Done right, chatbots are incredibly effective at driving sales, reducing costs & engaging your customers.

Done badly, they can lead to frustration, boredom and ultimately lost business.

Here are recommendations for building a great chat experience and some technical advice on getting started.


Customer on-boarding in Cleo's React Native application.

Projects built by our chatbot development team:

  • Chatbot mobile app to track spending habits through open banking APIs
  • Chatbot to onboard all new employees of a multinational corporation
  • Automated concierge that offers targeted recommendations and in-app booking for restaurants and hotels 
  • NLP-powered chatbot that helps wealth advisors prepare for their client appointments by researching relevant funds via conversational language
  • Chatbot to connect customers of a DIY site to the right expert

Creating Effective Chatbots

Create a friendly, intimate rapport

Mimic the friendly, clear, and conversational tones of talented customer service reps. 

Rich media like GIFS, videos, and emojis can make the experience fun!

Our client, Cleo, has created a personal finance chatbot  which resonates well with younger audiences and makes checking your balance less painful (if not a bit disappointing at times!) through liberal use of funny GIFS and emojis.

Use data to create a personalized, custom-tailored experience

The data you collect can improve the chatbot experience for your customer.

Even simple data points like location and date of birth allow you to feed relevant content.

One conversation at a time

Usually with software the MVP approach is taken. This can mean building the whole user journey in the most rudimentary way and learning what works and what doesn’t. Conversation doesn't work like this so focus on one part of the customer journey at a time.

Imagine a conversation with a human. Would you want a long, painful, non fluent conversation that ends up with nothing, or a short, clear interaction that gave you the outcome you were looking for.

Remove leg work

Use button menus and available tools like location services to point your users in the right direction. For customers, it can be difficult to intuit how to communicate with a bot so narrowing the options available not only helps simplify the experience, but makes the interaction more likely to end positively.

One company that has done this well is KLM  airlines who introduce a button menu to focus the conversation. This also lets them get away with a less sophisticated NLP.

Example Airline Chatbot Screenshot

Finding the right chatbot solution for your needs

In the same way we encourage potential clients  to prove their business model and utilize off-the-shelf tools, there are quick and easy ways to accelerate chatbot development.

The best technical solution depends on the complexity of the problem.  Some options:

  • Very simple: if your conversations & customer interactions are likely to be straightforward, even non coders can build a chatbot using the likes of Mobile Monkey.
    • This freemium tool is a great way for non-technical founders to build a bot & if you are in e-commerce the good news is it even integrates with Shopify, Paypal & Stripe for transactions inside
  • More complex: for chatbots with more firepower, Flow XO is the pick of the class.
    • With tons of integrations, great customer support & the ability to deploy your bot across many platforms;
  • Custom:
    • If off-the-shelf solutions don't offer the flexibility, integrations or total control your business is looking for, it is time to bring in a chatbot development team!
    • Your investors might also be nervous about a dependence upon a third party. Your business could be at risk if, for instance, you rely solely on Facebook Messenger and they change their terms and conditions or APIs.

Standard chatbot project architecture

Generic chatbot integration optimized

(1) The entry point of any chatbot is a messaging service. Whether you want to integrate with WhatsApp, Facebook Messenger, WeChat or any other service, even your custom chat interface — on the web or in-app — we leverage connectors (2) such as Smooch to avoid complex integrations and reduce development time.

(3) Structured conversations can then be dealt with by a custom API built, for instance, with Django or Node.js. The API will trigger actions in your existing workflows (4) or gather relevant information from your IT systems or even third party providers (5).

(6) You can also leverage artificial intelligence and machine learning, trained with the dataset of all past conversations (7), to engage more efficiently with your users.