A Luddite In Conversation With Dr Iain McKay, Self-Service Technology R&D Manager, Graham Technology.

You know about technology, right?
Well, you could say that. It's what I do for a living!

Can you settle a pub argument?
Go on, then.

In this day and age, why do aeroplanes need pilots?
Definitely a pub argument, by the sounds of it. Well, no computer has yet been able to replace the brain in terms of ability to nimbly respond to the vast array of tasks placed upon it, especially when making qualitative judgements and seeing the bigger picture. Autopilot systems are commonplace, automated take-off and landing is rare but possible, but handling an emergency situation is as yet not something that a microchip can handle.

Flying a modern plane is much easier than flying a bi-plane, though, isn't it?
What makes the relatively mundane task of flying an aeroplane easier for today's pilot is information and automated systems augmenting reality. They take the workload off pilots. Their jobs are now orchestrating other, complex systems that are performing their own task in their own right. Now, I have a question for you.

Go on, then.
What has this to do with contact centres?

You tell me.
OK, then. I like a challenge. More and more channels are opening up between enterprises and their customers. There are channels such as natural language IVR, text-chat, email and SMS, as well as the traditional voice call. These channels are increasingly capable of being automated. They can take over the Customer Service Representative's role for common, mundane customer contacts, such as identification and verification, order tracking, simple transactions, and the like. Therefore, the CSR's job is becoming similar to that of a modern pilot: using automated systems to make their job easier, augmenting their own abilities, handling more tasks at once, having a greater overall view, and fundamentally using the human brain where the human brain is best utilised.

So this is the beginning of the end for the CSR, then?
Not really, no. The same argument applies as for pilots. This isn't the death knell for the role of human CSRs. The human agents are simply moving up the value chain. They are still required in the loop when it comes to conflict resolution, exception handling, negotiation, sales and complicated queries.

So if the CSR role is interventionist, how is this managed?
When a human CSR is needed, it must be a seamless transition with minimal delay, so that the customer barely notices. The human agent is like a safety valve: a safe pair of eyes and ears overseeing the virtual agents' progress, and clicking 'OK' as the final confirmation, for example. The system can present virtual agents to customers in the first instance, and then automatically spot a struggling user (such as an automated agent not understanding a customer's answer a couple of times) and hand this session over to the next available human agent.

At which point the customer has to repeat the entire process?!
Not at all. The human agent picks up the process where it left off.

Yes, but then they will have to explain to the CSR the story so far!
Wrong again, ace. The customer should never have to repeat themselves. Customers hate repeating themselves almost as much as they hate waiting. The human agent is presented with a transcript of the dialogue already undertaken, and can then complete the interaction with the customer without the customer necessarily even realising that a human agent has taken over from a virtual colleague.

Apart from the voice being different.
Actually, the virtual agent voices are very good nowadays. They can even have regional accents.

Blimey. Sounds like Sci-Fi to me.
Wrong. It's possible today. Several organisations, both academic and commercial are working on virtual agents being able to enter into a dialogue with customers and executing the same business processes visible to the human agent on the Contact Centre desktop.

So, the human agent can sit with their feet up watching the bots do all the hard work on their behalf?
Yes and no. Virtual and human agents do not need to be viewed as separate entities where customer interaction is bandied between them. They work well as a team when blended in a single customer interaction process, such that there is a virtual agent running alongside its human counterpart, the virtual agent fetching information and making suggestions as to what act to carry out and what information or prompt to convey to the customer. The human agent can ignore the virtual agent and enter a full dialogue with the customer, slightly modify the action suggested by the virtual agent, or sit back and confirm its actions.

So you need a virtual agent AND a human one? Some saving.
Don't be dim. With human agents performing fewer motor actions and less of the minutiae, like typing less in text-chat with customers, it frees them up to look for sales opportunities and, indeed, to multiplex more than one customer at a time, making for a more efficient use of human resource.

So how does it all work?
Natural Language and Dialogue Modelling modules store and execute business processes for customer interaction and deliver them across various channels; namely, the contact centre desktop, web, IVR, SMS, chat, digital TV, PDA and mobile phone.

Sounds like good stuff. What are you guys working on now?
The longer-term vision is to provide automated, customer-facing services which are indistinguishable from their human agent-based counterparts, thus allowing the relatively expensive human agents to do what they do best. There will be greater use of Artificial Intelligence techniques in order that the system can learn user behaviour per customer, and across all customers.

Indistinguishable?! Again, this all sounds like Sci-Fi.
Not at all. Sci-Fi, if you cast your mind back, frequently triggers Sci-Fact. To help make virtual agents come to life with a sense of human ability, intellect, anthropomorphic appearance, responsiveness and awareness, computer scientists and academics from fields such as computational linguistics, natural language processing, artificial intelligence and 3D Graphics are racing to fulfil the promises of Sci-Fi writers.

So when will I be able to talk to a virtual agent without pressing numbers on a key-pad or barking one-word commends down the phone?
Now. Software is already able to understand everyday language to varying degrees, allowing for a natural, human-like dialogue between consumers and computers. Such software can not only recognise words and phrases, it recognises nuggets of information which can help it determine in advance what the customer is going to want to do next.

So I am going to be repeating myself to a machine rather than a person. Is that supposed to be LESS annoying?
That is covered as well. 'Next generation' software stores pieces of knowledge pertaining to the customer's overall goals, individual tasks and parameters given. It stores information for use later: so no more repeating your name or postcode six times.

It also becomes far more re-entrant in allowing the user to complete a dynamically constructed process, and nimbly move around the process. Customers are presented with many more open-ended "How can I help you?" prompts, and the business process is thus aware of the context of individual sub-tasks.

As a customer, it sounds great. What does the business get out of it?
Organisational costs are minimised. As, indeed, are costs of training and maintaining a human resource. Indeed, contact centre processes can now be blended, such that self-service telephony and text-based interfaces seamlessly blend with human agents into a unified service capability.

Great. And no more queues, right?
Right. The virtual agent is a computer, and so can handle a vast number of simultaneous customer interactions.

Wait, though. Nowadays, I type more than I speak. What about instant messaging – or even texting.
Natural Language is Natural Language—stop me if I get too technical—be that spoken or written. We have developed an instant messenger interface, and texting is the same idea. Similarly, these sessions can be handed over to a human agent who receives a transcript of the contact to date. In such cases, it is even more likely that the customer needn't be made aware of a hand-off, as there is no change in voice to deal with.

Looks like we've thought of everything, then. Research teams can disband and go home.
Not a chance. Things move so fast that we will always be playing catch-up with the latest consumer gadgets and telecommunication methods. Just as new plane designs emerge, so will new customer interaction methods.