We’ve had calls and emails answered at 2 in the morning but those were extremely interesting cases. Most of the queries can wait until next morning but why make people if you can have someone to talk to them in real time.
The idea of a chatbot that could engage in an informative conversation, not pretend like it’s a human and not be a dumb answering machine popped up when the technologies capable of introducing domestic AI started to gain traction. If there is a benefit in using a bot to perform conventional tasks and take the weight off the employees, where does it end at this point? Can there be a potential beyond just digital script’s ability to do the predetermined things fast?
We decided to put our own assumptions and abilities to the test but first, there had to be a statement. My task as a writer was mainly a two-part job:
- Do the research and see if a bot can be of any business use.
- And if so, create a formidable communicative framework.
We took it a bit further and hopefully, the things we’ve learned could later help us incorporate more meaningfulness into our own bot.
Chatbot’s business value
Mobile devices with various applications and responsive web technologies reshaped the way we operate on our daily basis. Today it’s possible to run a full-blown business out of your phone. You can build a digital empire based entirely off of people’s addiction to being plugged in all the time.
There is no detrimental issue in spending up to 5 hours on your phone though, if you know what you are doing and only using the technology for the good. Ultimately, great technological startups fire off with that mindset.
Virtual front desk
The aspiration to create a 360-degree positive experience of your brand can’t happen unless you provide interaction and support. Customer service that once started as a person in charge doing all the feedback management, followed by the establishment of a separate customer satisfaction branch. Call centers evolved into contact centers, with global trend pointing towards cloud contact centers.
With that said, technological advances in the field of machine learning, AI integration, and overall better user experience lead us to understanding that a voice inside your product might transmit more than just your message, but also be the helping hand and a reliable assistant. The one you can entrust problem solving to.
Virtual assistant
Ever since Siri was introduced as a virtual assistant capable of solving various tasks while bringing a certain level of personality, the AI and NLP engineers have been obsessed with perfecting the technology and creating the ultimate assistant.
The crossover of AI and anticipatory design is where chatbotic functionality may shine. After all, it’s not the excitement of interaction with the machine that makes up delightful experience, it’s the solution to user problems that we appreciate along with the time it takes.
Human interaction may sometimes get compromised with linguistic or cultural differences and a number of other variables that aren’t always under control. However, the satisfaction you get after performing a task that involves dealing with another individual is second to none. Chatbots can retain the feeling of accomplishment while getting rid of the modalities of real-life communication. Let’s leave those things outside the business.
This issues a specific challenge for the writer as not only do they have to come up with an extensive list of queries, commands, requests, and orders, that bot has to understand, but also with scenarios that create empathy and aid interaction.
Virtual seller
Social awkwardness is a phenomenon that for the longest time was not considered a real problem. Supermarkets, total internet anonymity, privacy protection, and overall secrecy that once became a norm, perpetuated the syndrome to create a distinguishable group of people who later switched to catfishing, trolling, and faking everything that can be faked. Some call it freedom, some call it overindulgence, we are not going to get into that now.
But guess where this doesn’t happen? Small towns and close communities, bound either by Dunbar’s number or whatever natural mechanism, are rarely characterized with sociopathy. If you have to go out and buy your food from a local store every single day, people will know something about you. Your preferences, tastes, and demeanor will be analyzed.
In a way, this takes us back to a bazaar or emporium culture where people interact with people directly. At the same time, it’s plain business, bargain, even argument where the truth is born. If a chatbot knows your tastes, can read your mind like an experienced merchant just by watching you, if it doesn’t require you to navigate dropdowns and filter things, but can understand your direct input instead, this will be a whole new level of shopping experience tied together with personal interaction.
Writing for a seller bot won’t be an issue. There are tons of strategies and sales technologies to take from. The underlying principle though, remains intact: give value, have meaning, save time.
So, you might benefit from a chatbot businesswise if:
- Your support service needs to be accessible, reliable, and you can put together a complicated algorithm.
- The procedures of your product are complex and require a narrative to guide users through the steps.
- You want to benefit from an interactive user experience and create a meaningful presence while saving on human resources.
Chatbot’s communicative framework
In my case, where I only have to come up with a bot’s voice/tone, and a replenishable number of input phrases of the natural language that a bot can understand. At the same time, writing becomes the bot and there is no way around a certain logic that has to be implemented.
At this point, I have no idea how technically difficult it is to make a chatbot. All I know is my friend’s stories from working on a translation bot for Yandex. They use real-life queries, conversations, texts, etc., everything they could find.
All the subtle meanings people put into words communicating with one another are recognizable within the context. Is it even possible to teach natural perception to a chatbot?
In order to break down the process, let’s take a look at a high-level architecture of a chatbot.
You can use any of the popular chatbot APIs to process the input and create a chatting experience, but we are more interested in the communicative aspect, which leads us to the field of psychology.
Chatbot psychology
The human ability to read undertones, gestures, timing, and tempo of speech gives 93% of the internalized information, which leaves a chatbot with only 7% of verbal input that it has to make sense of. This obliges the bot to adapt and react instantly. For this, the chatbot has to be:
- Accessibility. There has to be no problem of choice between asking a chatbot to do something or do it via UI. For this, the chatbot has to demonstrate its benefits right away, with those being saving time and effort, the initial text has to be clear and precise, leaving no room for doubt.
- Invisible presence. In order to make a user engage in a conversation, the chatbot has no right to be annoying. For example, if a person is a control freak and prefers to do their own work through UI, you can receive a rejection but have no clue about the level of irritation they declined with. For this, give them options to choose from. Along with that, provide them with the reason to come back. For that, your help has to be requested.
- Use the yes/no questions and action buttons. In the appropriate situations, the pre-formulated actions might be the easiest way to engage users. It’s important to do a real user research to collect as many patterns. Later, you can turn those patterns into actionable buttons. However, this will only work for stock situations. Don’t rid your users of the ability to express themselves.
Trial and error is the only way to get through the stage of collecting patterns. Misdirection is the inevitable thing here, after all, humans will always find a way to discombobulate a poor robot. Our task here is to avoid terminations as communicative failures and get as much information from a user as we possibly can.
Formulating action buttons and widgets in a way that can transmit not only the meaning, but also the context, and the undertones are a key to reducing conversation dropouts in the future.
Machine learning is fine and all that… But how about WE learn first?
Chatbot wordhoard
One of the basic laws of writing a copy that appeals to a wider readership, is “don’t dumb down your audience”. If your chatbot has a personality of a Windows XP popup, you can’t expect users to look forward to talking to it. At the same time, you don’t want to express familiarity. And of course, the last thing you want is your chatbot to be offensive.
The chatbot, like a human, has to have a conversational mindset. It has to embrace the fact that people will never associate a chatbot talk with a real conversation, but they will be pleased to know they are being understood and appreciated. For this, the chatbot has to possess special skills:
- Guide and help. You might not get user input right away simply because they might not necessarily know what they’re after either. The guidance here would be giving them the list of a chatbot’s abilities and the utility would be the in performing the actual tasks.
- Remember your users. Most people remember those they’ve interacted with. There is no reason a bot can’t remember millions of people visiting its home. A simple welcoming message with some personal information gathered from the last time they’ve been there might do the trick.
- Display variation. Every statement has to have at least two other ways of saying the same thing with different emotional colorings. To feel relevant, and avoid sounding like a broken record, your bot has to be able to react differently.
- Lead with emotions. Every now and then, it’s good to have some fun talking to a chatbot that doesn’t sound like a robot. Pretty much everything a chatbot can say, may have a couple emotionally conditioned variations. Those variations can be triggered either in response to the appropriate user input, or as a result of the context-reading abilities. Say, we put three groups of words that indicate the mood of the conversation a user wants to have. Plus a middle category I call quarantine as a correctional facility. A chatbot can relate to those categories and respond accordingly.
Takeaways
We are nowhere near finishing a viable chatbot, however, for me as a writer, this stage is the most exciting. With a proven business value chatbots present, it’s only a matter of time that everyone at least gives the technology a shot.
It’s cool to have a bot in charge of your website while you are away. As an overall study of your audience, the psychology behind their decisions, and the behavior patterns they showcase, there is nothing like a chatbot prepwork.