What we learned from bots, before they were cool

Last week I bumped into Robert Hoffer, the famed creator of SmarterChild, the automated chatbot that used to sit at the very top of everyone’s AIM Buddy List.

For many people, it was the first experience conversing with a pre-programmed tool over a traditionally human-to-human channel. While learning more about SmarterChild’s childhood from Robert, I was reminded of the time before chatbots were “cool.”

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1987655,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,dev,mobile,","session":"C"}']

Long before Messenger, Whatsapp and Telegram came on the scene, SMS was the most intimate way for a brand to reach a customer. And we must give credit to the brands that had the foresight to experiment with a new technology across this personal communication channel.

In light of the current chatbot revolution, I want to share some key insights developers learned from those early days building bots.

1. Scope and expectations

It is essential to clearly outline the scope of what the chatbot does, before the user engages with it for the first time. This establishes an appropriate level of expectation ahead of time, and puts the user in a desired frame of mind when communicating with the bot.

Today, while chatbots are still relatively weak, it’s best to tell users exactly what they can do up front. For example, if it’s a weather bot, users should know not to ask it questions about food.

Gartner predicts that by 2018 a full 30 percent of our interactions with technology will be through “conversations” with smart machines. Major players have invested in the AI space, with a focus on leveraging their core AI assets, such as the Microsoft Bot Framework, Facebook’s Wit.ai, Slack’s Bot Platform, and Google’s TensorFlow. However, the biggest frustrations in the chatbot experience come from unrealistic or overinflated expectations. When half of Hollywood movies are raving about the endless power of AI, it is even more important to set realistic expectations of what the chatbot can and cannot do.

2. Reactive vs Proactive

How a user engages with a bot for the first time will determine the bot’s virality and ultimate adoption. There are three basic approaches.

Reactive bots: the customer will see an ad and a method to communicate (SMS shortcode, QR scan, etc), giving them the initiative to reach out. This provides a high level of integrity and respect in your brand communication.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1987655,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,dev,mobile,","session":"C"}']

Back in 2013 before the bot-rush, DigitalGenius used to build reactive bots for companies like Panasonic, BMW and Unilever. The customer would see a TV Ad with a phone number to text. (Messenger was not yet a standalone app). They would manually enter the number into their device and send their first message as if they’re texting a friend. This was a simple, non-intrusive approach offering consumers a high level of integrity and respect — very fitting for a brand like BMW.

Proactive bots: these message the customer first. They are cool because they are unexpected and can drive a much higher initial engagement rate than reactive bots. However, they can be perceived as spammy, and therefore must deliver a pleasant experience and extremely valuable content to the user right away.

Examples or use cases would include surfacing a boarding pass when the customer arrives at an airport or sending a message when a food delivery order is about to arrive.

Mixed engagement bots: these combine both approaches. The user will send the initial message, showing their interest to engage. The bot will then take them through a conversational experience to learn more about the user, and eventually serve proactive messages as needed.

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1987655,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,dev,mobile,","session":"C"}']

Once the initial conversation takes place, and rapport is built, brands can proactively push valuable messages to the user when they need them most.

A great example is x.ai, a chatbot which helps you schedule meetings through the use of deep learning and natural language processing. The tool can be summoned (reactive) when you want to offload the scheduling process, or can ping you (proactive) if someone is trying to schedule a meeting with you.

3. Channel

As a consumer brand wanting to play the bot game, it makes sense to be where your users are. Probably living inside Messenger, Whatsapp and Snapchat.

This is one of the biggest reasons why companies like Telegram and Facebook have put a huge emphasis on chatbots and messaging. When a brand has a bot, especially one powered by AI and “continuous learning”, each conversation becomes a historical log, which should be used by the system to get smarter over time. Therefore, the more users on the platform, the faster the baby-bot will grow up and become a smart-bot.

[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":1987655,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,dev,mobile,","session":"C"}']

WeChat’s popularity has exploded, with approximately 700 million monthly active users in China, alone.The platform initially started out as a messaging app, but has expanded into the retail industry. Using WeChat users can conduct shopping-related activities, including buying movie tickets and ordering food.

4. Evaluation and metrics

As with most emerging technologies, there is a period of time with lots of action but not enough palpable insights on what’s actually working. This is why it’s extremely important to establish a way to evaluate and measure the chatbot experience for brands. A great example of this: ChatFuel’s integration with Botan.io — which allows bot builders to gather precise analytics for their creations.

Further, the quality of the underlying AI engine the bot is built on can be measured by its resilience to unexpected queries and the rate at which it can learn over time.

For instance, AI deployed directly in a contact center can help customer service agents answer questions. The agents are using the tool every day and achieve measurable efficiency gains, but as a byproduct, their usage is naturally training the AI to get smarter over time.

[aditude-amp id="medium4" targeting='{"env":"staging","page_type":"article","post_id":1987655,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,business,dev,mobile,","session":"C"}']

By deploying a bot, or an artificial intelligence layer in this environment, with customer service agents who actually know the most about a brand’s customer service process, it’s possible to train up a much stronger and more accurate model in a shorter period of time. Therefore the customer service agents unintentionally become the AI trainers, while the tool helps them perform better in their role.

Conclusion

The chatbot space is in hyperdrive. Though chatbots are very weak right now, we need to give them the time and  and a supportive environment to grow up. By setting realistic expectations, creating the right engagement strategy, and choosing the correct channel of communication, brands can benefit greatly from these early days of bot building.