat GPT in a contact centre is its suave communication skills. This technology can hold an intelligent conversation, which has the potential to streamline and enhance agent and self-serve processes. Rather than leaving customers to get frustrated with incoherent chatbots, or a scarcity of agents, Chat GPT has the potential to answer questions efficiently and in-line with our business’ tone of voice which offers an efficient and radical personalisation aspect.
- Agent Support
As well as its language skills, Chat GPT’s pool of training data is huge, which means this technology boasts expertise and knows lots of answers to lots of questions. Whether these are specific to our business and helpful to our customers, though, is another issue we’ll explore further down. But when this type of generative AI is supervised by agents, it has strong potential to streamline their workload by offering response options to customers in real time.
- Strong Signposting
Inevitably there will be enquiries that should and will be dealt with by real people, rather than Chat GPT, within a contact centre. But Chat GPT has the potential to reduce wait and transfer times for our customers, by quickly getting to the bottom of who needs to speak to who and helping to direct chatbot/IVR enquiries to the right agents and skills groups.
Now Give Us The Cons…
- Wrong Answers
Chat GPT has the power to sew together human-like responses and slip through conversations like silk. But as we know from politicians, it’s possible to sound smooth whilst spouting rubbish. The problem with the vastness of Chat GPT’s training data is that it may not serve the niche specifics of our businesses. Unless we arrange for this LLM to be plugged into our back-end systems which details things like client info and product delivery and order journeys, it won’t be able to offer our customers accurate answers to their burning questions. For example, if Frank’s t-shirt order has been put on hold because Arlo in the warehouse noticed that the packaging is damaged, Chat GPT won’t be privy to this intel unless it’s talking to our internal systems.
In this scenario, Chat GPT unfortunately behaves even more like an MP. Because when Chat GPT doesn’t know the answer to a question, it doesn’t stop chatting. Instead, it plays the guessing game and says something very articulate and inaccurate. Which means, much like many politicians, Chat GPT is not a standalone solution to anyone’s problems. In our contact centres especially, Chat GPT needs agent supervision. The LLM might support our agents’ ability to communicate quickly and effectively under high customer demand, but this generative AI is not yet ready to replace their expertise.
- Security
Feeding specific data into Chat GPT using a secondary system is one thing, the sensitive handling of this is another. Once trusted with our customers’ data, Chat GPT can’t just draw from and reproduce this any time it wants, it needs to honour GDPR rules like us humans. Security is therefore a concern for an LLM that likes to eat knowledge and spit it out again. This in turn raises a question about how easily Chat GPT could be integrated with our back-end delivery systems without giving the LLM indiscriminate access to sensitive data like contact and banking details. But without this information, can it really answer the questions that our customers are asking?
In a similar vein, we’re curious as to whether Chat GPT understands boundaries. Safeguarding limits need to be imposed on customer conversations that could be tempted off-script, like asking the LLM out to dinner or their views on Trump. (NoJitter.com)
- Text-focussed
Being a language model, Chat GPT is largely focused on text rather than rich media. But here at Adexchange we know the most powerful content for resolving issues is rich content. We’re talking videos, screen walkthroughs, animated explainers, infographics. All those bad boys. If Chat GPT can help deliver the right video to the right person, or show them the right infographic, then fantastic. But currently, this kind of media is not an LLM’s first language, so it’s limited.
The GPT gist:
It’s here to help us but not replace us. Even if we all wanted to retire from contact centre life tomorrow, we’d have to put those pina coladas straight down because we can’t leave Chat GPT unattended to talk to our customers.
What this generative AI technology CAN do for us is…
…provide a supportive skeleton to create better content for our customers, whilst being less taxing on our agents’ time. Chat GPT does this by offering informed, empathetic conversational experiences whilst chewing through vast amounts of information in real time. It can rapidly process requests and help to matchmake customer needs with agent skill sets, so offers intelligent routing in this sense. All of this is a fantastic opportunity to optimise our contact centres, we just need to harness it..
But what Chat GPT cannot do is…substitute agents, chatbots or even IVRs, or be considered a standalone solution. As it stands currently Chat GPT needs to be integrated with our business systems and monitored closely for inaccuracy, before it can be usable.
The Adexchange Take:
Here at Adexchange, we’re advocates of human-in-the-loop AI, which is where machines augment human systems but don’t replace them. It’s a great way of trickling down technical expertise and automating tasks without compromising on service. Chat GPT paired with specialised agents could allow our contact centres to achieve a level of customer service that neither humans nor machines could achieve on their own. We say: swap notions of direct replacement with the rich potential of a partnership.
If you’d like a hand working out where your contact centre could benefit, but not be mercy to, the advantages of AI technology – give us a ring. We’d love to help you capitalise on the sweet spot and get right ahead.