Until recently, automated processing systems have had a hostile relationship with the typical user. Recently though the trend has turned and many users are quite happy to interact with an entirely artificial servant.
So what has changed? Have we now welcomed our robot overlords with outspread arms? It’s probably best we set the scene first and determine what we’re talking about.
Artificial Intelligence, or “AI” for short, is a truly immense field of research that’s starting to take a decent foothold in our commercial and personal lives. There are many ambitious projects for self-driving cars and almost self-aware bots that pretend to be humans. There are threats of entire industries losing their jobs to automation, with not even lawyers and chess grandmasters safe from replacement by a more cost-effective and longer working transistorised counterpart.
While theoretically anything can be performed by AI, the area we’re interested in is customer service. More specifically, how businesses deal with general phone enquiries without the need for a human customer service operator.
While the capacity to automate customer services has existed for some time, the acceleration in successful deployment of services, and their adoption by the general public, is due to the significant advances in speech processing and text-to-speech. This has enabled systems to understand spoken commands accurately without training and then relay results in a more natural sounding voice. Previously, automated speech recognition systems would either require lengthy training using known sentences to understand the accent of the user, or use very narrow sets of commands to limit the possibilities of what could be said.
While the capacity to automate customer services has existed for some time, the acceleration in successful deployment of services, and their adoption by the general public, is due to the significant advances in speech processing and text-to-speech.
Recent advances in technology and their deployment in services such as Siri and Alexa have meant that algorithms can be tuned and accuracy iteratively refined, while all the time placing automation in the hands of almost every westernised user. And while the technology has kept improving, our ever-increasing exposure to AI-powered devices on a day-to-day level has made us more knowledgeable and comfortable in interacting with it. In essence, not only have AI systems got better at understanding us but we’ve got better at, and more accustomed to, interacting with them.
The potential of AI in the SME business is huge and compelling. The UK is aiming to be a leader in this field, with some big players making some bold claims, such as 85% of all customer care being handled by AI before the year 2020.The potential of AI in the SME business is huge and compelling. The UK is aiming to be a leader in this field, with some big players making some bold claims, such as 85% of all customer care being handled by AI before the year 2020. Click To Tweet
Would your business benefit from AI?
Before you go laying off your entire contact centre team it’s worth considering first whether AI is right for your business and fully understand what you’ll be losing out on.
Anyone with access to Alexa, Siri, Google Home or any of the other plethora of accessible AIs will be well aware of their limitations. They’re great at adding six bottles of sparkling water to your next delivery, but awful at recommending a decent bottle of Italian white wine that goes well with fish. The same applies to customer service automation – AI is great for requests that can be easily understood and automates processes that it can obtain all the parameters for with high confidence and pass to well defined interfaces to existing systems.Anyone with access to Alexa, Siri or Google Home will be aware of their limitations. They're great at adding a bottle of tonic water to your next delivery, but awful at recommending a decent bottle of Italian white to go with your fish. Click To Tweet
The good news is that within the average contact centre there’s excellent scope for automation. Most contact centre teams already operate to rigid parameters that are largely scripted. For example, your after-sales team may be authorised to only provide refunds on purchases that fit a particular set of criteria, such as:
- Transacted in the last 14 days.
- Paid by credit card.
- Product is faulty.
- Original packaging retained.
Anything which falls outside these parameters may be handled by a manager. Effectively you don’t give any powers to your regular customer services team that aren’t already clearly bound by rules that couldn’t be automated.
In analysis you may also find that a large percentage of calls coming into your contact centre are requests for information that could easily be relayed back to the caller, such as pricing information or opening times.
Lastly, booking systems are wide open for automation, as the information being collected is easily obtained and known in advance, with the reservation system typically being online and accessible by Application Program Interface (API).
Stop annoying your customers
Another area AI excels in is reducing the amount of menus needed to navigate a caller through your company. Some larger operations require three or even four levels of Interactive Voice Response menu before getting a caller through to the department they need to talk to. The upshot is a very annoyed and frustrated customer, especially if they are then expected to wait to speak to an agent. If they’re calling to potentially close their account then you’ve lost the battle before the first word is spoken.
The upshot is a very annoyed and frustrated customer, especially if they are then expected to wait to speak to an agent. If they’re calling to potentially close their account then you’ve lost the battle before the first word is spoken.
An automated voice recognition system that works alongside your existing IVR menu means that customers prepared to speak a request to a well-designed AI system can get through to the department they need with a single request. And if the AI engine can be trained for your specific environment then a wide variety of requests specific to your business can be directed efficiently. For example, a request such as “I have a problem with a vacuum cleaner I bought last week” can be transferred straight to the small home appliances after-sales team, whereas “I’m interested in stocking your products in our department stores” can be sent straight through to the key reseller accounts team.
How AI systems get better at understanding voice instructions
The key to a successful speech recognition system is “deep learning”. Unlike beating a chess grandmaster or navigating the streets of San Fransisco, this is a little more down to earth and merely involves cutting through colloquial terms and understanding the root context of a request. For example, “I would like to speak to sales” and “I want to buy something” would both trigger a call transfer to the sales team. This is despite both requests having very little in common in terms of the actual words used.
Deep learning uses a mixture of initial training, grammatical substitution and reaffirmation to identity requests. Over time, this allows AI-powered call management systems to build a better understanding of other ways to request the same intent. This is achieved by asking callers to speak a request differently and adapt its algorithms on the fly.
Once a request or “intent” has been identified, a more natural process of turn-taking can process the request to obtain any missing information. A deep-learning AI is capable of extracting provided information from the requests to avoid asking for the same information again. For example, “I want to book three seats to Les Miserables on Saturday night” contains a lot of information which the AI needs to allocate to specific “slots” and avoid asking for, whereas “I want to book tickets for a show” will require the same AI intent to ask for the specific show, day and number of seats. Effectively, a deep learning AI adapts to what is being said and can change the way it interacts with a caller to gather additional information.
Avoiding the AI pitfalls
AI is a great tool. It helps you to be available 100% of the time, while reducing the demand on customer care teams. This gives customers access to services you may not economically be able to provide any other way. It also makes you appear at the cutting edge of your industry and shows you can adapt to change. In short, AI can provide a better service to your customers while reducing your operational costs.
So what’s not to like?
Well, done wrongly, AI can appear intrusive, cumbersome for your customers to use and cause you to lose touch with your customers. Any implementation of AI in your business needs to be carefully considered and done right. Here are the key considerations:
1. Don’t let automation get in the way
Too many automated systems appear as barriers to talking to a human. And often, the route to speaking to a live customer services team has deliberate traps that route callers into an automated system. The real path to an operator is often hidden and obfuscated with callers going round in circles listening to their balance over and over again or looping through countless levels of IVR menu hunting for that illusive option that causes a human to answer. This just causes frustration and makes for poor customer experience.
To work effectively, AI should sit as a tool and offer automation as an option for callers. In many cases, callers will take the automated option as it offers immediate service without queueing. A well-designed system will offer callers the choice to use IVR to navigate a menu and speak to a customer care agent, or offer a spoken request. Callers can then use whatever approach they feel more comfortable with.
Xewave Cloud PBX has the ability to provide AI alongside menus which either route calls through an organisation or link into other back-office systems to automate workflows. This means callers can cut through layers of menus and get right to the service they need with a single spoken request. Alternatively AI can be injected into a call queue so that callers who know they have an expected three minute wait can try out an AI without losing their place in the queue, either completing their initial request and leaving the queue, or obtaining information on a new product or service while they wait to discuss a different matter entirely.
However you do it, make AI a choice and if someone wants to speak to a human, let them.
2. Pick the lowest hanging fruit
Before you automate, understand the nature of all your enquiries and automate the processes that fall into the “easy to implement” and “frequently requested” box. You don’t have to automate 100% of your business to make significant reductions in the amount of enquiries your team has to deal with. Look for the widely requested enquiries and speak to your existing systems vendors to identify what access there is into your online systems. In many cases, enquiries can be handled with simple prompts. For example, if callers often want to know your opening times, then a simple prompt could be played which satisfies their enquiry.You don't have to automate 100% of your business to make significant reductions in the amount of enquiries your team has to deal with. Click To Tweet
Approaching us with a list of types of enquiry and what systems vendors you use allows us to design a future-proof AI service that your customers will enjoy using and provides dramatic reductions in contact centre traffic.
3. Don’t erode contact with your customers completely
If you automate too much contact with your customers you lose the ability to understand how they feel about your service and lose the ability to up-sell services to customers. While some element of automation in sales is good – e.g. pro-actively offering better deals to customers can significantly reduce churn – many service sectors require a personal touch to up-sell services, and losing contact with a customer makes those opportunities less frequent.
If you automate too much contact with your customers you lose the ability to understand how they feel about your service and lose the ability to up-sell services to customers.
But aren’t Artificial Intelligence platforms expensive?
They absolutely don’t have to be, and compared to their human counterparts they are extremely cost effective. Some service sectors value each call into their call centre as costing them more than £50 to handle when accounting for all operations costs. 1000 AI requests through Amazon Lex costs less than £5.
Xewave integrates with Amazon Lex, Polly and Lambda to provide turn-key AI processing either to route calls throughout your organisation or to link in to back-office systems to provide true automation of customer care activities. Our default integration with Lex automatically builds a speech processing “bot” with all the capabilities to route calls throughout your office that you can easily adapt to handle the nuances of your business.
Need more advanced AI functionality? We can handle the entire development process for you as an end-to-end project, or work with your teams to establish connectivity with other systems in your business.