We all know the frustrations of fighting our way through the options menu when we call “customer support,” and the tedium of navigating corporate websites and filling in lengthy online forms.

Often we encounter these hassles at work, as IT help desks have been replaced by complex support websites. Companies increasingly expect employees and customers to “own” much of the workload, based on the assumption that users know best what they need and want.

There’s a method to this madness: it’s called customer self-service.

However, this kind of self-service often does not make for a happy user experience. We expect customers to find the right part of an app or website, assess the available options, interact with the system, wait for confirmation and know when a task is done.

Companies expect their customers to do too much of the work in self-service systems. This problem can whack a company’s bottom line, as a positive digital customer experience can be just as important for commercial success as more obvious factors like cost and productivity.

The underlying issue is not so much with the design or user interface, but with what I would call “cognitive load” – the amount of mental effort we need to interpret a complex situation and act on it. After all, most people’s working memory can hold only four to five pieces of information at a time; so the fewer pieces of information we expect customers to process, the quicker and easier it becomes for them to make a decision.

Take the success of Uber: If you need a taxi, you don’t have to go online anymore to research routes and cab companies, then quickly get cash to pay, and make a call to request a ride. All that load has been reduced to a few taps in a smartphone app.

The same principle can be applied in other industries. Companies that want to succeed must ask themselves: How can I shift cognitive load away from my customers, to ensure they have a better, happier experience?

That’s where cognitive technologies and agents like Amelia come in. Thanks to Amelia’s ability to understand natural language, she offers the most efficient and easy-to-use way of shifting cognitive load away from customers.

To be clear, shifting cognitive load is not a trivial task. Reducing customer-facing complexity requires plenty of thinking and heavy lifting in the background, and that’s why companies will need Artificial Intelligence solutions to do the job.

The best solutions here are not big data recommendation engines, scripted chat bots or – at the high end – the all-purpose AIs being developed in science labs.

Instead, companies need to make use of commercial AIs, cognitive technologies that have been designed for a specific purpose; their success is being measured by whether they can solve a specific business challenge, not whether they can fool people into mistaking them for a human.

These commercial AIs will help companies deal with the high-volume, high-frequency queries they typically receive. They get the job done quickly and efficiently, never losing their cool. At the same time, they free up an organization’s human employees to dedicate more time to complex tasks. And when these AIs encounter a query that goes beyond their knowledge, they can pass it on to human colleagues, who are still more adept than machines when it comes to handling new and unpredictable situations.

Cognitive technologies are based on linguistic principles and designed to understand natural language. They can extract intention and generate dialogue to clarify missing information before taking action on a customer’s behalf. Customers don’t have to change the way they think or talk to adjust to the limitations of a system. There’s no need for keywords, rigid phrases, or remodelling your thought process to engage with the system. In other words, we can speak to them in a normal way.

The self-service approach to customer service has gone too far. It’s OK for simple requests, like checking an account balance. But any company or organization that depends on transactions that place a lot of cognitive load on customers and staff should investigate whether it can shift this burden onto an artificial intelligence system, with its ability to understand natural language and mine huge data-driven systems.

Parit Patel is Head of Solution Architecture at IPsoft in the UK.