My chatbot is dead

Why yours should probably be too

google lens

Personal websites are usually like old books in a shelf. They languish, accumulate dust, and their wrinkles and cracks become more apparent over time. About 3 years ago I embarked on a simple experiment that would end up prolonging the shelf-life of my website by an unusually large margin.

Back in 2017, it seemed like Conversational UI was poised to take over the world. We saw Quartz turning news into a conversation, WeChat being featured as the poster-child of a post application world, iMessage turning into an unnecessarily complex mess, and chatbots popping up like mushrooms in moist forests.

Of course, any trend gaining so much traction and interest needs to be taken seriously. As such, I decided to familiarize myself with the topic, and turned my website into a chatbot.

Instead of being greeted by the internationally standardized greeting every designer used at some point in their career, there was no bold, dramatically oversized, and deep black sans-serif reading: Hi, I’m a designer. (To be fair, I didn’t use Proxima Nova either)

Instead, a couple of chat bubbles exuberantly ushered onto the canvas to greet users as if we had all been long time friends.

Conversational intro

It was witty, new, and slightly awkward. People would send messages that ranged from simple chit chat, to deep philosophical topics, to downright disturbing and ridiculous insults.

The experiment got featured on Hackernews, Medium, was used in psychological studies conducted by Dan Ariely’s team, and the source code was ripped and edited by various startups to fit their needs. One business in the Bay Area had an idea to use it to sell flowers in a conversational way. It looks like they went out of business.

The reaction and feedback was surprising to say the least. It was an idea so simple, so silly, that the outcome was in many ways unexpected. After all, the only one who really cares about your website, is usually yourself.

That didn’t stop me from revamping my website and kill the very thing that had turned it into a micro-celebrity before. With the death of my old chatbot, some angry emails by schools who are using it as a reference for “creative” web design, and a good amount of time that has passed ever since, I wanted to take a step back and set the record straight.

When chatbots matter

So let’s be honest with ourselves for a moment: when did you actually ever enjoy talking to a chat bot? And I’m not talking about the type of bots you talk to when you’re bored, but about those that provide a deeper purpose.

It turns out that the answer is, at least for most of us, almost never.

I love you Intercom, except when I don’t. 99% of time I don’t want to talk to a silly and obtrusive avatar popping up from some corner of the screen before I even had a chance to check out what’s going on. Somehow, I can’t help but think others feel the same.

In fact, we do know that others feel the same. Chat heads jumping at us unasked, are the quintessential equivalent of the infamous sales clerk who eagerly talks to us upon entering a store.

To further add to the challenges: as soon as users go off-script, chat bot’s don’t just become awkward and unpredictable—they turn into little sociopaths that might rub users the wrong way.

UX — Conversational UX News

The moment you create a chat bot is the moment you allow customers to have a conversation with your brand. Not with yourself, not with your friend, but with an uber entity—a symbol—that represents everything you and your team stand for. That’s not a step to be taken lightly.

This simple conversational entity can be a fun tool to engage with people but depending on how the conversation goes, it can quickly turn into a misrepresentation of the values of your team and your company. So building a chat bot should never be the default choice, but an intentional one.

That’s why it’s worth asking yourself the following three questions before venturing into this space:

1. Is your use case simple enough to be solved through chat?

Conversation is incredibly complex and it’s challenging enough to keep it on track in the real world. If the use case isn’t simple, chances are, chat bots are not the right tool for the job.

2. Is your NLP capable and sophisticated enough?

There are two types of bots: pre-scripted bots with a range of default answers users can choose from, and Natural Language Processing based ones.

Choosing the right one is hard. While pre-scripted can feel too limiting, NLP can break at every corner. Often times, teams quickly fall into the trap of spending a huge amount of time focusing on personality and silly jokes, instead of solving the problem users hired you for in the first place.

Therefore, building on top of the first point above, within the conversational landscape, simple always wins.

3. Are your users actually in chat based environments?

Chat bots work best where users already are. If your users are primarily spending time in messaging platforms where bots and micro-apps can be seamlessly embedded, great. That can serve as an effective and natural way to engage with your audience because it matches the “be where users already are” principle.

If on the other hand, people come to your website, a medium that has made great strides to provide content in a non-linear and quick way, it often unnecessarily slows users down.

Farewell chatbot

I don’t want to discredit chat bots as a paradigm. They have their use in certain industries, medium, and work well for a specific set of use cases. The important part is being deliberate, rather than jumping ship blindfolded.

So whereas turning my website into a chat was a fun experiment, I ultimately feel like it has slowly turned into a fad. I got fooled by the trend, and as a by-product became part of the trend itself. Fads come and go, and as they get refined and re-interpreted, they ultimately find their true purpose. What we’re left with is the age old insight that it’s only through experimentation, that we can unlock concepts and ideas that last.

Rest in peace chat bot, long live chat bots.