Writing Assistant Prompt Examples

Storybuilder’s Writing Assistant helps you generate memos, outlines, lists, and tables, turning your team’s documents and work product into structured, actionable content for your narrative. Whether you’re summarizing case materials, creating trial outlines, or organizing key facts, how you configure the Writing Assistant will directly impact the quality and relevance of its output. This configuration is done using prompts — specific instructions that guide your Everlaw AI Assistant.

In this article, we’ll cover the basics of prompting and explore practical examples for effectively using Writing Assistant in Storybuilder.

For a more in-depth look at the Everlaw AI Writing Assistant, visit Storybuilder and Writing Assistant.

What is prompting?

Prompting is the act of providing generative AI Large Language Model (LLM) with input such as questions, instructions, and relevant resources to factor into the output response. You may also hear us refer to this as “configuration” ot “task configuration” when we talk about Everlaw AI Assistant.

Prompting in the scope of Writing Assistant

When you write a prompt for Writing Assistant, you will use the form-like framework built into the Writing Assistant dialog. Everlaw then takes that prompt and passes it, along with additional instructions (e.g. “format the response as a list”), criteria, and information about your evidence, to an AI LLM. 

The LLM then processes the request and returns a response. When talking about Writing Assistant, we refer to this response as a generation.

How to Prompt Writing Assistant

In this section, we’ll cover the high level steps of prompting the Writing Assistant. For more complete instructions, visit Storybuilder and Writing Assistant.

Before you start:  Set a goal — what do you want your Writing Assistant’s help generating? Then consider what document sets contain the information that your Writing Assistant will need to produce the content you want. To be accessed by your Writing Assistant, these documents must be added to the Story you’re working on. 

After opening Writing Assistant from a Draft or Deposition:

1. Select the evidence: You'll be asked to choose which evidence the Writing Assistant should consider. Options include:
   • All Story evidence
   • Evidence in this Draft / Deposition
   • Evidence with a specific Story label
2. Choose a Format: Next, select the format for the AI's response. You can either use a template or start from scratch. The available formats are:
   
   • Memo: Generates a long-form document with in-line citations, organized into multiple sections and paragraphs.
   • Outline: Produces a structured framework with sections and subsections that would guide a longer Memo, each containing a brief description and relevant citations.
   • List: Creates a bullet-point list with citations for each item. Sub-bullets may be included depending on your configuration.
   • Table: Generates a table with rows and columns based on the setup you provide.
3. Provide Additional Details: Depending on the format you selected, you will be asked to supply further information, such as the desired tone and structure of the response.
   
   In some cases, the information you are asked for will be binary (e.g. analyze vs. argue). In other cases, you have the opportunity to write out more complex instructions. In such cases, you may want to consider the following qualities of well defined prompts: 

• • Clarity: A well-crafted prompt should be clear and unambiguous. The more specific and direct the prompt, the more likely the model will produce a focused and relevant output.
  • Simplicity: Concise and straightforward instructions help make for a clearer prompt and more focused output. Treat the prompt like instructions to a smart intern who is most likely to succeed with specific directions communicated in plain language.
  • Context: Providing context helps the AI understand the background or the scenario, leading to more accurate responses
  • Use of examples: Including relevant examples guides the AI towards the desired type of output.

4. Generate Content: Once your selections are complete, click Generate to let the Writing Assistant do its job.
5. Review and Refine: When the content is ready, review the results for accuracy and quality. Select Copy to a new task for an opportunity to tweak and improve your initial prompting.
   Keep in mind that your first prompt may not produce the best result possible, and this is okay. Good prompting often involves refining the prompt iteratively — making small adjustments based on the outputs until you’re satisfied with the quality of the results. All of your prompts and generations are also saved, so if you like an earlier attempt better, you can always fall back to that. 
6. Insert it into Draft or Deposition: Once you have a generation you're satisfied with, insert it into your Draft or Deposition. You can then directly edit it or use the Writing Assistant to rewrite a specific section with instructions on how you want to improve it. 

Prompt Examples

Below are four example prompts, one for each Writing Assistant format type: memos, outlines, lists, and tables.

Each section also contains a description of the format type, suggestions of how you might use the format in your work, and details about configuration options.

Memos

Prompting the Writing Assistant to create a memo results in a longform piece of writing with in-line citations. Memos consist of multiple sections, each with one or more paragraphs.

You should consider using the memo format for instances in which you want to draft long form content or arguments warranting a factual basis, such as a Motions for Summary Judgement, a statement of facts, a bench trial briefing, or a closing argument.

Memo configuration options

You should consider the following criteria when configuring a Writing Assistant prompt to generate a memo:

Memo example

Scenario: We’re working on a case alleging that major pharmacy chains contributed to the opioid crisis by failing to adequately train employees on proper opioid dispensing policies. We've collected a set of internal documents, employee statements, and corporate policies that reveal inconsistent training and insufficient oversight. We need to draft a statement, arguing that the pharmacies in question should have implemented more robust employee training programs.

To do this, let’s prompt Writing Assistant with all story evidence and the following input:

• Compose a memo
• that (tone) argues
• (substance) that major pharmacy chains contributed to the opioid crisis by failing to adequately train employees on opioid dispensing policies. 

Assessment of this example: This initial prompt is strong in its specificity and clear in its objective. Rather than saying “pharmacies,” it says “major pharmacy chains” to be more specific and avoid ambiguity between specific pharmacies (that need the training) and the big chains that are the parties at fault. It could be even better to call out the specific entities the user wants the memo to be about.

The prompt could also direct the Writing Assistant to including specific references to help anchor the memo in concrete evidence. In some cases, this is beneficial because it encourages a fact-based argument; however, we recommend initially avoiding calling out specific docs unless you’re doing it for a specific reason (e.g. if the training was laid out in specific training or policy docs). The prompt might also be improved by more clearly defining the scope of the memo. 

Avoiding direct requests for “legal standards” reduces hallucination risks, though reframing around concepts like negligence in training practices can still address relevant issues without prompting legal misinterpretation. 

Outlines

Prompting the Writing Assistant to create an outline results in content organized in a structured framework. This includes concise descriptions of what should be covered in each of the sections and subsections, as well as citations to relevant pieces of evidence. It can be a great starting point to guide you or your team writing something more in-depth. 

Consider using the outline format when you want to build out a story with compounding arguments. This can be helpful for tasks in which you might want to organize all evidence relevant to a particular issue, argument, or witness. It can be a helpful guide for when you write a longer piece of content, like a memo.

Outline configuration options

You should consider the following criteria when configuring a Writing Assistant prompt to generate an outline:

Outline example

Scenario: We’re preparing to depose a key witness from CompanyA in a case alleging improper narcotics dispensing practices. The case centers around whether CompanyA's policies aligned with DEA recommendations in the years leading up to 2013. We've gathered internal policy documents, emails, and DEA guidelines for comparison. To prepare, we want to draft an outline that analyzes the differences and similarities of CompanyA’s policies compared to DEA recommendations, year by year. 

I this case let’s prompt Writing Assistant with the following input:

• Compose an outline that
• (tone) analyzes 
• (substance) the scope of content covered in CompanyA's official narcotics dispensing policies compared to DEA recommendations. Organize this by year, and include every year leading up to, and including 2013. 

Assessment of this example: This initial prompt is effective because it clearly defines the task and specifies a comparative analysis by year, helping to guide a thorough examination of relevant evidence. However, it could be refined by clarifying the desired level of detail. For instance, specifying whether the analysis should focus on significant policy gaps, factual differences, or even legal implications could enhance clarity.

Referencing a specific official document by Bates number, such as CompanyA's narcotics dispensing policies, would reinforce this specificity. Additionally, indicating important years or events — like major regulatory changes — could help the AI prioritize critical information, making the output more focused on significant points of interest. While exploring legal implications could be beneficial, care should be taken to remain rooted in factual analysis to minimize hallucination risks.

Lists

Prompting the Writing Assistant to create a list results in a bullet-point list with relevant citations for each unique item in the list. Depending on the configuration you provide, sub-bullets may be included.

You should consider using the list format for instances in which you want to highlight key points or summarize important information succinctly during case narrative development. This can be helpful for situations such as collecting key case facts or arguments summaries.

You could also use this prompt type for brainstorming.  For example, you could create a list of unique arguments outlining why pharmacies contributed to the opioid crisis. This approach can help you identify perspectives and insights that may have been overlooked, leading to a more comprehensive analysis.

List configuration options

You should consider the following criteria when configuring a Writing Assistant prompt to generate a list:

 

List example

Scenario: We’re preparing for a case involving several prescribers who have raised concerns about opioid prescribing practices at various pharmacies. The case focuses on whether these prescribers’ concerns were adequately addressed and how those concerns may have impacted the pharmacies' opioid dispensing practices. We've collected statements from prescribers, internal communications, and relevant evidence. We want to create a list that will help us organize and present the prescribers' concerns clearly, allowing us to evaluate the significance of each in relation to the pharmacies' actions.

In this case let’s prompt Writing Assistant with the following input:

• Compose a list
• Where each item is a unique prescriber concern
• For each item, include information about Name the concern, describe the concern, name the providers who share that concern, quote some provider statements discussing their concern
• Order (left blank)

Assessment of this example: This prompt clearly breaks down the required elements for each item (summary, prescriber names, supporting statements), guiding the AI to produce detailed, multi-part responses. 

For long lists, simplifying the request by asking for just the name of each concern makes navigation easier. Additionally, it would be beneficial to specify where the AI should source statements. This allows us to guide which documents are used by including types or labels. Lastly, providing guidance on the desired length of summaries and statements can prevent overly lengthy or brief responses.

Tables

Prompting the Writing Assistant to create an outline results in a table with rows and columns that are based on your configuration. 

You should consider using the table format for instances in which you want to organize complex information for ease of reference during case development. This can be helpful for situations such as organizing witness statements, or evidence timelines.

Table configuration options

You should consider the following criteria when configuring a Writing Assistant prompt to generate a table:

Table example

Scenario: We’re investigating a case involving improper opioid dispensing practices by multiple pharmacies and prescribers. The case hinges on whether these entities followed DEA-owned policies related to controlled substances, particularly opioids. We've gathered various internal policy documents from the pharmacies, along with DEA guidelines, to assess compliance. To get started, we want to draft a table that identifies and organizes DEA policies governing controlled substances.

In this case let’s prompt Writing Assistant with the following input:

• Compose a table
• Where each row is a unique DEA owned policy. 
• Include columns for Name (inclusive of aliases or acronyms), Policy Type, Brief policy description, Policy publication date 

Assessment of this example: This prompt is clear and structured well for generating a table with multiple data points. It successfully instructs the AI to organize the output into rows, each corresponding to a distinct DEA-owned policy. The inclusion of specific columns (Name, policy type, brief description, policy publication date) ensures that the AI will create detailed and organized responses, capturing various facets of each policy. Asking for aliases or acronyms also adds a layer of comprehensiveness.

However, the row description could benefit from clarification — specifying whether you're referring to completely distinct policies or policy documents, even if they have the same or similar content. Additionally, the phrase "identify as many policies as possible given the sources" could be clearer. Specifying the source types (e.g. official DEA publications, legal documents) might help the AI focus on relevant data. 

Want more examples?

Check out our templates.

Everlaw currently offers the following four pre-defined templates in Writing Assistant:

• Factual Background: A memo template that analyzes the factual events and context of the selected evidence. Consider using this to generate a statement of facts.
• Misconduct Analysis: An outline template that analyzes any misconduct committed by the parties found in the selected evidence  
• Theme List: A list template that indexes and describes distinct themes present in the selected evidence. This template is particularly valuable for generating deposition questions based on a specific set of evidence.
• Cast of Characters: A table template that summarizes key entities found in your chosen sources 

To use a template, open Writing Assistant and select one of them from the dropdown menu in the Step 2/2: Configure task stage of the dialog.