The AI Assistant can assist investigators with a range of tasks, including generating an overview of the investigation, answering specific questions about graph content, supporting data cleaning efforts, and explaining its own capabilities.
For general queries – such as the Assistant’s own capabilities – the Assistant responds straightaway without needing access to your graph.
For more advanced queries, the Assistant uses Tools – specialized sub-agents trained to perform actions such as interacting with the Maltego Graph, accessing and interpreting graph data, retrieving resources, or executing predefined tasks.
Tools include:
- Structured Graph queries
- Deep search
- Creating/updating/deleting Entities
- Investigative tactic search and recommendations
The Assistant automatically determines the most appropriate tool to use for a given task. This means no manual configuration is required from the user in most cases. However, if you know that a particular tool is especially well-suited for the task at hand, you can optionally instruct the Assistant to use it explicitly.
For any critical actions such as accessing graph data or calling resource-intensive tools, the Assistant will ask for your permission before executing the operation.
To see which tools the Assistant called to execute your task, you can enable "Show > Tool Calls” via the Assistant menu:
Click on the collapsed Tool Call boxes in the thread to view details:
General Queries
General chat queries may include asking the Assistant about its capabilities, to which the AI Assistant will respond directly.
Investigation Overview and Recommendations
In more advanced scenarios, such as any questions regarding the current investigation, or recommendations for next investigation steps, the Assistant might call tools and interact with its sub-agents. In this case, the Assistant will request your permission to access the graph.
Currently, the Assistant cannot recommend specific Transforms. However, it can draw on general investigative strategies to suggest possible next steps. These recommendations will remain high-level for now. Importantly, the Assistant does not have the capability to execute Transforms, browse the internet, or access any external data sources independently.
Deep Search
The Assistant typically triggers Deep Search automatically for responses to complex, qualitative questions or report-building. However, if it does not activate Deep Search by default, you can manually request it by including e.g. “use deep search” in your message.
Because Deep Search performs a comprehensive scan of your graph, it may take time to complete, especially on large datasets. For optimal performance, it is recommended to avoid using Deep Search on graphs larger than 3,000 Entities.
Example Prompts to Use Deep Search:
- What topics does this person post about on social media? Use deep search and give examples.
- Which of these Entities should I follow up on to understand X? (you should select the Entities on the graph prior to this request)
Data Cleaning
Assistant can use a combination of structured graph queries and its entity editing and deletion tools to filter and clean data.
Example Prompts for Data Cleaning:
- Delete all email address Entities that do not end in @xxx.com
- Find duplicates or near-duplicates on my graph.
- Outline the connection between the selected Entities. (select Entities prior to this request)
- Find all Entities that contain Cyrillic characters.
- Find all Contract Entities, convert their prices to USD and list the largest 10.