How to add OpenAI Assistants to your website or WhatsApp

March 22, 2024

In today's competitive business landscape, customer support is increasingly a differentiator. Customers set a high bar - prompt and accurate replies, provided 24/7, is now taken as a given. 

The challenge for businesses is how to provide this level of support without breaking the bank. Human agent time and training costs are inexorably rising, and traditional technology solutions are increasingly expensive to configure and implement.  

The advent of AI LLMs (large language models), as popularized by ChatGPT, provide a great opportunity to solve these challenges. 

It’s now possible to create your own custom AI Assistant that can understand your business and products in infinite detail. These assistants can provide contextual responses to your customers’ queries without you needing to create complex workflows and conversation paths (required back in the “old days” of building a chatbot), making them quick and cost effective to implement.

This blog describes how to create AI Assistants on OpenAI’s platform and connect them to your website and WhatsApp account.

Why an AI Assistant?

AI Assistants’ key feature is their ability to be trained on specific information, such as your website, product documents, knowledge base etc, providing it with extensive and in depth knowledge of your business and products. This enables it to stick to the subject when asked a question, and not “hallucinate”. 

With this capability, AI Assistants have the potential to revolutionize customer support by providing fast and accurate responses to customers' questions to a standard as good as, or better than, a real life human agent. 

Of course, AI Assistants also have the advantage of keeping going 24/7, without breaks, holidays, days off sick - all at a fraction of the cost of a human.

Adding custom AI Assistants to your customer touchpoints not only reduces workload for first line support staff, but also unlocks some other advantages:

> Intelligent, context-aware customer messaging eg based on previous customer interactions  

> Less repeat questions, and therefore more interesting and satisfying work for human agents

> The ability to provide scalable and personalized engagement at low costs

> Data-driven insights for continual optimization

So AI Assistants can help provide you with that competitive edge your business needs! 

But how do you create your AI Assistant? You can go to one of many providers who have built “wrappers” on Open AI’s APIs, allowing you to build chatbots based on OpenAI’s LLMs and infrastructure.  

But in November ‘23, OpenAI themselves announced that they would offer the ability to create AI Assistants on the OpenAI platform. This is in effect a version of ChatGPT that businesses and individuals can train with their own data.

The challenge with Custom GPTs is that they can only be accessed through OpenAI’s GPT Store using a ChatGPT Plus subscription, which somewhat limits the reach and access for users…so here at Converso, we set out to fix that by connecting OpenAI Assistants to our helpdesk platform.

This enables you to easily add an AI Assistant to your website or WhatsApp channel, without needing any code or other tools.

It’s important to note first that when deploying your AI Assistant on a website or WhatsApp channel, it uses behind the scenes OpenAI’s Assistants API, which is chargeable. This compares with a user accessing a Custom GPT via their ChatGPT Plus subscription, which does not incur any usage cost for the user (apart from their Plus subscription) or the business.  

Adding an Assistant to Your Website or WhatsApp account

The advantage of using OpenAI’s platform is that it’s quick and easy to set up an Assistant, and does not require any technical knowledge. Below is a summary of the key steps:

  1. If you don’t already have one already, create an OpenAI ChatGPT Plus account.

  1. Create an Assistant and configure it for file retrieval. 

  1. Upload your business data eg product details, FAQs, knowledge base etc. This trains the assistant's understanding of your business and products. 

  1. Create a prompt that customizes the model's tone, formality, and response depth

  1. Link with Converso by inputting the assistant ID and your Open AI secret API key. This enables us to use the Assistants API to connect your Assistant to our platform.  

Congratulations - you are now connected to Converso!

From here you can deploy your Assistant on any network connected to Converso - these presently include webchat, WhatsApp and SMS.

To connect to webchat on Converso:

  1. Go to Admin > webchat, and configure your webchat widget with your brand colours, logo etc 

  1. Embed the widget on your website, this can be done by yourself (with a little technical knowledge) or your web designer. Alternatively our customer support can assist you. 

Your website visitors are now able to connect with your Assistant directly from your website. 

Alternatively, you can embed a link on a specific web page (eg on the internal HR pages of your site) or on social media, for a more focused chat interface.

Your Assistant can also be accessed via WhatsApp. 

To do this, it’s necessary to create a WhatsApp Business Account using the WhatsApp Business API. This can be done easily on the Converso platform in just a few steps, again without the need for any code:

  1. Go to Admin > WhatsApp and click on Connect to WhatsApp. This will take you to the sign up process for you to create your own WhatsApp Business Account. 

  1. Configure your existing phone number as your new WhatsApp business number or choose a new number.

  1. Go to Admin > AI, and connect your WhatsApp number to your assistant ID to route all customer messages to your Assistant for automated responses.

It’s as simple as that - in just a few minutes, your Assistant is connected to WhatsApp, and ready to answer your customers’ support questions! 

Managing your AI chats

Once you’ve got your Assistant connected, you also need to consider how you’re going to manage the resulting AI chats.  

This is a subject for a blog post on its own, but below are a few important operational considerations:

> Escalate messages to your support team

Whilst many customer support queries can be answered promptly and accurately by an AI Assistant, there will be a proportion that need human input. This is where it’s important to be able to hand over the chat to a human agent. This transfer should also enable full visibility of previous chat for context, as well as the ability to assign the chat to a specific agent with the necessary skills to answer the question.  

> Analyse all incoming messages

Being able to view live chat messages will enable you to analyse customer behaviour, and proactively spot issues or problems before they become significant. This is not possible to do within OpenAI’s UI, which can limit your ability to gain valuable intelligence from these customer interactions.    

> Collect user contact details for leads

Pre-sales customer support can often include sending follow up information, meeting times, etc to the prospect. Therefore it’s very useful to collect contact details during the chat (in a secure manner) so that proactive contact can be easily made via SMS, WhatsApp or email, otherwise that prospect may be lost. It’s also important to be able to respond quickly to any request for product info etc, and being able to save contact details from within the chat makes this much easier.

Converso's platform addresses all of these issues, making it easy to both integrate Open AI's Assistants into your digital channels, and also manage customer and AI chats, all in one place. 

The creation and integration of AI Assistants can help you embrace the future provided by conversational AI: delight customers, drive efficiency, and stay ahead of the customer support curve! 

September 18, 2023
September 18, 2023

The AI-Powered Chatbot Revolution: How Large Language Models Are Transforming Customer Service

read article
July 21, 2023
July 18, 2023

How to increase productivity by upgrading to the WhatsApp Business API

read article
July 8, 2022
October 28, 2022

Making WhatsApp work for small businesses

read article
January 17, 2022
October 28, 2022

SMS payments - the benefits of text to pay

read article