Demystifying AI for the Mid-sized call centre

The overwhelming attention to AI recently has led to some interesting situations for contact centre operators. There are a plethora of articles about the technology and the advanced uses of AI, but for most of us, we just need to understand how it affects our contact centre and our customers. Let’s start with the basics.

The three types of AI that matter to us can be defined as:

1. Reactive (or procedural) AI: responding to the intent of the customer with pre-defined actions
2. Artificial General Intelligence: determining intents and outcomes by analysing the actions of humans / agents when handling customer intents that reactive or generative AI did not capture
3. Generative AI: determining intents and outcomes by analysing relevant large data sets and suggesting or implementing the outcomes or actions

Almost every contact centre has been using some form of AI for decades, but we never called it that. We had IVRs, call flows, contact flows, and other names for the automation to do things like:

1. validate the identity of a caller (by PIN, post code, shoe size, etc.).
2. provide account balances, recent transactions, etc
3. request password resets

These things could be done by procedural AI using touch tones (DTMF) or voice, known as Natural Language Processing (NLP). Each pre-determined option had specific steps to gather data, interact with another system using API calls, and report on the success or failure of the requested action. For example, when called by my bank for an overdraft on checking, I could select Transfer, From Savings, and Amount. The IVR would call the API for the transfer and report back the result, such as Success or Not Enough in Savings.

As technology (and the media’s attention) advanced, they started calling this Artificial Intelligence, or the ability for technology to perform tasks that normally require human involvement. A rose by any other name, or something like that...

The next logical step in this evolution is Artificial General Intelligence, or AGI. AGI takes this time-honoured process to the next level by adapting to what the customers want the system to do, but it does not yet have that ability. For example, we can do balance checks and transfers, but not fraud alerts or credit increase requests. By recognising the need for this from agent results or unmatched voice requests, we can enhance the system to add these new functions so the customer can receive automatic help and the company can save money by not using more expensive humans where electrons can do the job.

The final type to discuss here is generative AI, where very powerful (and expensive) technology can analyse vast amounts of data and determine new capabilities BEFORE the customer asks for them. Beyond the cost, there are other equally daunting concerns, such as security and relevance. To gain the benefit of GenAI for your business, the data has to be relevant to your business, which often means using data that contains your customers' personal information. This increases the value of the insights but also runs the risk of unintended disclosure. To quote OpenAI, “We spent 6 months making GPT-4 safer and more aligned. GPT-4 is 82% less likely to respond to requests for disallowed content”. I read that there is an 18% chance that customer data will be exposed and put my company at risk.

The good news is that, well, there is good news. You can take advantage of the benefits of AI without becoming an expert or spending tons of money. We can reach great heights by climbing there ourselves (the aforementioned expensive effort or cost), or we can hitch a ride with a Titan like AWS. AWS has a variety of ways to do this, and the best way to get the honey without bee stings is to use Amazon Connect and all the AI-based CX services exposed through it. And if you want them as a finished product versus the ‘world’s best set of Legos’, you can use a partner like CloudWave to get it. Whether it is personalisation and automation of the customer experience in the channel of their choice or having AI ease the drop on communication and improve agent performance with real-time assistance, there are solutions you can use now. And for many of them, you can evolve with the packaged solutions for as long as you wish. If you decide to travel further down the rabbit hole and customise your AI yourself, you can migrate there at your own pace.