A year ago, I was talking to my dad, who is a professor, and he said that he studied and worked with artificial intelligence (AI) back in the 1970s and 80s; it’s not as new a concept as most people think. But at that time, it was not considered practical or even interesting to most. Nobody cared about it.
In fact, the first AI program was presented at the Dartmouth Summer Research Project on Artificial Intelligence (DSRPAI), hosted by John McCarthy and Marvin Minsky in 1956.
By 1970, Minsky told Life magazine, “from three to eight years we will have a machine with the general intelligence of an average human being.” But the computational power would not be available to realize this vision for many years.
Even five years ago, AI was mostly rules-based. As long as a system behaved within the rules, it was great. The skill really lay in creating truly dynamic rules.
Gradually since then, many IT leaders began to experiment and implement AI. It was probably because many businesses moved from on-premises to the cloud.
The contact center clearly needed greater efficiency which made the cloud attractive. So, the IT consumption model changed abruptly.
Suddenly, a whole slew of SaaS (software-as-a-service) and cloud-based companies became relevant since the IT buyer was no longer restricted to one single network within a premise. That really opened up a new world to many businesses that could tap into these cool, creative, innovative solutions.
Back in the 2015-2016 timeframe, when contact center-as-a-service (CCaaS) providers were migrating a lot of contact center customers to the cloud, there was no AI present yet. Predictive routing was the only use case that we all thought would be feasible at that time.
Now there are so many different AI use cases that providers are now considering that revolve around improving the contact center workforce.
How COVID Kickstarted AI
The COVID-19 pandemic accelerated the movement of applications to the cloud and revved up the incorporation of AI in the software. It forced a lot of these businesses into the cloud faster than they would have migrated on their own. And likewise for developers to utilize AI in their products: like workforce engagement.
During the pandemic, organizations became strapped for human resources. Maintaining a workforce was difficult as contact center workers went home, and some made new life and work choices.
The contact center industry was growing exponentially but desperate for enough agents to function. These factors combined to make operators consider how to make their workforces more efficient.
To drive efficiency, operators had to break activities down to a set of tasks performed by their agents. They then determined which tasks are easier to automate: those ones that return the best bang for the buck to show instant results.
Now there are so many different AI use cases...that revolve around improving the contact center workforce.
Leading companies began offering rules-based tools like topic detection, followed by sentiment analysis, disposition prediction, and theme detection. They became bolder and tackled harder problems like summarizing calls or creating cases in Salesforce or automatically generating action items that the agent should take after the call.
As these companies became more experienced and more technologically advanced, they solved more problems with a more holistic approach to AI - without the constraints of rules - with impact on agent efficiency, automation, and business outcomes. What was really exciting: it was all real-time.
At the same time, the shift to remote work created its own issues. Some agents didn’t have a way to work from home. Some may have worked in a closet, while dogs were barking, or kids were screaming outside the door.
The need for better workforce management (WFM), specifically agent care, increased dramatically. How do you give them the right resources when they can’t tap someone’s shoulder and ask a question, especially if they’re new and inexperienced?
Contact centers began adopting tools that provide a really good forecast of the strengths and weaknesses of the workforce. With an accurate prediction of how many calls your contact center will likely receive next week, you can staff accordingly.
If you tend to understaff, you could know the workload that is coming and can adjust for the anticipated 50% increase in calls per day. This saves everyone from a strained, stressed workforce.
After getting the right forecasting and scheduling tools, other tools arrived that detected when improvements are needed. Like when agents are demonstrating a lack of knowledge around a product or not handling customer objections around cancellation effectively. If you know exactly where and why they’re struggling, you can properly coach them.
AI Ascent at the Agent Level
Any change in agent experience is carefully looked at by contact centers. For good reason: because it would impact hundreds of thousands of people who need to be retrained.
Top contact center vendors were not just worried about the customer experience (CX), but also understood the importance of the agent experience and the need to make their jobs easier. So, they created AI-based tools that monitor agent-customer conversations in real-time and give agents guidance on next steps.
But imagine if the AI started bombarding them with myriad bits of data, creating information overload and critiquing everything they do? The agent’s focus would then shift from showing empathy to customers, negatively impacting the natural flow resulting in a stilted, scripted dialog.
So, vendors created tools that used AI to take several monotonous things such as writing call summaries off agents’ plates, enabling them to have genuine conversations with their customers and more effectively helping to solve their problems.
A good workforce engagement tool not only does a good job of forecasting and scheduling, but it also identifies areas where you can coach your agents and help them improve around your products and services.
Speech analytics review their language skills and quality monitoring enables the coaching of agents. AI-based scoring helps so you don’t have to listen to a small sample of five out of 5,000 calls. You can listen thoroughly to every single call with the help of AI and come up with those coaching moments.
It will take just a few big brands that have AI managing and populating their workforce to build trust and confidence.
Vendors are now using Generative AI to allow contact centers to identify these coaching moments faster, pushing training into an evaluation room for real-time learning. It’s also changing the landscape of the entire product suite where each product is getting better and more efficient.
Imagine an evaluator asking AI to quickly fill in the form before validating the call. Or a speech scientist trying to create a topic and all phrases are already there as recommendations.
So vendors are not only evaluating each call, or calculating NPS scores or detecting coaching moments in the products, but also augmenting their products to be AI-enriched.
AI’s Greatest Opportunity
We believe we can improve every aspect of a customer journey from customer experience to employee experience.
- Can I have my people handle this call more efficiently?
- Can I train and coach them so that they’re better at their jobs?
- Can I do certain tasks for them, such as create a case for them automatically?
- Can I create a set of action items, so they don’t have to go in and do it themselves?
Nothing is impossible now. We are literally looking at everything that an agent can do. If a machine is listening to and learning from agents all year, it becomes smarter on things like:
- How do I handle objections?
- How do I behave?
- How do I show empathy?
- How do I show emotions?
After one year of training, AI will do most of the jobs and gradually shorten the training duration to the point where it will perform better than a human.
In the short term, we will see more focus on task automation. There will be a transition to a hybrid model dominated by human-assisted machines or machine-assisted humans. It will be a journey over the next 10 to 20 years.
Gartner predicts that by 2027, chatbots will become the primary customer service channel for roughly a quarter of organizations.
There are already virtual agents that pass the Turing Test (SEE BOX). But as business leaders, we cannot cede total control to machines. There must always be human oversight even if AI logs many of the tasks humans handle today.
It will take just a few big brands that have AI managing and populating their workforce to build trust and confidence. It’s like cloud migration. At the start, no one wanted to put their customer data in the cloud, then Salesforce happened, and others followed suit.
Has AI Passed the Turing Test?
Has AI in contact center applications finally passed the famous Turing Test where machines showed the same intelligent behavior as people? It appears that way.
Predictive routing was the first revelation of AI. Suddenly, for a large telecom carrier, suppliers like us could improve their call handling time by 2% or more.
Imagine a 20,000 or 40,000 agent contact center improving their individual call times that much: it equates to millions of dollars in savings.
Then suppliers expanded that AI concept to different use cases such as topic detection and sentiment analysis.
When I first saw a summary of a call produced by a machine, it just blew my mind! A machine took a 10-minute conversation and wrote such a crisp summary in three to five sentences.
I listened to 10 different calls from the five to 10-minute range. I checked the summary and listened to the conversation across multiple customers to validate it. Right away I could see the machine could be trusted.
I went to my customers and showed them, call by call, row by row, how much more effective and efficient the machines were than the humans. They realized the business value immediately.
It takes an average of around two minutes for a human to write call summaries. One agent typically handles 40 calls a day. This saves 80 minutes of time per agent per day, not to mention achieving greater accuracy.
How Does It End?
No, this evolution does not end in Skynet conquering the human race. But I believe something amazing will happen.
AI will understand and predict customer journeys. Currently, all the AI work that is going on in the market is reactive, reviewing and automating tasks to make agents more efficient. That’s the problem we’re solving for the next 12 to 24 months.
It will get really interesting when the AI understands and anticipates a customer’s behavior and course-corrects them from their current trajectory to drive a better outcome for the business.
This is the evolution to a predictive AI. It will proactively influence the customer via a chat session or by launching a new marketing campaign based on a better assessment of customer preferences.
This predictive, hyper-personalized AI may send a text or write an article based on how well it understands customer behavior, moving you along in your customer journey.
This development will happen in the next three to five years and the possibilities are endless. Influencing the customer to change their course will take an anthropomorphic entity that can transform into a sales agent, a marketing expert, a product information specialist, perhaps even a psychologist.
This semi-sentient AI will start to act as a business unto itself. It will automatically take care of certain tasks, even potentially entire business functions that might take five humans in different departments across the enterprise to accomplish today.
Eventually, this shift will impact the labor market, shaping the types of roles and skills needed, but humans will always have a place in AI-powered experience. It will be a fascinating future, one in which we harness as yet unimagined capabilities through AI. With humans in charge.