Is Artificial Intelligence The New Snakeoil?
A new look at how artificial intelligence can be used in the contact center and why it may make sense to start deploying it in a step-by-step approach.
>> All right. I'm gonna find giving this talk in English a little weird, even though I'm actually American by birth, but after living 32 years in Germany I've been in the Call Center World for At least 15 years, I've never given a speech in English here. So I'm gonna work really hard.
I'll try not to speak too quickly. My speaker before me was really rattling it out and I unfortunately missed the trip to Tel Aviv. That would have been a great one. I'm gonna talk about artificial intelligence, what a surprise. And is it something for your business? How do you get started with artificial intelligence?
Some of the experiences out of it. I'm an evangelist for Avaya. An evangelist is basically a storyteller. What I do is I collect stories of successful, sometimes not so successful, experiences with our clients and partners around the world, and I share those with the marketplace. So everything you're gonna see here usually has a story behind it.
There's one or two new things going to show that were just announced right before the Call Center World. And hopefully I'll make that before the flags in the back start telling me time's up here. When I talk about any kind of new technology with clients, seven years ago social media was the big rage.
Everybody wanted to hear about how social media was gonna take over all the communication channels. I always start the conversation or start the deliberation about it asking myself why, what is artificial intelligence supposed to do? Why do we need it? What's what's the purpose of it there? And if I look at the situation of contact centers today, it doesn't matter if it's in Germany or wherever you might be coming from here, we have trouble with high agent turnover and the cost and the time it takes and consumes to train and bring in new agents or supervisors into a business here.
Dissatisfied employees in contact centers, is sometimes a very frustrating business, a lot of repetitive work, nothing new. I have clients that we've been working with for over 15 years and they're doing still the same thing 15 years later. And you imagine as an employee in some of those businesses, it's a bit frustrating.
I'd like to have a little bit of variety. I'd like to have some of the tedious work taken out of my day job and let me work on the real problems or issues or challenges that my business needs for me to focus on here. And customers are everywhere, this is something we're seeing.
And you're probably experiencing customers are coming in left, right, and center. It's not just voice or email or chat anymore. It's all these other communication channels. And how do you make sense of that? How do you take advantage of the data that's available across one channel and make sure you're using and applying it to another?
So when I think or when I talk to clients and partners about artificial intelligence, what role is it? What is it gonna do? And of course, you read this in the press and anybody who's selling AI solution is gonna tell you, it's gonna make the experience of your customers that much better.
It's gonna be great, customer is gonna be happy, it's gonna satisfy all those dreams. Just like social media was supposed to solve it all seven years ago. Your agents are gonna be so happy, they're gonna be given new tools. They're gonna have a wonderful world to work in here.
Again, this is my snake oil peace. This is the attractive piece. The number one reason every client tells me why they're looking to apply artificial intelligence in communication is they wanna cut costs. They wanna save money, right? The appearance of AI is I'm going to automate parts of my business, automate part of my communication, and therefore I'm gonna take cost out of it.
Costs in a contact center business, as you know well, is the people, right? If you can reduce a bit of the people cost, you can save money. Ignoring the fact of what it might actually take to bring a really successful AI project into your business. Very recently, Forrester released a predictions guide for 2019, I just ran across it right in January and I thought there was a couple of things I just wanted to bring out.
Not the typical statistics, but 40% of the companies, they expect this year in 2019, will be adopting some sort of artificial intelligence for their customer support. 60% of chatbot deployments will have a very poor experience of handing it over to a live support or live agent. This is something that we as a Avaya take very seriously, that a chat bot can be standalone, can be a dead end chat bot there.
But if necessary with the right type of business, it needs to be integrated with the rest of your business processes. They expect 60% are gonna have these pretty poor experiences. The last one I thought was funny, sad. They expect or they predict a crowdsource attack on a corporate set chatbot somewhere in the world due to frustration.
People are mad. Some of the chatbot experiences that are out there today are less than stellar, right? They're not the best thing. We're not treating our customers or potential customers like they should be treated. And they expect some sort of crowdsource group coming together and whether sort of a DDoS sort of attack and taking down chatbox for a certain industry out there.
I'll be waiting till the end of the year, see if that actually happens and report next year. So artificial intelligence in the contact center. At Avaya we break it into four types of areas of where AI in the context center might play a role. One is IVR Hell, right?
How many of you, and I wanna ask for a show of hands, have the same IVR applications running for at least six or eight years? I know from our clients there's a lot. And if the customers contact you regularly, they're hearing the same menus every time, which might be comforting, but it could also be very frustrating there.
We see a more seamless and more conversational, I'm sure you've seen this across the stands at the show here, talking about conversation using natural language to communicate with a company here, not just following a structured menu of items. Smart routing, the acronyms FIFO and MIA, I'm sure you're very well aware of.
So the longest waiting customer in a queue should be the next one that comes out. The longest idle agent should be the one who gets the next piece of work. It's just fare that way. I like to give the example of, if I go into an expensive Italian shoe store, and I'm walking around for 10 minutes in the aisles there looking at very expensive shoes.
And then another person comes in 10 minutes after me, walks into the shoe store, and the proprietor of the shoe store looks at the two customers who are in queue, and I'm the longest one in queue, I'm obviously not spending a lot of money on shoes. The other one behind me who may have showed up 10 minutes later spends a lot of money on shoes, who's gonna get serviced first?
The idea of FIFO is a great technology, is something we brought as AT&T back in the 1980s, and everybody is using it today. It's not always the best match. So we look at making better decisions to route work and pair it together. Going back to the making your agents happier, making agents smarter, augmenting agents with knowledge bases, bringing smart information to the agent desktop so they don't have to search.
We have examples of smarts lists of PDFs as help guides popping up an agent's desktops based on the language that was used, the wordage that was used to search through and prioritize those. So the user, the agent doesn't have to go dive through directories and find a PDF and open it up to see if it's the right one.
So helping them to access information, gather that information more quickly. And the last one is taking advantage of all the information that's available in your business. The information, the KPIs, the insights. Just imagine if every conversation that was had, every voice call that was had in your business was transcribed into text, and you can analyse all that text in near real time and take out insights out of that information during the business day.
If you sell sweaters, And on Monday morning, five calls in the last 15 minutes have been coming in about red sweaters. And you know, that could cause a problem. Because you might not have enough backlog here. You might be able to flag agents and information in real time saying try to move them from red to others here.
Otherwise we're gonna have a problem and fulfillment here. This site types of insights can be made available using Artificial Intelligence to do the voice to text transcription, and then the analysis on that amount of data here, whether short term or long term. So we break it into the four areas and the types of applications that we have we use the same four areas of a description here talking about our virtual assistant who is AVA here.
I'll come back to that in a moment and different types of applications. Transcription real time in detections of intent, why is somebody contacted me? What does that text message say? Just because they opened up there, they reacted to the web chat that popped up on the website, they enter something in there and you send it to an agent.
Maybe it's better to send it to a specific agent, just like we do with voice calls. But most people aren't applying that type of intelligence on their texture channels here. All the way to call summarization, of course, quality monitoring and productivity. You can see how they play a role in different parts of the business here and how they can be interesting.
So how do you get started, and I'm gonna do it specifically here around chatbots because that's probably from an artificial intelligence, the lowest hanging fruit, easiest one to understand here. And I have a slide from a consultant, not being German, of course, this is an international crowd and maybe I can do it a bit differently.
But I've lived in Germany long enough to respect the Germans and respect their engineering prowess and their skills here but I had a consultant at an event in October, and he presented this is his slide. So I don't wanna point any fingers at Germany. Germans tend to over engineer everything.
I know that's probably not the case in your businesses and your countries here, but the Germans like to have everything 120% perfect before they even begin. So when you approach something like artificial intelligence and say just chatbots and you wanna have all your communications running across text through a chat bot at the beginning and the get go here, you'll never get the project off the ground.
It's just not going to work that way. So his advice is here is to take it slow, go small, right? Don't expect to be able to burn the fortress with the first rollout of your AI chatbot here is a business here. Be able to step back here and understand I need to start small, smart narrowly focused here because the chances of success will be that much greater here and then you can expand it over time.
The step by step approach that he talks about here is to pick five topics, right? Just find five scenarios, five typical things. It might not be the five most requested items in your contact center. It might be some niche areas but you have good information, in order for AI to work it has to have information has to have data, it has to know something, whether it's my knowledge bases, FAQs or other conversations to be able to feed into it, you have to be able to teach it from previous conversations here.
You may be able to find information quite easily if you find those five areas and you have no data to help feed it. Then take one of those items out and find another one. You have to have the information in order to begin training here, collect web chat information.
I had a Danish bank approached me or approached us a year ago and said, I have two years of web chat transcriptions that our customers and our agents have had. We've been collecting these here, can we feed it into the AI and have it ready to go and we said sure.
You can do that we can take all those conversations and feed it into AI and learn the insights and we can tag it, we can mark it and repair it. We have a running. But the question that we asked them is, are you sure that all of those conversations that were had between those agents and those clients are really pristine and good conversations?
Would you wanna teach your child all of those conversations that were had? Where some of them not so pleasant? Not so nice? And they said, well, we've never looked at it all. We don't know. And I said so sanitizing the data it's a big effort. So the idea of saving money, cost savings of rolling out AI is a little bit more trickier than most people assume.
And then it's the iterative part. You're gonna repeat this process over and over again. That's why, [INAUDIBLE] consulted from marketing consultant, his ideas, take a few examples of a few scenarios of how you wanna start doing AI here. Take more examples, take what he calls, role model interactions, right?
Things that are really good interact. That's the kind of conversation I wanna have my agents having with customers more often and keep repeating at that depth. Chat bots, this is something we work together with a number of our consultants around the world to understand. And it doesn't mean you have to go through this process but it's what we're seeing a number of customers are doing is where do you start and where might you get to?
You can think of a chatbot is gonna solve all my problems back to the snake oil. I'd love to sell you more chatbots here. You might start with very simple. It might just be on the web page, right? It might not go to a live agent. It's just a dead end.
It's just an informative FAQ bot. I was shown last year that Lidl in the UK has a chatbot that all it does is suggest wines, right? It's a very narrow use case, right? Not connected to any agent not connected to any store, you can order through the chatbot here, but you're asked some questions.
What are you preparing for dinner? Is it gonna be meat, it's gonna be poultry, is it vegetables, and make some suggestions of certain wines that they might like to put in front of you here. You can't even ask if that wine is available in your local Lidl or not there but.
The idea was a very narrow, very, very successful chatbot for what it does here with no live interaction added to it. The next stage might be to expand the number of cases maybe it's not just wines, maybe you add cheeses, right? Now you teach your chatbot to add cheeses here instead of trying to over engineer it and cover all the different scenarios in the beginning, start simple.
Living here in Europe and Germany here we have a lot a lot of multilingual challenges here. A lot of contact centers are challenged to have agents that can handle all the languages that are necessary here, even in Germany. Within our cities here, I worked as public administrations that are trying to roll out chatbots to do customer or citizen assistant sort of chatbots there and the citizens come in and don't type very well in German.
Very hard for a well taught German chatbot to understand gibberish typed in there and so they want to offer will type in Turkish type in Arabic type in Russian and polish it whenever you're in English you might want to do there. We can handle the translation the language very very easily.
So you might add multilingual support and maybe even a live seamless support to a live agent. And going through and then perhaps at one point you get to the point where you have proactive bots interactions going on? This is going to be a really interesting point there your proactive notifications, we're doing it already you're sending SMS to the customers, informing them something's happened to their order, or something's gonna be arriving input.
But usually when they get that SMS and I was talking to a large retail here in Germany yesterday, they said you might not know we send a lot of SMSs to our customers to update them on status that their situation, their orders and here it's a customer response to that SMS.
What happens with SMS? We never read any of them. We get millions of them. We never read them. If you send out the email that says do not reply at company ABC, that's a really frustrating one as well. But if these notifications, having that channel open that you can-
>> [INAUDIBLE] Starts to please feed. [INAUDIBLE]
In the conference wire.
>> At least they did it in English. So there is a guided tour in 15 minutes. So proactive notifications with a response channel for the customer to come back to you using the AI to be able to automate some of that conversation.
So you don't have to always have a live agent. So one of the new things we just announced right before the call center world is a social network for Bots. And every time I present this, whether in large groups or small over the last few weeks, I can't help but smiling and I see the smiles and people visit seems a little interesting or weird.
So imagine your chatbot for your company goes on the Facebook, and finds other chatbots. We just let that sink in for a moment, right? The idea is a successful chatbot is taught to be very good at one subject, wines, or cheeses, or bank accounts or order status here.
But we as humans, don't always only ask those questions. If I go to the little wine chatbot and I say what would be the right cheese to buy with that wine that I've chosen? That chatbot knows nothing about cheese, right? if I ask that same chatbot where is the closest leader?
It doesn't know anything about location and direction management here. It was taught to do a certain skill very well. Again a narrow use case around an AI and a chatbot or whatever type of scenario can be very, very successful. But we as humans want to ask other questions we don't always go the straight and narrow.
The idea is here chatbots can be brought into a very well controlled network of friends and befriend it, can be decided they want to friend other chatbots. And the scenario I have up here are three different organizations. The pharmaceutical company which produces medicines, logistics company which obviously brings the medicine to the hospital, to the end user here.
And the government is example is the one who approves the medicines. So the hospital who's not shown here, the hospital contacts the pharmaceutical company on a chatbot and says, when is the medication going to be available? I have patients waiting for this medication. The pharmaceutical company says, the chatbot would say, well, we've turned it in for approval at the government.
I'm a chatbot, I don't have the answer that information here. The logistics company will leave out at this point of the example here. Already here in this controlled environment of our social network for bots here platform, if the pharmaceutical company's bot had friended the governmental bot here, they would be able to ask the question to the government little bot.
Me as doctor at the hospital, I don't see any of this happening. I'm communicating with the pharmaceutical bot, it's asking this question. When am I gonna get the approval for this medication? If the information comes back and the trusted value is high enough, this is something that the user, the administrator, who said I want to put the pharmaceutical bot on the platform.
They make the decision who the friends are. This is not us, this is not just Facebook, such as has nothing to do with Facebook to be very clear. They make the decision, do I trust the answer that's coming back? Do I wanna give that information to the user, or not, to the requester?
If yes, give the information to the user, I got the information. It looks like we'll get approval for the medication in April. The next question for me as doctor might be well, when will I get it? So the question then if April, we have to ask the logistics bot, when we'll be able to pick it up and move it from the production to the hospital here.
The idea is be able to take very narrow bots that know their domain extremely well, put them together in a network of bots, let them communicate with one another under extremely strict guidelines. Where the administrator of each of these bots can decide which way you want to go here.
I mean and can unfriend bots at any point, can see the communications going on there in order to give a seamless communication to me the doctor in the example here. Of course, there's recommendations of bots here we're bringing this it's in beta right now which should be for general access next month in March.
And of course in the Avaya world, we see a seamless network of bot communications through live agents as being an absolute must in most of these cases. Because bots aren't perfect and we don't wanna frustrate our customers, and just leave them hanging in a conversation. Also with Avaya, and to be very clear Avaya, most of you will know obviously a contact center company and also Unified Communication company for 135 years.
I don't wanna go into our history here, but we've been very active in working with AI companies, companies that have Artificial Intelligence capabilities here. Because Avaya is not an AI company. We're not trying to be an AI company. We know there's others out there that do things very, very well.
There's companies here that do email management, email information extraction, all the way up to automatic email responses really, really well. Why would we wanna bring that into our network? So we started something last year called the A.I.Connect program here. A lot of logos of different companies there, you can go on to the web avaya.com/aiconnect, you can read more about it there.
This is expanding all the time as more and more companies come to us. Question that I often get when I meet with clients well. Andrew, you've got me excited they're so interesting here but, I can't make the decision for my company, right? These are bigger, bigger conversations that are happening.
This is not just I want to have a narrow chatbot in my business here that's gonna do one thing. I need it to be able to integrate it with other data and other departments that have it. How do I get my company on board? And I brought a couple of examples of what we've done just over the last year companies and partners have been using our technology to show to their own companies at times here.
The first one at the top is in Geneva was the hospital. The Geneva hospital organization, they organized a hackathon for over a weekend. They invited people to come in and design scenarios, technologies, communication. Wasn't just communications but this company, Choobs, our partner in Geneva used our platform to build a smart doctors and nurses application on a mobile device.
Right, over the weekend, they won first prize. Xtrasource is a business process outsourcing in the Netherlands our partner is Dimension Data there. There they organize an internal hackathon over three months to very large business. They have development capabilities there and over three months next to their day job, they had people in six teams I think it was that we're trying to figure out what would be the right type of new application do we need in our business.
These are the things that we don't often even know. They know who knows better, what you need in the business than often people within your own business. And then we organize ordinary applications or hackathons ourselves. That one was in Tel Aviv in the summer, last year where seven or eight companies were invited.
This had nothing to do with contact center. This was all UC, Unified Communications, smart email management, smart voicemail that using AI to drive that all the way up into sort of a Siri in your conference call applications out there. I organized an event here in Frankfurt. So I'm based here in Germany last October where I invited six companies to come in and show off their skills.
What do they do in different parts of artificial intelligence and communications. I'm doing another one which will probably be in June this year. So if you are in Germany, not sure this is the international forum where you might be coming from but we're putting on these organizations. I have one also happening in Switzerland and hopefully in the Netherlands later this year.
We are 135 year old company who have been doing telephony from the very beginning. And this is what most people know about Avaya. They know us as a telephone company, whether it is from AT&T, or whether it is from Northern Telecom, or from Germany here from Tenovus, and Bosch Telecom.
Our history or heritage is very much based in voice and telephony and connecting two lines together. But we've made a major shift in the way we approach communications as a business. And that's what I wanted to share with you today and show that we're applying things smart technologies, including AI in every part of our communication.
So not just in contact center, and we're looking for the best partners out there. Even in certain regions, they may be just a very local partner that we say, bring your own bot, bring your own AI. If you have somebody you're working with, you found a local organization that does something really well.
And you think this might be something you want to use, but you're not sure if it could work with your supplier platform, challenge your supplier. Bring it to them and say, can you work with them? And they should be able to do it because that's the way the communications are gonna be built today.
The way I tell this story is I talk about all these different suppliers, all the different players, whether it's email management systems or chatbots out there. These are all just different instruments that we're trying to bring together. And I use the example of a symphony, so each symphony is made of a lot of different instruments.
And if the instruments aren't brought together in a harmonious way, they don't make good music. And this is what Avaya does. We don't see ourselves as the whole symphony, but we are the conductor. We make the music, we bring it together, wherever those instruments might be coming from, and some of them obviously we bring ourselves.
But if there are specialties that are out there we bring them together in our symphony and we make this happen. In closing, I'd like to say that we're of course here, the ContactCenterWorld. I got the five minute notice, thank you. We're over here in the next hall, but we have four stories.
We're not showing product this year. We finally made a change, I'm just showing off our shiny products out here. But we've four different stories a customer journey story, very interesting customers moving between channels, voice into an email to a chat, to an outbound call, and being able to visualize that.
So that an agent can see where the customer has recently been in my business instead of just reacting and say, how can I help you? The Social Network for bots is being demonstrated over there wells and analytics reporting, and we have a service center of the future. What could that future desktop look like a little bit of a visionary look into things going forward.
We're over here directly on the other side. You might pass by and you can't miss the zebra looking as our colleagues like to call it, the new branding. If there's anything that was of interest that I shared, you can easily find me in the Internet. This talk has been recorded.
I'm doing the same talk tomorrow, but in German, it'll be up on my YouTube channel probably by Friday night. If there's somebody in your business that wasn't here today that needs to hear some of this, so you think it might be of interest, just search for me in the web somewhere you'll certainly find me.
This you won't miss anybody else, and I thank you for your attention for being here today.