Despite its unassuming name, the AI power-up is a messenger-based sales feature that will take your chatbot’s performance to the next level — all with the help of artificial intelligence. This feature utilizes natural language processing (NLP) in order to help your chatbot understand the questions that your clients ask, and provide them with the proper answer.
This feature is only available on the Enterprise plan.
How to set up AI power-up
To begin the setup of the AI power-up feature, head over to Settings section of your amoCRM account, and click Communication tools.
Click the Enable button to turn the feature on.
There are five columns that determine the AI power-up’ functionality: intention, keyword dataset, action, threshold, and times launched.
The Intention column describes a common topic (or an intention) the client may communicate through different questions while chatting with you.
I.e: the sole intention of buying a bicycle could be communicated in a variety of questions, such as: I want to buy a bicycle, what’s the price for the bicycle, etc.
To set up the Intention, click Add a new intention, enter the name for it and click save.
If you wish to add some common questions to the dataset right away, click on the name of the Intention you’ve created and select Add a new phrase to dataset. Enter it as many as you need and click Save.
You can have multiple intentions set up that the system will recognize, all with their own question datasets and actions.
The name of the Intention can be changed at any time by clicking on the pencil icon.
The Keyword Dataset column is for uploading the file with the table of questions your customers may potentially ask, to help the system determine the intention behind each question. The more questions narrow down the intention, the more precisely the system will operate.
To upload the file with the questions dataset, press on in the … button and select Upload. You will be prompted to either add the questions to the existing dataset or replace it entirely.
The accepted file should be an Excel spreadsheet file, either in .csv or .xlxs format. The questions should be in a single column, with each phrase on a separate line. It’s best to keep no other information in the file apart from the column with the questions. You can download a template to get a better idea of what it should look like.
There is one more function – Enrich dataset with synonym base – that you can activate. It will automatically search for synonyms and enrich your database with similar, yet different questions and keywords.
The Action column represents the automatic action the system will use if it deems the client’s question to have the required intent. This happens during the conversation in real time.
As the action, you can either have a certain pre-made Salesbot sequence, or a pre-made Response template will be sent.
To add a new action, click on the … button and select either a Response template or a Salesbot sequence.
The Threshold column determines the percentage of required similarity for the question asked by the client to be similar to those you have in the dataset. If the percentage of similarity is met, the action will be triggered.
E.g: your client is asking if they can buy a pair of rollerblades. Your dataset has no question with “rollerblades” in it, but the question is similar in structure to other questions you have (“can I buy rollerblades” as opposed to “can I buy a bicycle” present in your dataset). With the 100% threshold setting, the question has to be precisely what you have in your dataset in order to trigger the action. With a lower percentage set, the system will trigger the action if the question meets the set level of similarity.
The Times launched column is a simple indication of how many times this particular intent has been detected and action was triggered as a result. It displays a total number of launches since the Intent was added.
You can also enable Mark messages unread function, which will make the conversation as unread even if the Salesbot has responded to the client, as well as set the shutdown time for the chatbot if a human representative takes over.
The Phrases to ignore can help you create a dataset of questions that will not trigger the action regardless of other set intents.
Click on the Phrases to ignore button in order to add questions to the dataset manually, or click the … button in order to upload a file with the questions.
How to use AI power-up
As an amoCRM user, there isn’t much you need to do once the AI power-up are set up.
During the conversation with you (on any messenger channel), AI power-up will automatically respond if there is no representative active in the chat at the moment.
Be sure to visit the Settings section from time to time and clicking Update dataset in order for the system to synchronize more questions into the database.
Not a user yet? Visit our tour page or contact us for a demo.