Generative AI tools are becoming increasingly capable at helping people organize, analyze, and make sense of complex information. For many performing arts organizations, low-cost AI-powered tools are now within reach and unlock features that make emergency planning more efficient and consistent.
This article is the second in a two-part series on using AI for emergency preparedness. I wrote the first article, Enhancing Emergency Readiness in the Performing Arts with Free AI Tools, in the recognition that many performing arts organizations only have access to free AI accounts and focused on basic prompting techniques. This second article looks at low-cost versions of tools such as ChatGPT, Claude, and Google’s Gemini. It highlights features like Custom Instructions, CustomGPTs and other custom assistants, Projects, Deep Research or Extended Thinking modes, along with key data security and privacy considerations. The examples here reflect the state of these tools as of December 2025.
The goal of this article is to help organizations use advanced AI features responsibly while preserving the authority of human review and expertise. The tools should be treated as a drafting partner, not a decision-maker. Final plans must be reviewed and approved by leadership, staff, and, when appropriate, first responders or consultants.
If any of the steps in this article are unclear, you can upload the entire article to your favorite tools and prompt with something like: “Act as an emergency planning assistant focused on the performing arts. Review the attached article and explain how I can create [Custom Instructions / a Project / a Deep Research prompt] in my ChatGPT Plus account. Ask me a few questions about my organization and then walk me through the setup and draft any text I can use.”
Prompts like the ones in this article are only the beginning of a conversation. Refine them iteratively, correcting the model and adding details until the output matches your needs.
With most providers, including ChatGPT, Gemini, and Claude, users may choose from several different large language models (LLMs) that may respond differently to your prompts. Advanced models can be more consistent on complex planning tasks, but actual performance depends on the specific model and vendor. It is worth looking at the options in your account and, if you are unsure, asking the tool itself to explain in plain language which model it recommends for the models available to you and the task you are working on.
I. Free vs. Paid AI Accounts
While free accounts are excellent entry points, paid subscriptions offer distinct advantages for developing and maintaining emergency plans. These benefits include larger “context windows” (the ability to process more information at once), enhanced file security, greater access to deeper reasoning models, the ability to create focused assistants, and customized project-style workspaces where uploads, chats, and notes can be kept together.
The ability to upload documents and maintain long-term context means you must pay closer attention to what you share and to each provider’s privacy settings. When opening an account with an AI service, review what data the vendor stores, how long they keep it, how you can opt your data out of training, and how you can delete histories. Never assume that an external system is the right place for confidential information.
These accounts tend to offer stronger privacy and security controls than free versions. Many plans comply with security frameworks like SOC 2 and ISO 27001, which require providers to follow standards for protecting sensitive information. Even so, organizations must use caution about what is provided to any external system.
Safe to Use: Publicly published information, general facility descriptions, aggregated data, and generic planning templates
Avoid Using: Personal staff/patron data, passwords, alarm codes, financial account details, or security system schematics
II. Using Custom Instructions for Organizational Context
Many types of accounts let you set “global” instructions that apply to every conversation. In ChatGPT these are called Custom Instructions; Gemini and Claude offer similar profile or preference settings. You can use these to describe your organization and tell it how you want it to respond, so every conversation starts with the context of your organization’s details.
Helpful items to include in these Instructions are:
- The name and type of organization (theater, music venue, touring company, presenting organization)
- Your general size and scale (staff numbers, volunteer structure, number of performances per year)
- Other specifics (age of your building, dates of renovations, any history of emergencies)
- Typical hazards in your region or location (such as hurricanes, wildfires, flooding, winter storms, or railroad crossings)
- A statement about expectations for accuracy, citation of sources, accessibility, and clear structure in all responses
You can include as much non-sensitive detail as you wish and even ask the model to do a web search for useful information about your organization and then create a draft for the instructions. Here is a brief example of custom instructions for a venue’s account:
“You are an assistant to the Fake Not Real Theater, a fictional concert venue in Atlanta, GA that owns their 50-year-old building that seats 500 patrons with 10 full time staff presenting over 100 performances per year. The building is in a dense in-town neighborhood with a parking lot on both sides and is near a railroad crossing. The location is subject to minor flooding from storms each Spring. Provide clear, well-structured responses that staff can easily adapt to our internal documents and procedures. Base your suggestions on established best practices for cultural organizations and clearly label any assumptions you make.”
A useful strategy is to keep Custom Instructions fairly general, describing the organization and preferred style, then use individual prompts (or Projects, described below) to provide specific tasks, documents, or scenarios.
III. Organizing Work with Projects and Similar Features
Many AI subscriptions include project-style workspaces where you can keep related files, instructions, and conversations together over time. In ChatGPT these are called Projects, and other providers offer similar workspace features that let you attach documents and maintain context across multiple chats. Always be sure to redact any sensitive information in documents before you share them with any external tool!
For emergency planning, you might create a workspace titled “Emergency Preparedness Update” and upload items such as prior versions of your emergency or continuity plan, board-approved policies, public facing evacuation information, and regional hazard assessments. Inside that workspace you can start separate conversations focused on topics like risk assessment, crisis communications, continuity planning, recovery strategies, and training and exercises. Each conversation can draw on the same shared materials, which reduces the need to re-upload or re-explain information.
Example prompt:
“Act as an expert in ChatGPT Plus accounts [adapt to your account type]. Explain how Projects [or a similar workspace feature in my AI tool] work. Then help me set up a Project for updating our emergency and continuity plans, by drafting Project Instructions, suggesting which documents to upload, and explaining how best to use this feature for my purpose.”
Using Projects or workspaces in this way helps staff keep planning materials organized and makes it easier to update plans consistently over time.
IV. Using Deep Research and Extended Reasoning Modes
Some paid plans offer two kinds of advanced modes. One is a deeper ‘thinking’ mode that simply allows the model to spend more computation on a single answer. The other is a separate research tool that runs a longer, multi-step process. In ChatGPT, Deep Research is the dedicated research feature that can search the web, work with uploaded files, and produce a structured, cited report. Other tools offer similar extended reasoning or research modes that may have different names. They are especially useful for tasks such as:
-
- Conducting extensive research into areas of risk or planning
- Reviewing long, multi-section emergency or continuity plans
- Comparing guidance from FEMA, NCAPER, state cultural agencies, and other authorities
- Identifying common strengths, recurring gaps, and priorities across several sample plans
Example prompt with Deep Research selected in ChatGPT:
“Analyze the four uploaded public emergency plans for performing arts venues and any key guidance you find on the web. Identify common strengths and recurring gaps, then recommend the top five improvements a mid-sized performing arts venue like ours should prioritize in its next plan update.”
If you are not sure how your account’s deep research feature works, you can ask the tool for a brief explanation and a recommended way to use it with your own planning documents.
V. Supplementing AI Workflows With Existing Documents
Your organization may already have documents with information useful for planning with an AI tool. Examples include older redacted emergency plans, consultant reports, public facility descriptions, training materials, risk assessment spreadsheets from dPlan/ArtsReady 2.0, and generic checklists.
Before uploading any information, be sure to remove or redact any sensitive content (names, direct contact details, passwords,…).
VI. Building and Using Custom Assistants (CustomGPTs, Gems, and Similar Tools)
Many paid platforms let you create custom assistants (similar to Projects) that are tuned for a particular purpose and use a set of reference documents. In ChatGPT these are called CustomGPTs, in Gemini they are called Gems, and in Claude similar behavior is handled through Skills (task specific configurations that Claude can load when needed). Each custom assistant has its own instructions and knowledge that are separate from your general account settings.
Custom assistants are useful when you want a repeatable helper for specific planning tasks, for example:
- Risk Assessment Interviewer, which walks staff through key questions about hazards, vulnerabilities, and existing controls
- Continuity Planning Helper, which turns uploaded worksheets into draft continuity plan sections
- Tabletop Scenario Generator, which creates rehearsal scenarios tailored to your venue and hazards
- Evacuation Checklist Reviewer, which checks public evacuation procedures against basic life safety and accessibility concerns
For a single organization, you might create an assistant by adapting these setup instructions:
“You are an emergency planning assistant for the Fake Not Real Theater. Help staff create structured drafts for risk assessments, continuity plans, and communication procedures. Use the emergency planning templates, existing emergency plan, risk assessment questionnaire, and guidance from established emergency preparedness frameworks in your knowledge base. Identify missing information and suggest questions that staff should resolve.”
In many tools, these custom assistants can be shared with colleagues so that everyone can use the same configured helper. Sharing settings vary by provider, so review how access and permissions work before including any sensitive materials.
You can prompt one of these assistants with something like: “Search the web and find information about regional storm risks in our area and identify which items in our emergency plan’s hazard summary should be updated.”
VII. Advanced Prompting Strategies for Emergency Planning
In the first article I recommended a simple structure for prompts where the user assigns the AI a role, gives it a brief context about your organization, describes the task, and specifies the format of the output. If you are using Custom Instructions, a Project, or a custom assistant that already describes your organization, you usually only need to describe the task and output format. The role and context are already in place, so you can focus on the task and the desired format.
Scenario based prompt: “Guide me through a scenario where a transformer failure causes a partial power outage during a performance. Present this as a step-by-step checklist that identifies immediate safety actions, key decision points, and internal and external communication needs.”
Document comparison prompt: “Using our uploaded public evacuation procedures, compare Version 2023 and Version 2025. Identify changes that improve accessibility and crowd flow and list any sections that still need clarification.”
Vulnerability scan prompt: “Act as a strict assistant fire marshal or insurance assistant. Read the attached evacuation plan and any logical gaps or safety failures that could occur during a real emergency. Be critical in your report and provide practical steps to address these gaps.”
These techniques give the model a clear target and help generate more useful drafts that staff can then review, correct, and finalize.
VIII. Integrating AI Into Institutional Emergency Planning Workflows
These tools are effective when built into your existing planning processes rather than used ad hoc. One practical approach is to define a simple “AI step” inside tasks you are already doing. For example, you might:
- During annual plan updates: Upload last year’s plan, then ask it to highlight sections that may need revision based on new hazards, staffing patterns, or productions.
- For tabletop exercises: Ask it to propose scenarios tailored to your venue and then refine them together with staff and local first responders.
- After incidents or near misses: Use it to help structure an after-action review, summarizing what happened, what went well, and what needs to change.
Across all of these, AI should be treated as a drafting assistant. Human review, staff consensus, and professional consultation remain essential.
IX. Conclusion
Low-cost AI tools give staff practical ways to manage the volume and complexity of emergency planning work. Features such as Custom Instructions, custom assistants, Projects, Gems, and Deep Research modes help teams synthesize more information, keep materials organized, and turn rough ideas into clearer drafts.
When used with careful attention to privacy, accuracy, and the limits of AI, these tools can help performing arts organizations keep plans current and embed preparedness more deeply in day-to-day operations.
Clear, efficient prompting also reduces the number of unnecessary AI calls and revisions, which can lower the computing time and energy needed for a given planning task. Thoughtful use of these tools saves staff effort and modestly reduces the environmental footprint associated with your emergency planning work.
This article is informational and does not replace legal, safety, or emergency management advice. Organizations should review all plans and procedures with their leadership, legal counsel, and appropriate local authorities to ensure compliance with applicable laws, codes, and regulations.
To get started, you can upload this article (and the previous one) into your preferred tool and adapt a prompt such as: “I have uploaded two articles on emergency planning for the arts. Based on the strategies outlined in the text, guide me through setting up a Project for the Fake Not Real Theater’s new Business Continuity Plan.”