Over the last few years while doing conversational projects, we have learned the following lessons that helped our customers and partners to create smarter chatbots and voice actions, and deliver them faster than the competition. We’d like to share some of our experiences with you in the form of 5 tips.

Being a forward thinker, you want your company to stand out in the digital arena. Customer experience is at the heart of your digital strategy. You have heard of Conversational AI, and realise that it is the next step in your plan. You know that doing this right gets you closer to your customers and can set you apart from the competition.

If you are or already have been exploring the market but don’t yet know what exactly to look out for, this article is exactly for you. In the period to come you’ll have stakeholders to convince, a team to build or transform and decisions to be made. Doing this right will give you a head start, whereas making mistakes here can set you back for months. The lessons here will help you avoid the common traps.

1. Don’t go for turnkey if your goal is to stand out

For many niche markets you see turnkey solutions popping up, advertising one-click chatbots for events, hotels, planning, et cetera. This almost sounds too easy, doesn’t it? Most of the time, that’s exactly what it is: a shallow solution with limited possibilities to customize.

But what if customer experience is one of the major differentiators in your market? Wouldn’t that make you just one of the rest? Can you actually create your unique customer experience this way?

Turn key solution often are easy to get started with, but will soon limit you as you are trying to move beyond the first steps.

Though it’s good to start with a proven templated solution you want to make sure it has a broader foundation that allows you to integrate, customize and grow with your business.

2. Avoid the greenfield trap

Everyone knows: building anything from scratch is risky and expensive. Remember how websites were built 15 years ago? If you were not careful, you would have your team or digital agency build a proprietary CMS for you and handcraft all HTML and javascript from scratch, costing a fortune and months of work.

Don’t be tempted to make the same mistake only because we are doing conversational UI now instead of a web UI. Software developers are often tempted to think that creating a conversational UI is just a matter of writing a decision tree. However, conversations with a human being flow much more naturally.

Choosing a technology provider or an agency that has years of experience under their belt will help the users of your chatbot immensely.

3. Stand on the shoulders of giants but don’t get locked in

Natural Language Processing (NLP), is one of the main parts of any good conversational platform. Its role is to classify the intent and extract entities from what the user says. Most of the solutions use Machine Learning for. Trained on a big publicly available corpus it provides a general language “understanding”. Adding your own domain specific training data is required for the AI to understand your customers language. This training data is your capital. The rest is commodity.

So concerning your choice of AI provider, choose wisely. Which solution doesn’t lock you in? Which solution can rely on the biggest datasets in the world for it’s machine learning corpus? Which solution can rely on the best scientists for its algorithms. And considering these AI’s are commoditizing as we speak, which player will be most cost efficient in the long run?

In the short term it might seem that local niche players can outperform the big ones in a certain domain or language, or you can even try to build it yourself. But realising that the future of digital is conversational, should you choose a small player? Who will be the winners and who will be losers? On which side will you be?

If you choose wisely, you can save a lot of money which you can then spend on actually building better conversations, which is where you can make a difference in the marketplace!

4. Design to scale from the start

You will not be building one chatbot, you’ll be managing a dozen. In the near future there will be chatbots for all kinds of purposes. Think of them as little assistants, each fulfilling a specific purpose. A well architected platform should allow you to manage your bots at scale in an efficient way possible. Your team will use it to build, develop and test dialogs, analyse KPI’s, manage content, integrate backend systems, operate human takeovers, manage channels and collect and govern contact data.

Your bots will be sharing certain traits, but they will also have their own specific ones. These traits are based on knowledge, and learning is never finished, the same as with humans. Therefore, expect a majority of the cost and effort for maintaining a chatbot will come after first deployment rather than before.

A good conversational platform allows for sharing of traits between bots. For this to work at scale you need to have proper version management. Due to machine learning and the fluent nature of conversation, your chatbots will change faster than any other technology you ever managed before. Continuous integration, allowing for instant deployment without losing quality assurance, is a must-have for any good conversational platform.

When selecting your platform, ask yourself: will this platform allow me to manage bots at scale? Will it allow for continuous integration in real time?

5. A well trained base bot gives you a headstart

How often do you come across a chatbot that responds “I don’t understand” the first time you ask it something? What happened to finding you the right answer, even if it isn’t trained on the subject. We are talking to a computer for pete’s sake. The least it can do when it doesn’t know the answer is look it up on your website right? This, among other basic skills, like properly opening and closing a conversation and answering basic questions, is something you want your chatbot to be able to do right from the start.

We learned to accumulate basic training, basic behaviour and general knowledge into our basebot so everytime we do a project we start with our smartest chatbot. This way we get a headstart and only need to focus on the specifics for this particular purpose and audience.

Don’t think of it as a standard chatbot. That is something else. As discussed, you want to stand out as a business. But a headstart is a welcome shortcut to a great conversational experience, because you can spend the majority of your budget on what matters: the specific traits, tone of voice and functionality for your business. You’ll rest assured the rest is already working and tested.

Want to know more?

I hope these insights have given you an idea what to look for when embarking on your first Conversational AI journey. And so you know, here at Botsquad we are always happy to provide you with guidance to help you navigate the ever-changing conversational landscape. Just contact us us if you need to know more.

Thanks for reading!

Niels Wildenberg

Niels Wildenberg

Co-founder Botsquad

email me