Tuesday 7th March and Thursday 9th March, we invited a small group of customers. Our goal? To discuss the present and the future of bots. 2 sessions, 10 chairs, 1 screen and the much-needed refreshments. But above all: 2 animated discussions, and 2 times the input we need to continue improving our bot solution. Because, as the expectations are extremely high, it still is a new and young technology that keeps evolving into a more mature solution.
At RingRing, we are at the forefront of bot implementations. In customer service, as well as in transactional communication. Not long ago, most customers came to IPG for time and means. In other words: they needed manpower, and extra capacity for their call centre. That has become increasingly more a commodity product. Today, clients talk about their challenges, and want an innovative solution. It’s not about calls or contacts anymore, it’s about interactions.
The Koramic2Engage group, with RingRing and IPG as vital parts, combines both human and digital solutions. And that is where the future of customer support lies: partly human, partly digital. Let’s explain how we integrate bots into customer support.
A bit of history
The first bot was introduced in 1964. ELIZA was the first natural language processing computer program. It simulated conversations by using a pattern matching, giving the illusion of understanding. From then on, bots kept evaluating. In 2001, Smarterchild was an intelligent bot widely distributed across SMS networks. This was the first step towards more advanced bots, such as Apple’s Siri and Samsung’s S Voice. In 2006, IBM introduced its Watson, a bot initially designed to compete in the American game show Jeopardy. Today, bots such as Apple’s Siri, Google Now, Amazon’s Alexa, Microsoft’s Cortana and Facebook’s bots for Messenger pave the way for a new era. An era in which bots combine artificial intelligence and machine learning.
How do bots work today?
A bot is a piece of software that connects with different sources. It’s an Artificial Intelligence ‘person’ you can interact with through different channels, such as SMS and Facebook Messenger. You could say a bot is a hyper-connector, which connects with a multitude of sources. There are four types of sources that make a bot functional:
- Any public API: they tap into big Web services, from Google and Microsoft among others, and give the intelligence the bot needs. For example, if your bot gives your customers directions on Google Maps to your office, it’s thanks to Google API.
- CRM: thanks to a connection with your CRM, your bot has insights about the history of your client. And is able to write new data into your CRM.
- Conversational API: this is where your bot finds language API’s. Therefore, he’s able to understand your customer and create meaningful conversations.
- Communication channels: these are the platforms on which your bot creates a conversation with your clients.
What can bots do today?
Bots today are able to have simple multimedia conversations. Let’s assume your customer has a question about his invoice. The bot asks all relevant questions, such as the client number, (a picture of) the invoice, where the issue lies etc. If the bot has gathered all information, it has two alternatives. The first is to respond itself – if he’s able to. The second is to send all gathered information to a human agent. The latter has the head start, and can go directly to problem solving. Allowing your customer service to be more efficient, and faster.
Bots can also communicate with all systems attached. As most systems today have an API, connecting isn’t complicated. The bot can simply log all relevant details of the conversation into your CRM.
Yet, today a bot still needs human back-up. The language component is one of the hardest things to develop. The bot connects to conversational API’s, which are connected to language databases. These language databases are not formed by software itself, but by humans. Bots today have no real learning mode. However, the (near) future can be different, as the AI technology evolves at the speed of light.
Talking about the future. We predict that bots will be able to:
- Have more advanced conversations based on machine learning
- Replace apps, as you will be able to ask any type of question to a bot
- Enable voice: a bit like Siri does today, but even smarter
Some observations from the clients that were present:
“We had a huge recall action, and turned to IPG to help us with handling all calls, which was about 40 calls per minute. That’s huge. The integration of a bot would be great to capture most of our clients’ questions.”
“I really thought bots were already more advanced. Language recognition and smartness are really in their early stages. It’s too bad a bot can’t really interpret the questions of clients, and still is dependent on language libraries.”
“The good thing about bots, is that they can go multichannel, and handle a thousand – simple – questions at once, without getting tired.”
When to introduce a bot?
Not all roads lead to bots. But they are relevant to companies wanting to achieve better high-value interactions. By letting bots take care of no-value or low-value interactions – such as Q&A’s or basic questions about invoices and order statuses – your customer support agents are able to create extra value. And a meaningful service.
How to start?
Every situation is different, and needs a profound analysis of your needs. Generally, we recommend our clients to start with the implementation of a ‘basic’ bot with restricted answers. From then on, they can gradually add new features to it. That’s the good thing about bots: they can grow organically and can connect to new and improved API’s. Providing news features, and better interactions.
Bots: 8 do’s and don’ts
- Don’t fool your customers. It should be clear to them when they’re talking to a bot, and when they’re not.
- Be strategic. And think about the possible added value of a bot compared to your other channels.
- Know your audience. Test your customer and analyse if there’s really a market for your bot.
- Monitor your Bot and ask your customers whether they would recommend it.
- Keep it simple and plan an exit to a human.
- Use structured input whenever possible. Primarily at the start, to avoid misconceptions.
- Handle the first contact with extra care. Your first impression is vital.
- Keep your tone of voice. You’re still in contact with your clients. Keep yourself in the way you communicate through your bot.