In recent years, customer service has been abuzz over the AI revolution and what it means for contact centres. Will chatbots spell the end of human agents? This was more hype than reality, but there has indeed been a revolution and its mascot is ChatGPT

AI needs custom training
Before ChatGPT, most people had never heard of conversational AI or Large Language Models (LLMs). Now, they are household names. Undoubtably, ChatGPT is remarkable and can do many useful things from writing to translation, but for conversational AI to be useful for organisations, it must be precisely trained for its industry.
    
Briefly, LLMs like ChatGPT are built using huge data sets, covering a broad range of topics. Using Natural Language Processing, they understand language and then generate an appropriate response, and also create ‘original’ texts and artwork from prompts.

Custom Language Models are the engines powering industry-focused conversational AI
However, general LLM-based chatbots tend to hallucinate (fabricate ‘facts’) over 15 per cent of the time, which is unacceptable for organisations. This is where Custom Language Models (CLMs) come in. CLMs train AI in an industry-specific language to be accurate and relevant.
    
In the end, organisations are not interested in novel tech; they seek solutions that offer compelling benefits and address real issues. And CLMs fit this bill perfectly.
    
For instance, in customer service, the public cares about trust and quick, accurate information, whereas for organisations compliance, costs and efficiency are important.
    
In the future, a proliferation of CLMs will focus on the regulations and language of individual industries. They will use similar tech, but for different purposes, much like how GPS is used differently by the military and athletes.

Relational AI chatbots help and connect
People engage with AI chatbots using unstructured language as bots understand intent (what someone is saying), pick out entities (dates, numbers, etc.) and use propensity to guide people to the best outcome.
    
Also, conversational chatbots are suited to helping people in financial stress - if they are trained to recognise vulnerability and to respond empathetically. They offer more than just FAQ answers; they combine empathy with efficiency.

CLMs are where the fun tech jumps into the business world
Custom Language Models take generalist AI and turn it into something powerful, showing benefits like reduced waiting times, reduced call volumes, 24/7 service and accessibility, all which help government departments fulfil their obligation to citizens.