Skip to main content
Interacting with Generative Search

Get answers and insights using natural language with Generative Search

Mark Jones avatar
Written by Mark Jones
Updated over a week ago

NOTE: AlphaSense Generative Search is currently available through our limited private beta. If you have any questions or would like to inquire further, please reach out to your Account Manager.

Ask Gen Search in real time with natural language in a conversational experience, and it will formulate an answer that utilizes all key market perspectives that make AlphaSense's content library so valuable.



Generative Search Interactions

Gen Search is purpose-built to help you get up to speed on topics quickly, by answering questions like:

Type

Example

Company Information & Metrics

What is Caterpillar's AI Strategy?

What was Meta's 2024 guidance?

Segment Information

How did Apple's wearables segment perform last quarter?

Industry Information

How has shrink impacted the Dollar Store industry since Covid?

Product Information

What is the commercial potential for oral GLP1?

Screener

Who are the players in biologics?

Macroeconomic Impact

What impact would an EU oil embargo have on crude flows?

Change Questions

"How has Google's smartphone market share changed over time?"

Comparisons

"Compare Apple and Google's approach to Gen AI"

From any query asked, Gen Search will leverage natural language to extract relevant parameters from your question in order to run the most effective search. The parameters that can be extracted are:

  • Keywords and/or Topics

  • Companies

  • Industries

  • Regions

  • Timeframes

  • Source Documents and/or Voices

    • Broker Research

    • Company Docs

    • Expert Insights

    • News

    • Regulatory

    • Patents

    • Internal Content

As with all of AlphaSense's AI offerings, Generative Search will cite its sources. AlphaSense will not only cite the snippets that it referenced while constructing a response, but we also conduct a subsequent evaluation to confirm that each cited snippet contains supporting information.

Gen Search will also tell you when it doesn't have an answer. Rather than hallucinate an answer, it will let you know when it cannot find any relevant content within your AlphaSense library to generate an answer to your question.

If you want to discover how to access Generative Search, take a look at our "Accessing Generative Search" article.


Structure of a Response

After asking a question to Generative Search, you'll see the question you asked (and its corresponding plan), the content summarized with citations, a conclusion/summary, and related questions.


Question asked and corresponding plan

At the top of Gen Search's response, you'll find the exact question you asked along with the ability to view the plan created.

AlphaSense Summary/Planning Agent

In order to provide complete and concise answers to very complex questions, AlphaSense utilizes a "Planning Agent" to break down these complex queries into multiple questions. These multiple questions are then used to retrieve the correct information before AlphaSense analyzes the information to create a single, concise response.

Selecting "See Summary Steps" allows you to see what questions the Planning Agent broke the original query into.


Response & Citations

Generative Search will structure its response to your question, providing an organized overview of the major takeaways.

Each summary will provide a direct citation to the source material it was summarized from. Simply select a summary and the Keyword Hits pane and document preview will open alongside a condensed Gen Search view. This allows you to conduct further exploratory research, while also retaining the ability to ask follow up questions to Gen Search.

Depending on the response, Gen Search may compile a Conclusion/Summary related to your question.


Regenerating & Sharing

At the end of the Generative Search response, you'll have the ability to generate a new response, share the Summary, and dictate if the response was Unhelpful.

Generating New Response (1)

You can generate a new response if you'd like Generative Search to create a new answer to your original question. You'll likely generate a new response if:

  • You want more summaries to reference/look at

  • It didn't adequately answer your question and no rewording/rephrasing is needed on the original question.

Share Responses via Clipboard (2) or Notebook (3)

Easily share responses wherever you need them! Select "Share" and either copy to your clipboard (and paste them in a document, presentation, etc) or directly add them to any Note within AlphaSense.


Related Questions in Generative Search

At the bottom of a response, you'll find several related questions that can be used to immediate ask a follow up question related to your initial ask, or can be used as inspiration for follow up questions.

You also have the ability to ask any follow up question you'd like directly from chat bar.


Best Practices when Interacting with Generative Search

Asking Questions

Best practices if Gen Search didn't answer your question

  • Generate a new answer - at the bottom of the response, you can have Gen Search answer your same question without adjusting your original question

  • Adjust your question:

    • If it provided an answer but it was focused on the wrong content sets, we'd recommend selecting "Clear Chat", starting a new Generative Search session OR launching Gen Search after conducting a traditional AlphaSense search and setting your filters manually

      • This will ensure you're searching against all content available to you in your AlphaSense account

      • If you're wanting to hone in on a specific content set, include that content set/market perspective in your question (ex: "What does broker research say...", "...within PubMed articles?", "How do former competitors think about..."

    • If Gen Search provided an answer but it was focused on the wrong companies or keywords, we'd recommend altering your question and being more specific.

      • It's best to include tickers and use words that speak directly to a given topic (ex: searching "electric vehicle industry" is more optimal compared to "electric vehicle space").

      • It's also ideal to use keywords that activate Smart Synonyms whenever possible to optimize results

Adjusting Timeframe

By default, Gen Search will index the last 12 months. If you'd like to change the timeframe it will search against, use natural language in your question, such as "...in the last 3 years", "In Q2 of this year...", and "...last month..."

Setting Generative Search's Content

Check out our best practices around Prompt Engineering to learn more about setting Gen Search's content and example prompts

Follow Up Questions

  • Conversational Memory: Generative Search does maintain conversational memory. If you ask "What are the positives around Nividia's supply chain?", you're follow up question can speak indirectly to the first question. Example "What about the negatives?"

  • Content context is not carried over. If you're wanting to isolate a specific content set in a follow up question, you'll need to specific it.

    • Example: When asking a follow up question to "How are brokers speaking positively about Rivian's R&D?, instead of asking "What about the negatives", ask "What about the negatives from brokers?"


FAQ

How does Generative Search work?

Gen Search uses a number of natural-language process and machine learning techniques to interpret questions, retrieving all relevant information, and reason over the data to analyze/summarize findings.

It starts by interpreting what you're asking - extracting relevant parameters from your question to run the most effective query.

Then, it retrieves the top 200 snippets for that query from documents across AlphaSense relevant to your search. These snippets are then used by our large language model to generate an answer.

What content in AlphaSense can Generative Search use to reason over?

Gen Search can source it's answers from all content that's available to you in your AlphaSense account.

How does Gen Search tackle complex questions?

With the Summary/Planning Agent, Gen Search can break down questions into multiple queries to retrieve the right information before then analyzing those queries to produce a single, concise answer.

Gen Search didn't answer my question - what should I do?

If it didn't answer your question, you have a few options

  • Generate a new answer - at the bottom of the response, you can have Gen Search answer your same question without adjusting your original question

  • Adjust your question:

    • If Gen Search provided an answer but it was focused on the wrong content sets, we'd recommend selecting "Clear Chat", starting a new Generative Search session OR launching Gen Search after conducting a traditional AlphaSense search and setting your filters manually. This will ensure you're searching against all content available to you in your AlphaSense account

    • If Gen Search provided an answer but it was focused on the wrong companies or keywords, we'd recommend altering your question and being more specific (include tickers, remove keywords that might have drastically different meanings (ex: "space" can refer to industries or aeronautics)

What questions does Generative Search not answer well?

  • Screening - while Gen Search will attempt to answer questions like "What companies are working on Generative AI?", it doesn't yet provide a comprehensive list. We recommend using the traditional AlphaSense search experience (via 'Company in Search Results' module) until this skill is added to Generative Search

  • Calculations - while Gen Search will answer questions like "What is Apple’s revenue growth trend over the last 5 years?" by looking for explicit discussion on the topic, it does not yet perform mathematical calculations or provide charts in its answer.

  • Librarian - Questions like ‘Summarize Meta’s last 3 earnings calls’ do not presently work well because Generative Search doesn’t have the ability to source an exact set of documents. Instead, ask something like ‘What are key positive and negative trends from across Meta’s earnings calls’

Generative Search didn't produce an answer to my question - what does that mean?

Gen Search will tell you when it doesn't have an answer: Rather than hallucinate an answer, it will let you know when it cannot find any relevant content to generate an answer. If your question is not producing responses, you might try reformatting your question.

It could be that there is no/not enough content around the topic you're searching for, which you can verify through the classic AlphaSense search experience.

Did this answer your question?