Rasa is a powerful AI solution used for chatbots and natural language processing. The best part about Rasa is that it's also Open Source, meaning a few things:
- Companies such as Google, Facebook, and Amazon don't get your data.
- You can use it as often as you wish and have as many users as you want without paying additional fees.
- You can customize the functionality if you wish (although what they provide out of the box is excellent.)
- It's easy to leverage the data collected by Rasa for your own processes (personalization, quantification and categorization.)
Beyond just using it within your website as a sales chatbot (which can be very effective in itself and lead to a wealth of information), you can also use Rasa as a customer service bot – and with limited effort it can integrate into your existing support solutions and processes.
Rasa: What's in the Box
Rasa offers the following capabilities with their solution out of the box:
- Ability to train using text-based processing using three different algorithms.
- Ability to integrate to a website using their web-based code or via a 3rd-party React component
- Ability to test and collect training data (Using RasaX - a separate platform)
- Ability to automate tasks based on text - for example if a person says they want to cancel, Rasa can run a 'Cancel' task into the process.
- Able to support multiple channels - Rasa supports website integration in addition to major platforms such as Facebook and Slack.
- Ability to respond randomly - Rasa supports multiple answers and will randomly choose an answer from the training information provided.
Using Rasa with Zendesk or Freshdesk
Using Rasa in existing customer service platforms is a fairly simple process. Both Zendesk and Freshdesk offer amazing APIs for integration that cover every feature available in their system.
Collect Training Data
This is arguably the hardest step when automating customer service. With APIs, it's easy to extract the questions and answers that are provided. It takes a little bit of effort to filter out bad answers from good.
API Integration
This sounds technical, but it's really not. It just means running a job that picks up tickets and responds. Rasa provides APIs that can retrieve an answer and supports some context-level capability. The important thing in this step is to track the 'Predicted Answer' from Rasa which is as simple as storing it in a database.
Testing
Because both Zendesk and Freshdesk support internal comments, when you deploy your robot you can have it respond as an internal comment, then your customer service agent can determine if that's the correct response and either send the Rasa-generated response, or can correct the response. By tracking the predicted answer vs. a human-generated answer, this can give you the data you need to re-train and update your scripts without a big effort.
Automation
Identify which tasks to automate relative to customer service and create automations that perform those tasks. It could be as simple as looking up account status and logging it as a note within a ticket, saving a customer service agent time when identifying crucial customer details. It could be as simple as automating common tasks such as cancellations, refund requests, or free trial extensions.
Evolution: Smarter Onboarding and Smarter Engagement
The best part about a solution like Rasa is that you can integrate it into your end-to-end processes. It doesn't have to be a silo that only exists in one area of your business. The data from Rasa can be collected and analyzed as part of the entire customer journey. Here's some examples of ways that you can become smarter:
Prediction
Predict KPIs such as probability of customer churn or quantify lost revenue due to missing features or bugs.
Sentiment
Identify customer sentiment based on content, interactions and activity. This data can also be used as a predictor for churn.
Personalized Campaigns
Create personalized campaigns based on customer experiences and customer service problems. It's a terrible experience to be told 'We hope you've loved your free trial' when I haven't logged in once.
Categorization
Identify, prioritize, and categorize issues causing customer service requests. Invest in proactive training or in new features based on quantity of issues causing customer churn.
Getting Started
Getting started with automating your customer service is a fairly straightforward process:
- Identify the timeframe to collect support information from. Going back too far may have information from old tickets that isn't relevant anymore. Identify the sweet spot of tickets.
- Identify the top 3 most useful automations – either something commonly performed or useful information making it easier for someone to do their job.
- Create an aggressive timeframe for launch. A basic solution can be implemented in just a few days.
- Think differently about how you leverage your human talent. Maybe instead of supporting customers they are creating personalized content that can be used in personalized onboarding campaigns, or perhaps even prospecting for new customers and asking for referrals.