Why Bots Suck and How to Improve them Quick!

Chatbots need some serious CPR… here is what we can do.

Stefan Kojouharov
Chatbots Life

--

Over the past year I have been working on bots. During this time period, I have consulted many of the top Bot Companies like Swelly, OutBrain, NearGroup… just to name a few. Here are a few major take aways. I hope this helps in your bot project and that ultimately bots go in the right direction.

Even the Giants are Struggling

When compared to their App counterparts many of the top bots are struggling to have similar engagement and retention numbers. This is a huge issue! Ultimately it tells us something about the platform that we should have intuited a long time ago, conversations are limiting.

Who’s Killing It?

There are a few narrow use cases where engagement and retention are much higher in a bot, and I mean MUCH HIGHER. These uses cases often REQUIRE Conversations. NearGroup is a perfect example of this… If you took the conversations out of the NearGroup Bot, the Bot would fail!

So, the first thing to consider when building a Bot is ‘Are Conversations a Must?”

If Conversations are Not a Must

You have to be very honest here…. If conversations are not a must, your Bot should be build around a very rich GUI. You can do this using webviews! This feature allows you to practically create an Instant App. This is the future for most bots, instant apps that you can talk to.

Instant Apps you can Talk To

Most bots will become instant Apps you can talk to. We are already headed this way, and personally I think a GUI is better 90% of the time. In the cases where it is not, users should still be able to talk to the bot.

If you agree that Bots need to become more like ‘Instant Apps you can Talk To’ let me know here.

When Should Users Talk to the Bot?

Conversations will have their place, but only when the value add is much higher than a GUI. For example, if someone hacked into your Bank Account, the best way for the bank to let you know would be via an instant text message. In this example, a text notification is by far the best way to communicate what happened.

This is how high the bar is! The text notification has to be of such high value and relevance that your users would get Mad if they did not receive the notification!

Let’s go deeper into the example. The Banking Bot would be 90% GUI and 10% conversations. If someone hacked your account, they can quickly text you via the bot. The bot can then answer FAQs and automatically route you to a customer service representative as needed. All of this can happen within the Bot, without you leaving. In this way, the Bot has leverage conversations where they are required and become An Instant App you can Talk To.

Managing Expectations

This is a big one for me… Simply put we should begin looking at bots differently. We should start seeing them more from the point of view of websites and less from the point of view of apps.

This means that we have to strongly consider which types of metrics apply to our bot. For example, most bots will not be daily use cases, so it is unreasonable to expect a high DAU numbers.

Does your use case really warrant frequent usage?

There will be a large portion of Bots that service one off cases. They may quickly help a customer complete a task, charge a fee, and never interact with that person again. For these types of use cases, the metric that matters most is… revenue growth.

Is your Bot Valuable?

During the Product Market Fit (PMF) phase, I would focus on the one thing that matters more than anything else: Does this bot solve a painful problem in the easiest way possible?

If the problem is painful, the solution is easy and users are successful… your 9/10th of the way towards PMF!

Ultimately you bots is a business and businesses solve problems. Even entertainment bots solve a problem, the problem of boredom. As such, when customers are successful, you become successful! Make sure your bot solves a problem and that your users are successful.

Long Tail Effect

This goes along with managing expectations and solving problems. For the first time, an app technology will have a long tail effect. This means that your Bot does not have to be in the top 10 in its category and have a million users in order to make money. As long as your bot is helping solve people’s problems, you will be able to generate revenue. Think, instead along the lines of a 1,000 true fans.

Can your Bot get 1,000 true fans and solve their problems?

If you agree that Bots need to become more like ‘Instant Apps you can Talk To’ let me know here.

What’s Next?

First, I really hope this helps in some small way :) If it does let me know and I can begin sharing insight from each project. I have done a good number of these and I learn something major every time. If you’re interested in this let me know here.

--

--

Building AI Agents since 2016. Today, I am creating AI Agents for Wellness & Personal Growth and Sharing my Insights. Join me at: stefanspeaks.substack.com/