In China, where almost all the buyers we work with are based, some believe that chatbots are the new websites and that they will replace 99 percent of other apps in the years to come.

The most advanced chatbot designers today are creating lifelike, 3-D rendered characters to interact with customers online. Forget the basic text messaging bot, this will look like you’re having a video call with a real human. It will be some time before this level of technology finds its way into real estate chatbots (although perhaps less time than you think).

Unfortunately, the best AI is not yet widely available for chatbots. In the real estate world, most chatbots will be relatively basic for years to come.

Even the biggest organizations have struggled to make their bots work effectively.

• When customers asked CNN’s news bot to unsubscribe them, it wouldn’t do it. It turned out the bot could only understand “unsubscribe” as a one-word command and not when written with any other words in the sentence. So, “Please unsubscribe me” would get you nowhere.
• The Wall Street Journal bot had its own fail. Unlike CNN, it would successfully unsubscribe users at their request. But it would just as reliably send notifications immediately thereafter — no matter how many times users tried to unsubscribe.

One of the key lessons for real estate chatbots is to not get too clever. A bot that does a few basic things well will be much more successful than one that tries to do too much and fails.

Even with a basic chatbot, you can do things to make the experience more pleasant and even enjoyable for the user. This will improve user satisfaction, reinforce your brand values and improve bot completion rates. 

Making your bot more human

Don’t fall into the trap of aiming for a full natural language experience in your chatbot. Instead, establish your users’ expectations upfront so they know they are dealing with a bot. That way they are more likely to accept any mistakes the bot might make.

Also, those who tend to get annoyed when not dealing with a human, might be more responsive to a chatbot if it’s got human-like characteristics, such as humor.

Although no one is likely to confuse your chatbot with a chatty person, you can make it seem more human — and thus more pleasant to interact with. Here are some tips for making this work.

Use emojis

Austrian real estate chatbot zoomBot uses emojis to be friendly. But it has an ulterior motive as well. Emojis also allow zoomBot to prime the user with important information visually. That way, the user can understand the gist of a message at a glance, without having to read all the text.

Let’s say your chatbot plans to ask a user the following question, “How much do you want to spend?” If you insert the emoji that looks like a stack of dollar bills into the text, they will know at a glance that you are referring to money. This will increase user comprehension and response rates.

Given that few users dedicate their full attention to their chatbot conversation, anything you can do to improve comprehension is valuable.

For some examples of emojis that could be useful to a real estate chatbot, take your inspiration from the Realtor-centric emoji keyboard released by The California Association of Realtors, which you can download in the App Store.

Casual language

It’s important that your bot be conversational. If the conversation is too stiff, customers won’t be into it. To help people feel like your bot is warm and friendly, choose your responses wisely.

Rather than “Hello. My name is (insert name of chatbot),” you could try “Hi there. I’m (insert name of chatbot).”

One highly regarded weather chatbot, Poncho, greets new users like this: “Oh, hey! I’m Poncho. I can send you daily weather forecasts and make your day a lil’ less boring.”

In reality, the type of casual language you use will depend on your brand’s voice. If your chatbot can be casual while always staying on brand and in character, it will be a success.

Loose grammar

You may be one of those sticklers for good grammar and spelling. But in the world of chatbots, grammar can get in the way. Think of chatbots as text messages rather than as email, and you will be in the right frame of mind.

We are very loose with the rules when text messaging. We might skip punctuation marks, abbreviate words and create new acronyms on the fly.

Chatbots do need to maintain some brand standards, but they don’t need to be as uptight about grammar as an English teacher (with all due respect to English teachers).

For an example, let’s look at AIRA, a finance news chatbot on the Facebook Messenger platform. Despite having a relatively sophisticated subject matter, AIRA often skips commas and periods in shorter sentences. For AIRA, only longer sentences merit proper punctuation, and even those don’t get it all the time.

AIRA is also notable for the little mistakes it makes. In a chat with me, AIRA called Toyota a “manufacture” — missing the requisite “r” at the end of the word.

ARIA wrote a question but included too many spaces between the last word and the question mark. And when ARIA called someone overly positive, the bot left out an “o,” as written here: “to optimistic.”

Humor

Your chatbot won’t be mistaken for a person, but it should be as lifelike as possible, and that means having a sense of humor.

But remember, though it’s important to give your chatbot a personality, you should never lose sight of the ultimate goal, which is to help the user accomplish something — such as set up an appointment or arrange for an agent to call them back.

“Use personality and humor sparingly, so it doesn’t get in the way of utility,” said chatbot expert Emily Withrow of Quartz Bot Studio. “If the personality gets in the way of utility, people will abandon ship.”

The safest humor to use is self-deprecating. Obviously, you will want to stay away from politics, religion or innuendo.

The weather app Poncho, mentioned above, is a case in point. When you allow it to access your location, it replies with a joke.

“Oh, Los Angeles, California? Is that the right city?

“Oh, my ex is from Los Angeles, California. Looks like it is 71 degrees and sunny there RN.”

When to stop talking

You can program your bot to be polite, friendly and funny, but you can’t control how people react to it.

There will always be some users who want to type profanities into the chat, so be sure to build in a mechanism to deal with that behavior.

The best thing your bot can do when someone speaks rudely is to politely end the conversation.

Carrie Law is the CEO of Juwai.com. Follow her on Facebook or Twitter.