Researching company information - This action can be used to research information about companies, such as their products, services, financials, employee count, and competitive landscape. It is described as one of the easiest and most common go-to-market use cases for this action. You can provide an open-ended research question like "What products does Company X offer?" or "How many employees does Company Y have?" and the action will find and return relevant information and citations to answer the query.

Researching a topic to write long-form content - This action can be leveraged to research specific topics or elements that need to be incorporated into long-form written content like reports, articles, etc. For example, if you are writing a brief that requires discussing a particular topic in-depth, you can use this action to research that topic and then chain the output to a "Generate Text" action to incorporate the researched information into your long-form content.

Researching topics to generate an account plan - In a more advanced use case, you can use this action to research a variety of topics related to a particular account like industry trends, competitive landscape, company details, employee counts, etc. The aggregated research output can then be used to generate a comprehensive account plan with an informed perspective on the account.

Detailed question or a research statement or topic - This is the main input required for the "Research a Topic with AI" action. It should be an open-ended question or statement describing the specific topic you want researched.

Model selection - This allows you to choose which specific language model will be used behind the scenes to actually conduct the research and generate the answer. There are two models currently available: Perplexity Sonar Large and Perplexity Sonar Small. Typically, Sonar Large is smarter and take longer to run. Sonar Small is faster, and is good for lighter weight research tasks.

No answer or poor answer returned - This issue typically arises due to limitations of the model being used or the way the question is framed. If the model does not have sufficient information in its training data to answer the query, or if the query is not framed in an open-ended way that allows the model to effectively search its knowledge, it may return a poor or no answer. To mitigate this, try rephrasing the query in a more open-ended manner or provide a specific response for the model to return if it cannot find a satisfactory answer.
​
For handling errors, it is helpful to provide an alternative response for these scenarios: {{Your Research Topic or Question }} If no answer is available please respond with N/A

Research citations no longer available - The research action provides citations or URLs as supporting evidence for the answer generated. However, since the model providers rely on a snapshot of the internet rather than live web access, some of these citations may become invalid over time if the original web pages are moved or deleted. While the citations were accurate at the time of training, they can become stale. It's advisable to validate important citations, especially if you intend to rely heavily on them.

Research response not structured as needed - The research output provides the answer and citations in a basic JSON format. However, if you need the response structured differently for your use case, you have a couple options. You can try providing guidance on the desired structure in the query input similar to how you would construct a prompt. Alternatively, you can chain the research output to other actions like Generate Text, Categorize Text, or Extract Data from Text to post-process and reshape the raw response into the required format.

Perform Internet Search - Performs a basic Google search and returns a list of web links related to the search query. It does not provide any summarized information or answer. Whereas Search just returns web links, while "Research a Topic with AI" actually attempts to provide a direct answer by summarizing information from multiple sources.

Research Agent - An advanced AI agent that conducts in-depth research on an open-ended topic and provides a comprehensive report. It uses an agent-based approach to gather information from multiple sources and synthesize the findings into a broad analysis.Research Agent aims to provide an extensive report covering many angles of the topic, going beyond a simple summarized answer.

Research Statistic Agent - A specialized research agent focused on finding statistics related to a given topic. It is optimized to extract and present relevant statistical data from various sources. Research Statistic Agent is laser-focused on locating and presenting statistics, whereas "Research a Topic with AI" covers general information on the topic.