Cognitive reasoning makes bots less artificial and more intelligent

Cognitive reasoning can deliver new products and services, better outcomes and reduced costs, with a blended approach of artificial intelligence and human interaction


Cost reduction efforts in customer services have largely delivered atrocious results. Consumers now have low expectations and are bored of long call queues and pesky IVRs (interactive voice responses). It’s not just enterprises that want a virtual relationship with customers, a significant percentage of us now prefer to self-manage our utilities, banking and insurance, but with human support when we need it.

So what role can cognitive reasoning play in improving service?

You can automate a task, but you can’t automate a person. Artificial intelligence or AI works best in narrow use cases and there are still few successful customer-facing cognitive projects. Who would want to delegate customer relationship management to two dimensional, non-conversational AI bots?

Siri, Cortana and Echo are trying to be your personal assistant, to help you find a restaurant, book an UBER or check the weather. But these cognitive interfaces are also a gateway into more advanced services. Imagine being able to ask Siri, “What mortgage is right for me?” You can see how these services could be monetised and what this access to consumers could do for enterprise.

The number of active users on messaging platforms now exceeds the number using social media. Facebook have opened a “bot store” where customers can engage in a way that is better for them. If I want to check if my flight is on time, why do I need the airline’s app? Isn’t it easier to message my airline’s bot?

Really simple bots can be powered using decision trees, but this technology is unwieldy for complex scenarios and only useful for answering a single question. Decision trees cannot learn or be “joined up” to solve more complex problems.

Machine-learning technologies tend to fail because they start with data, not an existing model of human knowledge. They require engineers and data scientists, as well as a lot of time, money and faith. You also cannot audit the reasoning.

Most technologies fail to build conversational AI that can ask smart questions to get the data needed to deliver a good outcome.

There is a different type of cognitive reasoning based on inference, one that can give you the best of both worlds, models of human knowledge that can also connect to real-time data, asking intelligent questions and making good judgments.

In many contact centres, critical tribal knowledge is locked away in individual’s heads, documents and personal notebooks. Even putting information into a shared wiki doesn’t turn it into knowledge.

Being able to model existing human knowledge in a system that can still learn is essential if we are to deliver smart bots. This technology also needs to be able to deal with the real world, which is full of bad or missing data, uncertainty and ambiguity.

Current virtual assistants are very limited in scope. Can you imagine asking Siri, “Why is my mobile signal low?” and it telling you that the rain has reduced the signal strength of your nearest mast. A glint of the future perhaps, but only possible by joining up human knowledge, real-time data with a powerful conversational interface.

Cognitive reasoning technology is best deployed alongside humans, augmenting them in a seamless way

Such cognitive reasoning-based bots could discover new facts and learn about customers through experience. Most importantly they could explain why each judgment was made, essential for building trust and demonstrating compliance.

But even cognitive reasoning-based bots must be focused on narrow areas if they are to be successful, providing consultations to questions such as, “How can I reduce my debt?” or “What is my diabetes risk?” Open-ended customer service bots will only lead to more disappointment and frustration for consumers.

Today, cognitive reasoning technology is best deployed alongside humans, augmenting them in a seamless way. The Facebook M project blends AI and humans together with a natural language interface so seamless you don’t know if you are talking to a human or a machine, and what’s more you don’t care because it works.

Cognitive reasoning can deliver new products and services, better outcomes and reduced costs, but the customer journey will improve only through this blended approach, not through automation alone.

Exploring the roles of cognitive technology starts with understanding human strengths and machine weaknesses.

To learn more about cognitive reasoning and how others are using it, go to www.rainbird.ai/raconteur/

Rainbird_Logo_WEB