The WorkWithPlus AI Assistant will determine the best way to answer the end-user messages to fulfill their needs.
When the end-user asks for some information in which the more adequate answer is in Natural Language, the WorkWithPlus AI Assistant will let the application about it.
The following steps occur internally in the WorkWithPlus AI Assistant for answering a text in natural language:
- The AI Assistant will send the LLM metadata about the Knowledge Base (available Transactions, attributes, and filters) together with the end-user message.
- The LLM will return the query to get the data needed by the end-user ('For each' Transaction, filters, sorting, etc.).
- The AI Assistant will run the query and get the result.
- The AI Assistant will send the result to the LLM service to get the answer in Natural Language.
- The answer message is shown by the AI Assistant to the end-user.
Example
"Who is the youngest person registered in the system?"
The answer to the end-user will be:
Another way of displaying the result of an end-user message is in a data table. This will be used for displaying for example more than one record and some information about each one.
The following steps occur internally in the WorkWithPlus AI Assistant for answering with a visualization of Data Table:
- The AI Assistant will send the LLM metadata about the Knowledge Base (available Transactions, attributes, and filters) together with the end-user message.
- The LLM will return the query to get the data needed by the end-user ('For each' Transaction, filters, sorting, etc.).
- The AI Assistant will run the query and get the results.
- The AI Assistant will send the result to the LLM service to get the answer in a Data Table.
- The answer message is shown by the AI Assistant to the end-user.
Example
"which are the top 5 most expensive projects? Can you display a table with the info?"
This type of visualization is useful when the answer is a numeric value, like a sum, average, etc.
The following steps occur internally in the WorkWithPlus AI Assistant for answering with a visualization of KPI:
- The AI Assistant will send the LLM metadata about the Knowledge Base (available Transactions, attributes, and filters) together with the end-user message.
- The LLM will return the query to get the data needed by the end-user ('For each' Transaction, filters, sorting, etc.).
- The AI Assistant will run the query and get the result.
- The AI Assistant will display the KPI with the information obtained.
Example
"Show me the total price of all projects"
The AI Assistant supports three types of charts: Timelines, Ranking, or conventional charts (Pie, bar, etc.). The chart used for the response will depend on the end-user's need and will be determined automatically by the AI Assistant. The end-user can also include in the message the type of chart to display its information.
The charts support displaying multiple series (up to 10 series for a chart).
The following steps occur internally in the WorkWithPlus AI Assistant for answering with a visualization of a chart:
- The AI Assistant will send the LLM metadata about the Knowledge Base (available Transactions, attributes, and filters) together with the end-user message.
- The LLM will return the query to get the data needed by the end-user ('For each' Transaction, filters, sorting, etc.).
- The AI Assistant will run the query and get the result.
- The AI Assistant will display the chart with the information obtained.
Timeline Example
"I want to see the price of all projects (no matter their status) over time by start date"
Then, the end-user asks to add another series to the same chart (by taking advantage of the contextual chat)
"In that same chart I want to see also their real cost"
Ranking Example
"which are the top 5 most expensive projects?"
Pie Example
"Can you show me the distribution of projects by status?"
Bar Example
"Can you show me the amount of people born in each decade from 1960?"
The AI Assistant has also the capability of redirecting the end-user to a list with instantiated filters or to a detail of some record. This capability is the same feature provided by the AI Intent-based navigation but inside the AI Assistant.
|