Agentic AI: let's reinvent, not recycle
Agentic AI describes a form of AI in which autonomous bots take instructions from humans or other bots, then perform a variety of tasks, in particular by accessing different computer systems, to solve a higher level problem.
It sounds a lot like a contact centre! But therein lies a risky illusion.
It is human nature to anthropomorphize. The tendency to detect human traits, emotions or intentions in non-human bodies is a side-effect of humans’ evolutionary advantage in complex social interactions. But one effect of this tendency is that whenever a new AI technology is announced, our first instinct is to ask, “how can we get this to carry out a human task.” In the customer service world, that means getting AI bots to take on the roles of Customer Service Representatives.
To apply AI agents like this in customer service would, I believe, be needlessly expensive, would under-utilize the potential of Agentic AI, and could risk making customer experience worse.
To explain, I need to start with one of the basic principles of customer service:
With very few exceptions, most customers do not want to spend time contacting customer service
If a company were to offer customer service in the form of an array of agentic bots, they would be setting in stone an approach which always requires customers to contact customer service. This means customers would always need to:
Find out how to contact the company
Be careful to describe their need in the right way, so that they get the right outcome
Try to navigate the complexity of the company’s internal structures and processes if the bot is unable to help first time
In fact, if the bots are relatively simple to deploy and companies choose to use these rather than building robust self-service journeys, then contact volumes - and customer effort - may go up.
Plan for the strengths of Agentic AI
Instead of simply replacing human tasks, more interesting possibilities emerge when companies use the unique capabilities of agentic bots to reinvent their customer service model. These strengths include:
An ability to remember all previous interactions with that customer, and provide a personalised response
An ability to run multiple processes in parallel, to more effectively diagnose a customer’s need and to find the best solution
Being always available and running at full productivity without needing a break
What this points towards is the development of bots which don’t just answer customer’s questions, they solve customer’s high level needs.
For example, a customer’s question may be “When is the next available flight from Bangalore to London,” but their need may be, “I want to get home before my son’s birthday so I can give him an amazing surprise.” Imagine an agentic bot that can not just book the flight, but also the taxi home, and a birthday cake and a big teddy bear! An agentic bot could orchestrate all of this, with the service itself being delivered by a combination of bots and humans. An experience which would previously have only been available from a high-end concierge service could now be made available to all customers.
One implication of this model is that agentic makes more sense as a customer-side tool than a company-side tool. For example, if you have your bag stolen, you may need to contact your bank to get your cards blocked, you telephone company to get your phone replaced, and a locksmith to get back into your home. It would be a far enhanced experience to interact with a single AI assistant to get all of that solve, rather than three. I think over the next few years we will see a battle to own the front-end customer service interface between service providers, hyperscalers and device manufacturers, all vying to be the one-stop personal assistant of choice. First movers are likely to learn valuable lessons.
The full potential of this tech, running at scale, is still likely to be many years off. Now is the time to start imagining and planning the new customer service models that it can enable, so that when the tech is ready, we don’t just follow the obvious path of replacing human tasks with bots.
Recommended news articles
CX Today: Google launches an AI agent that will call customer service for you
Financial Times: Virgin Money chatbot scolds customer who typed ‘virgin’
Forrester: You’re overlooking the key to customer service automation: tacit knowledge
Latest perspectives from BCG
US workers are checking out. Here’s how employers can reconnect
Right now, employee engagement in the US has fallen to the lowest level in a decade.
Less than a third (31%) of respondents to Gallup’s survey of almost 80,000 US workers say they felt engaged— defined by involvement and enthusiasm in their work and workplace.
Employees choose every day how much energy they bring to their work. There’s a real difference in the value created by someone who enjoys what they do versus those who just show up and bide the time until the workday is over.