Webinar Recap: Artificial Intelligence in Cloud-Based Solutions
With the advent of things like chatbots, artificial intelligence (AI), interactive voice response, and machine learning, novel technologies continue to disrupt the contact center industry. These advances often fuel the fear that automation will someday replace humans. To dissect the hype and explore the real opportunities around AI, I teamed up with Customer Contact Central to discuss AI in cloud-based solutions. Access the recorded webinar here or read below for a summary of how customer service centers should realistically think about AI, complete with five questions to ask vendors when evaluating AI solutions.
Reality vs. hype in artificial intelligence
So what exactly do we mean by artificial intelligence? There are many specialized fields within the broader category of AI, and we often see confusion around what each field actually covers. First and foremost, AI is a subset of computer science. It focuses on incorporating simulated human intelligence into machines. Underneath the umbrella of AI comes machine learning (ML), natural language processing (NLP), and deep learning (DL).
Machine learning refers to techniques that make machines learn from data and then use those learnings to provide value back to the end user. NLP involves making machines “understand” the meaning of natural language, including the intent of the words humans use to communicate with one another. Deep learning pertains to algorithms that are inspired by the structure of the human biological brain. DL has generated a lot of excitement recently because it’s the closest machine equivalent to simulating how the brain actually works.
So now that we’ve defined AI, where are we with this technology?
AI has been “The Next Big Thing” for a long time. Since the advent of computing, the ultimate goal has always been to create a technology sophisticated enough to act as a peer to humans. AI has had so many ups and downs that we refer to these trends as “seasons”. When things are going well, we call it an ‘AI spring.’ And when things are not going so well, it’s an ‘AI winter.’ Right now, we’re in an AI spring.
Coming off the heels of the last huge technology shift – cloud computing – the data and processing power needed to make AI work is now more accessible and affordable than ever. It used to fall to companies to build their own environments for storing the vast amounts of data and the computing power necessary to facilitate AI, but the advances in cloud computing have enabled AI to be operated more easily. Now, we see the big players in cloud computing – Amazon, Google, Microsoft – all providing not only the actual processing power and data of cloud computing, but AI services as well. Companies can now leverage and harness these technologies to pull together AI-driven solutions.
Improved enterprise user experience (UX) and easy-to-use interfaces have also exponentially increased the growth of AI. The easier a software is to use, the more data it’ll generate. And the more data that can be leveraged to train AI, the better the solution. Enterprise software has recently gone through a consumerization: the software we use at work is becoming just as enjoyable as the software we use in our personal lives. The idea of UX being a core focus in enterprise software has helped push this technology forward and generate a lot of excitement.
With excitement comes hype
Given all the excitement around AI, it’s important to level set as to what exactly is possible with this technology. Gartner releases what they call a “Hype Cycle” that plots tech trends to show which emerging technologies are most hyped. Hype grows as solutions scale the Hype Cycle, then peaks at Peak of Inflated Expectations, and then dips again when tech inevitably loses hype and enters the Trough of Disillusionment. The end goal is to transcend the curve and climb the Slope of Enlightenment into the Plateau of Productivity.
Some technologies fall off the Cycle and never make it up the final curve, but many do. The example above is the latest iteration of the Hype Cycle, and has much-hyped deep learning at the top. In 2009, cloud computing was at the top of the Cycle. We saw the same behavior happening back then as we do now, so it’s interesting to contrast that with today’s most-hyped technologies.
When a technology is overly hyped, we see all sorts of crazy articles written and ominous movies made about it. Between shows like Westworld and articles that claim that AI will soon be writing better novels than humans, the way AI is portrayed in media and pop culture is often confusing, and plays into the fears of what could go wrong if it goes off the rails. There is a very real and prevalent fear that AI poses a threat to humanity.
A common manifestation of this fear is customer services teams wondering whether they’re going to lose their jobs to machines. All the talk of AI replacing humans and automating away processes misses the real opportunity of how AI can be transformative.
The real opportunity for AI in contact centers
A lot of technology adopted in the workplace has traditionally been applied as ways to save money. As businesses, we make cases to buy technology based on potential cost savings. But in the case of AI, a lot of these cases are being made based on potential revenue increases instead. Business are not asking how AI can save them money, but how AI can make them money. How can AI help customer service agents convert more customers from free plans to paid plans? How can AI help customers understand products better so that they’ll renew?
It’s a really interesting reframing, this shift from cost savings to revenue generation. That transitions nicely into the customer service field making a similar shift from cost center to a revenue center.
Customer Service is transforming from a cost center to a revenue center
Customer service teams have the most enduring relationships with customers, long after sales has closed the deal and moved on, so how can we help them have better conversations and better relationships with those customers? Despite those close relationships, a disproportionate amount of the AI conversation applies to solutions that remove the customer service agent from directly talking to customers rather than bringing them closer together.
4 types of AI for CX
Deflection refers to intercepting customers who are reaching out with simple, repeat questions and answering them before they even have to ask them. The tech deflects an interaction with a support agent from actually happening. This is a cost savings approach, not a revenue generator.
Bots simulate human customer service experiences. However, being cognizant of the customer experience, I think the best bot designs make it clear that it’s a bot on the other end of the line. Good bots don’t try to simulate humans, they augment wait cues and provide value.
Processing or workflow AI tends to be agent-facing. These solutions speak to identifying and alleviating common pain points. As a human, it’s hard to pinpoint where customers tend to get stuck because it involves indexing all tickets, categorize them, identifying topics, trends, and sentiments. Machines are better suited than humans to bucketing and analyzing, so that’s where processing AI usually comes into play.
Coaching AI is also agent-facing rather than end customer-facing. This type of AI aims to help and empower humans to be better at their jobs. It aims to help agents have better conversations with customers so they can spend more time creating white-glove experiences rather than digging around for the answers to questions. Coaching is the way we think about and deliver AI at Guru. Empowering humans is a great way to create long-term value for customer service teams. This AI tech is 100% focused on helping someone be better rather than automating them away.
Top 5 questions to ask your AI vendors
When considering a new AI solution, it’s important to ensure that the initiative you’re thinking about is aligned to best set you up for success. Here are five considerations to keep in mind along with questions you can ask vendors during the evaluation phase.
1. What metrics should we expect your solution to improve?
Beware the “Jack of All Trades.” A mistake that some AI systems make is trying to do too much. Today’s AI systems only have the capacity to do so much, which makes it super important that they be super focused on solving specific problems. The training data an AI system uses to make its suggestions is directly correlated to its success. If you’re trying to solve three or four business problems with one AI system and one set of training data, you should expect mediocre results.
The question to ask to get at the heart of this problem is “What metrics should we expect your solution to improve?” You need to suss out the ultimate outcome and how it will be tied back to the metrics you use to measure performance. You want a specific answer here; be wary of any solution that claims to solve seven or eight things at once. If a solution specifically focuses on a particular outcome, that gives you a great likelihood of success. Invest in AI products that focus on solving clear problems with access to valuable data to train from.
2. What will our customers experience?
Empower your agents and your customers. Whatever AI system you’re contemplating, be very focused on the end customer experience. Forrester has a report that talks about the risks companies face by too aggressively driving customer traffic (chat, phone calls) to AI systems rather than to humans in a race to save money. In doing so too aggressively, companies take a hit in customer satisfaction. You want AI to help you save money and drive revenue, but you certainly don’t want to that at the expense of customer satisfaction.
By asking “What will our customers experience?”, you can determine whether a solution is in line with how you think about providing a great customer experience. What your end customer is going to see when interacting with any system should be your primary concern.
3. How does your AI solution learn and improve over time?
Watch out for the “secret sauce”-ers. Transparency is important. Vendors should be clear and direct about what data they gather and why. AI systems are built off data that you will be feeding to it, so it’s super important that any AI provider tell you exactly what data it will use to train itself, how that data is stored, and how long that data is stored.
By asking “How does your AI solution learn and improve over time?” you’ll get a clue into the data sets your AI provider will need from you in order to do its thing.
4. How will we keep our knowledge up-to-date and accurate?
AI without up-to-date knowledge will fail in contact centers. This is related to the jack of all trades concept. When you think about the knowledge that’s in your environment, it’s the encapsulation of the know-how of your subject matter experts, of your products, of your systems and processes, and how all of those things works together. Any AI that is leveraging that know-how needs to have a way to assure you that that knowledge will stay accurate and up-to-date.
There’s a concept in AI called the closed loop. Over time, the knowledge and things training your AI systems will change because your products change; and the technology your products are dependent on will change; and new competitors will come to market and you’ll have to adapt to them; and as team your grows, the way you do support will change. With all that inevitable change, what you don’t want is an AI system that doesn’t have a good closed loop of evolving its learning. You’ll see examples of this when AI systems start returning degraded information over time. When the system reduces the quality of output, that’s a leading indicator that it’s not learning and evolving with your organization.
The problem is that you may not see this until a few months in when the knowledge starts degrading. So a great question to ask upfront is “How will we keep our knowledge up-to-date and accurate?”
5. How will your solution make our agents better at their jobs?
AI should empower people, not replace them. Be sure to ask “How will your solution make our agents better at their jobs?” to find out what the immediate impacts of that AI solution will be on your company. Over time, there will be profound opportunities to automate tasks, but for now, it’s important to get an answer to this question that doesn’t sound like lip service. Terms like “automation” and “virtual agent” tend to indicate AI solutions with less near-term, practical applications.
Because again, it’s still relatively early days. AI is profound in long term capability and impact, but it’s still a long way off from understanding things like empathy. If you put an AI system in front of your customers directly when they’re upset, a machine isn’t going to improve the situation. These are the types of questions that make sure you’re thinking about the outcome of the product in the best possible way.
Like cloud computing before it, AI is transformational not just for enterprises, but for all humans. While the hype is huge, and many people misrepresent its capabilities, there are real gains to be achieved today if you’re focused on the right outcomes. Instead of thinking about AI as “automating us away,” and ultimately creating this superior class of machines, what if instead we talked about AI helping us grow? AI helping us improve as humans, both personally and professionally? That’s the mindshift we need to make that will be really exciting as to what’s possible this technology.
For more information about using AI to empower humans in your contact center and throughout your entire organization (and Guru’s answers to these five questions), contact email@example.com.