"

8.1 Summarizing Skills

A summary is meant to inform your reader—who has not read a text—of what the text is about. It provides an objective, condensed (shortened) description of the content of the text. It describes its purpose or main idea, and includes the supporting arguments that develop that idea. Unlike a review, it does NOT analyze, evaluate, or critique; your opinion of the work is not typically part of the summary (unless you have been asked to add your thoughts afterward). The summary often forms the foundation upon which a critique or new ideas rely. Following a good summary, a reader will then know if they will find the information useful and if reading or viewing the original is worthwhile.

Knowing how to summarize is the first step in being able to ethically make use of research information in your business reports. Whether you are writing a proposal, recommendation, audit, or feasibility study, research information must be conveyed in an accurate and objective manner. Since summaries usually occur within a specific context (e.g., part of your report), your thoughts about what you have summarized will probably follow in an analysis. But when writing the actual summary of someone else’s ideas, you must neutrally and accurately describe what you take to be the important ideas in the author’s or presenter’s work in as few words as possible. Occasionally, if the work you are summarizing has an unusual form, style, or tone that affects the content, your summary might describe HOW the author presents those ideas.

In business contexts, you will find the following types of summaries, which serve different purposes:

  • A routine summary of someone else’s ideas, in the context of a report, helps you to support and develop your ideas. You may summarize someone’s ideas because they support your own, or because they differ from yours and allow you to introduce the idea you want to argue. Someone else’s theory may provide a framework for your analysis, so you might summarize the theory before beginning your argument. A summary can act as a springboard to launch your ideas
  • An executive summary of a long or formal report that you have created informs the reader of the key points of the report. In this case, you will summarize all the main ideas, including the conclusion (and recommendations), of your own report. Executive summaries are written by the report authors, and thus do not use signal phrases.
  • An abstract, written by the author(s) of a published paper, describes the content and purpose of the report and is included at the beginning of the document. An abstract is pretty formulaic and consists of the following types of statements: problem, purpose, methods, findings, next steps or implications. Abstracts are written by the authors, and  do not use signal phrases.
  • Conclusion: Appears at the end of a report; reiterates key ideas discussed and draws inferences based on the discussion or findings.
  • Briefing notes are often written by government workers to give the busy minister a summary of important information needed for a meeting or for a decision.

Being able to create a clear and useful summary is a valuable skill both in academic and professional contexts.

 

Knowledge Check

Should I Use an LLM?

Summaries can be created through a careful reading of the original text, or it can be quickly created through the use of an LLM. The method you choose will depend on whether or not you need to have a good understanding of the entirety of the original text.  Consider the following factors before you decide to use an LLM to summarize a text:

  • Do I need the depth of knowledge contained in the original text? Nothing replaces the knowledge you can gain from reading an original text from start to finish. By reading the text, you are developing foundational and critical knowledge about a topic that increases your expertise in a subject. Whether you agree or disagree with the ideas presented, you are developing important critical thinking skills that are essential for making sound differentiations and decisions.
  • Can I sharpen my argumentation skills by reading the original text? Not only are you learning important information when you read the original text, but you are also following a flow of ideas that supports a thesis and makes an argument. There is no better way to learn how to organize ideas and draw on evidence and other forms of support than through the texts of accomplished writers.
  • Will I have to report on complex ideas and concepts? If you do not read original texts and learn the concepts, you will not be able to report on the information gathered in any meaningful way. While an LLM can give you the key ideas, it will not convey all the nuances of meaning and detail. In situations when you really need to relate important subtleties about a subject, it’s best to summarize the text yourself.
  • Am I 100% sure that the LLM will produce a completely accurate summary? The only way to know if an LLM has produced a completely accurate summary is for you to know the details of the original text and to do that, you have to read through it yourself.

Creating a Routine Summary: The Traditional Method

Writing summaries can seem to be a daunting task, but much of the work can be done throughout your research, as long as you take notes as you are reading through the texts. Following the basic method described in the video and text below will enable you to develop a process that will with time become routine and enable you to make effective use of your time as you gather key information.

For an overview of summary writing strategies, review the following video, How to Write a Summary (2021).

 

 

The following stages in summarizing will guide you through the process.

Pre-writing Stage

Before you summarize, you must gain a good understanding of the original text and do some pre-writing in order to avoid getting bogged down in the small details and neglecting to present the main idea. The pre-writing stage helps you to gather information to connect the thoughts into a meaningful whole.

  1. Actively read the text. Make notes. Make sure you understand what you are summarizing: What is its main purpose? What is the “thesis”? What are the main points that support the thesis? Explain it verbally to someone else or in free writing. Use your own words to make sure you really understand what you have read or seen.
  2. Reread the text and break it up into sections or “stages of thought.” This would apply for longer original texts. Breaking the text into parts or segments helps to manage the summarizing process. Briefly summarize each section and indicate how it relates to the main idea by rephrasing key ideas. Omit the finer details.
  3. Keep your purpose and intended audience in mind when you design your summary; remember, your intended reader has not read the text. Why are you summarizing it? Why is your audience reading your summary?

Writing Stage

Now you are ready to begin writing your summary. Follow these steps:

  1. Provide the author’s name and title of the text being summarized. If the context is important to your summary, give some details about the intended audience, etc.
    • In Daring Greatly (2012), Brené Brown explores the pressures created by relationships in which “never enough” dominates . . .
  2. Paraphrase (write in YOUR OWN WORDS) the author’s THESIS or main idea:
    • … and reveals how it impacts negatively people’s ability to be vulnerable, take risks, and succeed.
    • About Online Paraphrasing Tools: Online paraphrasing tools, often referred to as spin bots, are being used by many students to switch up the language of an original text in order to avoid plagiarism. It is highly ill-advised to use such a tool. The text that the machine creates is often nonsensical due to a haphazard use of vocabulary that is devoid of context. You are strongly advised not to use such tools.
  3. Describe, in a neutral and objective manner, how the author supports and develops the main idea. Do not editorialize (evaluate, critique, analyze, etc.); simply describe as if you were a reporter. Keep the following in mind:
    • Summarize the key points used to develop the main idea by rephrasing them in your own words.
    • Leave out details and examples that are not important to the main idea.
    • Do not quote from the article, or limit quotations to a single keyword or important phrase. Padding your summary with quotations is not an accepted method for summarizing.
    • Do not accidentally plagiarize. Use signal phrases, such as “Brown explains” and “Brown asserts” to show that the ideas are not yours, but that they come from the article you are summarizing.
    • Cite and document your source using the APA method; e.g., (Brown, 2012).
    • Keep your summary at about 1/3 the length of the original.
  4. Pay attention to verb tense: Summaries of ideas are generally given in the present tense, while results and findings are often given in the past tense.
    • Brown explains …  (present tense)
    • Hollander’s study found that …  (past tense)

Summaries of presentations are generally given in the past tense since the presentation happened only once in the past, while a text can be read and reread several times, making it more “present.” However, a video presentation, such as a TED Talk, would likely be summarized in the present tense, much like an article, because it can be reviewed over and over again. The verb tense you should use is not subject to absolute rules; you will have to use some judgment to determine what sounds best (and what sounds awkward).

Sample Reference

Brown, B. (2012). Daring greatly: How the courage to be vulnerable transforms the way we live, love, parent, and lead. New York: Avery/Penguin Random House.

Rewriting Stage

Review and revise your draft using the following steps:

  1. Revise content and organization: Is it complete? Should you add any key ideas? Is it well organized? Is the content accurate? Does it follow the order of ideas of the original text?  Can you get rid of any unnecessary content? Have you used your own words and phrasing?
  2. Edit for flow:  Do ideas flow smoothly together? Are sentences clear, concise, correct and coherent? Or do they require effort to decode? Do transitions effectively indicate the relationships between ideas? Have you effectively introduced, developed and concluded?
  3. Proofread: Look for mechanical errors (typos, spelling, punctuation) and for grammar and usage errors that may have crept in during revision and editing.

 

Knowledge Check

Signal Phrases

Signal phrases allow you to clearly indicate when words, phrases and ideas you include in your writing come from someone else:

In Communication Essentials for Business, Robin L. Potter asserts that. . .

Such signal phrases include verbs (in bold above) that introduce summaries, paraphrases, and quotations. In general, it is best to avoid verbs like

  • says (too vague)
  • writes  (too vague)
  • talks about (too informal)

Instead, use a verb that more precisely and accurately describes the author’s rhetorical intention — describe what the author is DOING in this quotation, or what rhetorical purpose the author is trying to achieve. Chapter 8.2 contains a useful table of Signal Verbs for various purposes.

“I Can’t See the Forest for the Trees”

A summary should move from a statement of the general subject and purpose to the specific ideas used to develop that purpose; it should be neither too vague nor too specific. There is an expression: “I can’t see the forest for the trees.” It means you get too focused on the details, so you miss the “big picture.” Avoid being too general or too detailed. Rather, give an accurate description of the forest as a whole, and quickly go over the main characteristics of the types of trees that comprise it (the key examples used to illustrate the main idea). Don’t let your summary get bogged down in the details, specific examples, and precise data (the species of fungus on the leaves of the trees).

See examples of paraphrases and summaries on the website for the Online Writing Lab at the University of Purdue.

 

Knowledge Check

Creating a Routine Summary: AI-Assisted

An LLM can create a useful summary in seconds from a browser and within research tools like Elicit. While Elicit will automatically create a general article summary for you, when using Copilot, however, you must input a prompt and provide a text that the LLM can summarize. The quality of the summary will depend on your prompting strategy. A weak prompt will result in a very general and potentially inaccurate summary; whereas, a carefully worded prompt will result in a better summary as characterized by density of information. You can also request a summary for a document you upload or one that is on a webpage. Below you will find information on how to work with Copilot to create summaries.

Prompting Strategies for Summaries

Researchers have realized that LLMs will provide very generalized summaries that contain little concrete information unless you use a prompt that specifies that you would like the summary to be dense with information. You can do this in two ways: by using a simple prompt or a Chain of Density prompt, both of which are discussed below.

Simple prompt: You can create a simple prompt for Copilot like the one below to elicit a summary that contains the key details of the original. The request to “apply density” is to ensure that the output contains specific and concrete information.

Sample Simple Prompt and Draft Output for a Summary

Prompt used in Copilot on February 13, 2025:

Please create a specific summary of Chapter 2.6 on the adjacent web page. Apply density and ensure that the summary is at a university level. Include all key points and format the summary as a paragraph. Use signal phrases to mention the original author of the chapter occasionally in the summary.

Draft Output: 

In Chapter 2.6, Robin L. Potter explores the integration of generative AI (genAI) into workflows, emphasizing its dual role in cognitive and material processes. Potter explains that traditional workflows involve steps like gathering information, planning, and revising, but genAI can simplify or complicate these tasks depending on its application. The chapter defines a workflow as a sequence of steps to complete a task, highlighting that genAI fits into both cognitive tasks (e.g., ideation) and material tasks (e.g., formatting). Potter notes that while some companies have integrated genAI systematically, many employees adopt it individually, often within organizational constraints. The chapter discusses “workflow thinking,” which helps situate genAI in the co-creation process. Potter outlines three approaches to using AI: “Just Me Tasks” for human-only tasks, “Delegated Tasks” for tasks better handled by AI, and “Automated Tasks” for routine tasks. The chapter also presents two use cases: Damon, who minimally engages with AI, and Paige, who actively integrates AI while maintaining control. Potter emphasizes the importance of evaluative judgment and intentionality in using genAI, suggesting strategies like learning about quality standards, revising AI output, and inviting peer review to ensure high-quality co-created content.

The summary produced with this prompt is informative in that it includes most of the key ideas expressed in the original text. It effectively signals the original author (Potter) of the chapter; however, it does not mention Mollick, who is referenced as the source for the task types. The user (human) would have to clarify the source for the information in the summary revision. In addition, if using Copilot, you must specify that the summary must be formatted as a paragraph; otherwise, it will be presented in point-form, which is not the typical way a summary is formatted. Clearly, the summary created by Copilot could not be used as is, which is where human co-creation becomes necessary: your choice is either to revise the one given, or to refine your prompt and work with the LLM continuously until you achieve a version that is more suitable.

Chain of Density prompting requires that the LLM engages in repeated or iterative summarizing such that it self-identifies the gaps in the information and repeats the attempt for the specified number of times. Drawing on a study by Adams et al. (2023), Ethan Mollick (2024a) shared the following Chain of Density prompt and when used to summarize the same text as above the results are as depicted below:

Sample Chain of Density Prompt and Draft Output

Chain of Density prompt (Mollick, 2024b), slightly adapted for use with Copilot Chat (sidebar), February 11, 2025:

Please generate increasingly concise, entity-dense summaries of the article on the adjacent web page.

Repeat the following 2 steps 5 times.

Step 1. Identify 1-3 informative entities (“;” delimited) from the article which are missing from the previously generated summary.

Step 2. Write a new, denser summary of identical length which covers every entity and detail from the previous summary plus the missing entities. A missing entity is: – relevant to the main idea, – specific yet concise (5 words or fewer), – novel (not in the previous summary), – faithful (present in the article), – anywhere (can be located anywhere in the article).

Guidelines:

  • The first summary should be long (4-5 sentences, ~80 words) yet highly non-specific, containing little information beyond the entities marked as missing. Use overly verbose language and fillers (e.g., “this article discusses”) to reach ~80 words.
  • Make every word count: rewrite the previous summary to improve flow and make space for additional entities.
  • Make space with fusion, compression, and removal of uninformative phrases like “the article discusses”.
  • The summaries should become highly dense and concise yet self-contained, i.e., easily understood without the article.
  • Missing entities can appear anywhere in the new summary.
  • Never drop entities from the previous summary. If space cannot be made, add fewer new entities.
    Remember, use the exact same number of words for each summary. Answer in JSON. The JSON should be a list (length 5) of dictionaries whose keys are “Missing_Entities” and “Denser_Summary”.

*Reformatted with bullet points to increase readability for this text.

Draft Output: 

{ “summaries”: [ { “Missing_Entities”: “workflow; genAI; LLMs”, “Denser_Summary”: “This article discusses the concept of workflow and how generative AI (genAI) and large language models (LLMs) are integrated into it. It explains the traditional workflow steps and how genAI can simplify or complicate these processes. The article also highlights the importance of understanding one’s workflow to effectively incorporate LLMs. Additionally, it touches on the varying degrees of genAI integration and the need for evaluative judgment to maintain human agency in the co-creation process.” },

{ “Missing_Entities”: “cognitive; material; McKinsey survey”, “Denser_Summary”: “This article discusses the concept of workflow and how generative AI (genAI) and large language models (LLMs) are integrated into it. It explains the traditional workflow steps and how genAI can simplify or complicate these processes. The article also highlights the importance of understanding one’s workflow to effectively incorporate LLMs. Additionally, it touches on the varying degrees of genAI integration and the need for evaluative judgment to maintain human agency in the co-creation process. It differentiates between cognitive and material aspects of workflows and mentions a McKinsey survey on genAI adoption.” },

{ “Missing_Entities”: “Lockridge & Van Ittersun; enterprise LLMs”, “Denser_Summary”: “This article discusses the concept of workflow and how generative AI (genAI) and large language models (LLMs) are integrated into it. It explains the traditional workflow steps and how genAI can simplify or complicate these processes. The article also highlights the importance of understanding one’s workflow to effectively incorporate LLMs. Additionally, it touches on the varying degrees of genAI integration and the need for evaluative judgment to maintain human agency in the co-creation process. It differentiates between cognitive and material aspects of workflows and mentions a McKinsey survey on genAI adoption. The article references Lockridge & Van Ittersun’s work and the use of enterprise LLMs.” },

{ “Missing_Entities”: “secretive use; organizational policies”, “Denser_Summary”: “This article discusses the concept of workflow and how generative AI (genAI) and large language models (LLMs) are integrated into it. It explains the traditional workflow steps and how genAI can simplify or complicate these processes. The article also highlights the importance of understanding one’s workflow to effectively incorporate LLMs. Additionally, it touches on the varying degrees of genAI integration and the need for evaluative judgment to maintain human agency in the co-creation process. It differentiates between cognitive and material aspects of workflows and mentions a McKinsey survey on genAI adoption. The article references Lockridge & Van Ittersun’s work and the use of enterprise LLMs. It also discusses the secretive use of LLMs and the impact of organizational policies.” },

{ “Missing_Entities”: “productive struggle; environmental costs”, “Denser_Summary”: “This article discusses the concept of workflow and how generative AI (genAI) and large language models (LLMs) are integrated into it. It explains the traditional workflow steps and how genAI can simplify or complicate these processes. The article also highlights the importance of understanding one’s workflow to effectively incorporate LLMs. Additionally, it touches on the varying degrees of genAI integration and the need for evaluative judgment to maintain human agency in the co-creation process. It differentiates between cognitive and material aspects of workflows and mentions a McKinsey survey on genAI adoption. The article references Lockridge & Van Ittersun’s work and the use of enterprise LLMs. It also discusses the secretive use of LLMs and the impact of organizational policies. The article emphasizes the importance of productive struggle and considers environmental costs.” } ] }

*A line space has been inserted between attempts to improve readability.

This second attempt produces five versions of a condensed summary that conveys some subtle nuances of meaning, but it is not as comprehensive or complete as the previous example and the ideas do not follow the flow in the original text. It also contains noticeable gaps in key information. It is interesting to note the “thought” process of the LLM as it identifies key areas that should be included but are missing and how it integrates the new information in subsequent iterations. The user (aka human) would have to revise the contents to differentiate between key ideas and supporting information and make decisions about what to retain and what to leave out. The user would also have to either ask for key points to be added or add these themselves. For example, reference to the McKinsey study offers interesting information, but it is not key information for the summary, yet does not mention the important machine-in-the-loop and human-in-the-loop distinction. Regardless of the flaws, the LLM can offer a draft summary that can become a good start for one that is completed by the user who is thoroughly familiar with the original text or reworked with the LLM using refined prompting.

Using Copilot to Create a Summary

You can use two methods to create a summary with Copilot: by uploading a text or by opening Copilot Chat (sidebar) on your browser so it can scan and summarize a text in a browser window.

  1. Uploading a text: Copilot will provide a summary upon direction when you upload a document as a reference text into the context window. Use the “Add a File” paper clip feature to load your document. Or, if it is a short text, you can copy and paste the document into the context window. Your instructions should be specific as to what you would like the LLM to do and how it should go about it. The LLM will work best when you upload or paste your document and ask it to read the text. Then you can load your prompt asking it to summarize the text.
  2. Using Copilot Chat (sidebar): Copilot Chat will create a summary of text on an adjacent web page. It offers a convenient “Create a Summary” button that you can use to prompt the LLM, but you can also create your own more specific prompt instead. Copilot Chat does not allow for document uploading. To access Copilot Chat, you need to have the Copilot icon loaded to your browser window. If you are using the Edge browser, you can do this by following these steps:

    Loading the Copilot Chat Icon to the Browser Toolbar

    1. Click on the three dots (…) in the upper right side of the browser.
    2.  Scroll down to Settings.
    3. Key in “Apps” in the search window.
    4. Scroll way down and click on Copilot.
    5. Select “Show Copilot on the toolbar” and “Allow Copilot to use page content.” You may have to shut down the browser and open it up again to see the Copilot Chat icon on the toolbar.

    To use Copilot Chat, navigate to the webpage you would like summarized, and click on the Copilot Chat icon on the toolbar. The Chat will open up in the sidebar.

     

Declaring Your Use of an LLM

When making use of an LLM to co-create a summary, you should include a brief declaration of LLM use following the completed summary. Include the following information:

(Co-created with Copilot, [date])

Doing so signals your transparency and will help to boost your credibility. It is also important information for the reader as it helps them to make a judgement about the summary. In some cases when the information is critical in making decisions, the reader may go to the original text to verify the accuracy of its contents. Keeping documentation of the prompt(s) used along with various versions of the draft summary is also good practice.

Summarizing Tips for LLM Co-Creation

As with any content created by LLMs, a summary created by an LLM requires careful review in which you apply your critical thinking skills. As seen with the examples provided above, the work is not done once the output is provided. You must carefully review the summary for accuracy and reformat the text as needed so it conforms to the conventions of the genre. Here are a few additional suggestions:

  • Mention the format in which you want the summary written: Copilot is inconsistent, as we have seen in the examples above. Mentioning in your prompt that you would like the summary in paragraph format would save you the step of reformatting the summary.
  • Mention the required length: Typically, summaries are 1/3 the length of the original text. Specify that in your prompt.
  • Ask the LLM for density of information: Copilot in particular tends to produce general output. For more specific and information-heavy summaries, ask for density.
  • Ensure the original author is mentioned and signalled: To avoid plagiarism issues, be sure that the summary includes information about the article’s original author and signal that authorship.
  • Check that only essential research sources are mentioned: If key ideas are based on research sources, ensure those authors are signalled in the summary. If these pertain only to unessential details, remove them as they are extraneous to the key idea of the summary.
  • Ensure an objective tone: Summaries must be conveyed in a neutral and objective way, so be sure that the text is written using this tone.
  • Ensure that all key ideas are included: The LLM will produce a draft. It is up to the user to ensure that the summary faithfully represents the key ideas of the original text.

Mastering the art of summarizing is essential for both academic and professional success. Understanding the original text thoroughly before attempting to summarize it is the first step in summarizing. Key strategies include active reading, note-taking, and paraphrasing to ensure comprehension and accuracy. Different types of summaries, such as routine summaries, executive summaries, and abstracts, each serve distinct purposes. A key component of every summary is the signal phrase, which attributes ideas to their original author. In addition to creating summaries in the traditional way, LLMs can be helpful in completing the task. By following guidelines provided, you can create clear, concise, and accurate summaries that effectively convey the main ideas of the original texts. (Note: This conclusion was co-created with Copilot Chat on February 12, 2025).

References

Adams, G., Fabrri, A. R., Ladhak, F., Lehman, E., and Elhadad, N. (2023).  From sparse to dense: GPT-4 summarization with Chain of Density prompting. arXiv 2309.04269

Brown, B. (2012). Daring greatly: How the courage to be vulnerable transforms the way we live, love, parent, and lead. New York: Avery/Penguin Random House.

Mollick, E. (2024a).  How to boost AI summaries with Chain of Density | Ethan Mollick posted on the topic | LinkedIn

Mollick, E. (2024b). Chain of density prompt. Generate Summaries from Articles. ChatGPT.

(Online Writing Lab (OWL). Paraphrase and summary exercises. University of Purdue. https://owl.purdue.edu/owl_exercises/esl_exercises/paraphrase_and_summary_exercises/index.html

Smrt English. (2012). How to write a summary [Video]. Youtube. https://www.youtube.com/watch?v=eGWO1ldEhtQ

License

Icon for the Creative Commons Attribution 4.0 International License

Communication Essentials for Business Copyright © 2019 by Suzan Last (Original Author) Robin L. Potter (Adapter) Tricia Nicola Hylton (H5P) is licensed under a Creative Commons Attribution 4.0 International License, except where otherwise noted.

Share This Book