Do...

     Don't...

• Use AI for grammar and spell check.
   
• Use AI to analyze and present data that YOU have provided (like the zip codes of your donors, or the hours your volunteers contribute and when those hours have happened, or the kinds of phone calls you are getting, or common themes by survey respondents, etc.). But be sure to keep data secure and confidential - make sure you aren't exposing such by using AI to analyze it.
   
• Use AI to analyze images and photos that will help further your mission. For instance, if you are an organization that works to educate people about invasive plant species, AI might be used to “look” at photos submitted and identify what the plant might be. Or perhaps you work with a marginalized area of a city where residents want better attention from the government, and residents of that area take photos of poor or dangerous road conditions, and AI could help you categorize those for presenting to the government body responsible for that road. When war broke out in Sudan in 2023, MercyCorps used 10 years of satellite imagery showing the health of the country’s crops in each of those years, based on how plants reflect light in different wavelengths, and then used AI to compare real-time crop health with historical patterns. The MercyCorps team could then see areas where people were most vulnerable and could deploy its resources where they were needed most before disaster struck.
   
• Use AI to translate essential text in another language so that you, yourself, can read that text (don't use it to translate your entire annual report, however).
   
• Use AI to communicate with a client or member of the public when a human translator is not available, and when you are certain that AI will protect the conversation if it needs to be protected (e.g., if it's going to reveal home addresses, medical conditions, or other information that should be confidential).
   
• Use AI to write something short, like a social media post or an event announcement, if you truly do not have the skills to do this, nor the time to do this yourself, AND you have all the details needed for an accurate announcement. AND, you still have to read what is written and make sure it is accurate - AI written summaries of books or long texts are especially prone to misinformation and inaccuracies. Also, a human should still be writing MOST of your social media posts - and when they don't, when it's mostly AI, yes, we know, and it affects our opinion of your nonprofit (not in a good way). 
   
• Use AI to write a poem or a song if you really need a poem or a song and you truly cannot do it yourself, nor are you willing or able to recruit a volunteer to do it. But declare that AI is the author for any presentation and, in presenting that poem verbally or in performing that song, please keep it human - authentic human speakers or human performers only!
   
• Use AI to screen a mountain - an onslaught - of applications for jobs or programs, if your organization is regularly overwhelmed by the numbers of such applications. Having one job opening every few months and having 100 applications for each job is not an onslaught. Remember that there are a variety of built-in automation tools for most online forms that will already allow you to screen out anyone outside your service area for a certain program, for instance, or allow you to screen out every person who declares "no" on the question, "Do you currently have legal permission to work in the USA?" A better option most of the time for a nonprofit's job or program applications is a screening committee of volunteers, if the HR staff or program staff can't handle it on their own. Humans reading applications not only can be the best for identifying job applicants and program participant applicants worth a closer look by HR or program staff, they also can see trends and information that can advise you on how to improve forms and application processes - humans can often see the barriers for qualified applicants in your process that AI cannot. 
   
• Use AI for a chat bot that will answer questions via your web site for clients or the general public if you feel that your clients and most people who would interact with that chat bot would get their questions answered accurately, that it would not take away from the essential information gathering and relationship-building that happens through staff interacting with clients, volunteers, the public, etc, and if you believe that it could substantially help your staff use their time to focus on other, much needed tasks.
   
• Use AI to chat with a bot that speaks in order to learn another language (but don't use it as a substitute for practicing conversational skills with actual humans). Duolingo is a very popular language app that uses AI in a number of ways, including allowing paid subscribers to have short video chats with its characters. It's excellent practice for learning another language, but don't use it as an outright substitute for practicing speaking with other actual humans. But test that chat bot regularly (volunteers are great for this). A chat bot is always, ALWAYS, a work in progress.
   
• Be transparent with the public about your AI use and your policies for using it, just as you are regarding why and when your organization reserves roles for volunteers (right? right?). This will build trust with your community in your messages being authentic and human-centered. Also declare when AI was NOT used to write something: add a note to emails, blogs and other long form creations that says “a human wrote this, not AI; while we do use AI to check spelling and grammar, our authors are always humans.” Be proud of your humanity and authenticity!  
   
• Implement guardrails regarding AI use with regard to privileged, confidential and proprietary information, like board meeting minutes, financial information, employee performance reviews, donor names and addresses and the amounts they have given, volunteer names and addresses, confidential employee and member or client information, privileged communications with counsel, etc. Such data should never be inputted into a free AI tool. If you want to use AI with any such data, you need to pay for a tool that ensures data protection within your network and significantly reduces various risks.
   
• Draft a policy for staff use of AI in their work, particularly in the creation of content and the handling of confidential data, and be ready to continually discuss this and revise it. Feel free to use this table that you are reading now as a basis for your own policy, as well as the oh-so-excellent Top Ten AI Usage Policy Considerations for Nonprofits, by Holly Peterson , Esq., for the Nonprofit Alliance. Your policy should include a list of AI tools that staff should, and should not, use regarding certain data, and provide explicit examples of what is, and is not, appropriate.
   
• Connect your staff with trainings and resources on how to leverage AI tools, as well as what NOT to do.
   
• Create ways for staff to share with each other how they are using ANY tools in their work, not just AI, but shared calendars, voice mail, Reddit, WHATEVER. This could be a regularly moment at staff meetings, or something mentioned in their monthly reports.
   
• Keep in your budget plans to hire humans - professional translators, web designers, graphics designers and a person to answer your phone when a person becomes exasperated by your AI chat bot. AI is not an absolute substitution for those roles - if you think it is, prepare to lose donors, volunteers and others who feel your nonprofit looks like everyone else and has lots its humanity.
   
• Track the time it takes for staff to learn and to use AI tools, and the costs associated with using AI tools (buying the software, buying the subscription, paying for training, etc.), and have numbers to show your board and your funders regarding the financial costs of using AI; if you have ways to show AI as saving staff time (and therefore, money), you should do that too.
   
• Read about and listen to information about how nonprofits are using AI and consider new uses at your organization.
   
• Believe that YOU are the best person to decide when and how to use AI, not a tech company, and feel confident in saying "No."
  

• Use AI alone to find answers to any question whatsoever. No matter what answer AI gives you in response to a question, you need to double check and research further to ensure that the answer is accurate - and don’t be surprised that it’s not. Even AI-generated answers to math questions can be suspect.
   
• Use AI to create most of your social media posts. WE CAN TELL. We see your AI slop, because it's general and reads like all other AI-generated posts and you embarrass yourself and your nonprofit when you do this.
   
• Use AI as your final editor. If you use AI to write something or to edit something, you, the human, MUST re-read or review its work, because if you don’t, you are going to send out incorrect information - AI regularly hallucinates information! The Internet is filled with examples of this, including in courts where AI has created and referenced non-existent court rulings and legislation. And here's an account where an unauthorized AI re-write almost ruined a very personal family event.
   
• Have AI write a blog or a white paper or a song or a summary or someone’s biography or something else substantial and then not declare that AI was the author, rather than a human, even if a human double-checked the work to make sure it’s accurate (which should always happen). If AI authored something substantial, you shouldn't hide that, and you are hiding it if you don't declare it.
   
• Use AI creations without any human alteration. If you use Canva, for instance, to create a graphic, you need to alter that graphic at least somewhat to make it your own and also to make it accessible (very light text on very dark background, or vice versa, a font size that can be read by most people, etc). Otherwise, you are going to have graphics that look like other people’s and organizations, and also have graphics that many, maybe most, people can’t actually read. AI does NOT understand graphics best practices!
   
• Use AI alone to translate something substantial (a guidebook, policies, a form, your annual report) into another language for clients, volunteers, the public, etc., without having someone who is fluent in that language review the translation, especially if whatever needs to be translated is longer than a paragraph.
   
• Use AI tools that present voices or live text that makes the person interacting with such believe they are interacting with a human. Make it clear that “chat bot” means “talking to AI”, not to an actual human. 
   
• Present AI created artwork without declaring that the graphic was created by AI, rather than a human, if there would be any doubt in someone’s mind about such. 
   
• Use AI to create what looks like a photo or a video of a real human, animal, building, environment, etc. To present an image or video that looks like a real human talking or dancing or moving or volunteering whatever, but you've created it from a computer program, is misleading at best and dishonest for most.
   
• Use AI to author a performance review or employee reprisal. AI can, and should, review it for grammar and spelling, and if you need AI's help in saying something more diplomatically than you know how to write it, okay, but to use AI to author something that should be absolutely authentic and YOUR words is reprehensible.
   
• Use AI to remove or hide most avenues for someone to talk to an actual human at your organization.
   
• Use AI to narrate a video, a slide show, etc. that promotes your organization and its work. There is ZERO reason for a credible nonprofit to do this! If no employees or clients feel comfortable doing it, recruit a volunteer specifically to do it. A voice that sounds human should be an actual HUMAN voice, except, say, for a person with a disability who needs AI to speak their words.
   
• Think that AI can entirely or even mostly substitute for humans talking to your customers, clients, donors, volunteers, etc.
   
• Use AI to replace human judgment in critical decision-making or in program or activity oversight.
   
• Brainstorm with AI. I have so much to say about this, I wrote more about it later on this page.
   
• Have AI analyze a range of data from a variety of data sources, not all of which you provided, and assume that its findings are accurate. You need to at least spot check the results to make sure the AI tool really has used the data sources it says it has, and that it really has interpreted them accurately.
   
• Input privileged, confidential or proprietary information into a free AI tool or a tool that you do not know whether or not it ensures data protection within your network - or give a free AI tool or insecure AI tool access to such on your network.
   
• Ignore that AI has stolen other people's work and words, and that narratives "authored" by AI may, in fact, be the words of someone who was not compensated for their work, or that graphics produced by AI may have outright copied or at least alterned the original work of a human artist who was not compensated for use of their work.
   
• Forget that AI tools, especially free ones, require users to click on an agreement which confers broad rights to that AI platform to utilize, learn from, and disseminate inputted content. That means they can use WHATEVER you share with such HOWEVER they want to.
   
• Ignore the negative environmental impacts of AI. Keep educating yourself about them and about your nonprofit's impact on the environment, not just through its use of computer and Internet tech, but also its use of cars, trucks, etc.
   
• Let someone from the corporate world direct you how you should be using AI. Listen to their pitch, but if they don't have examples of nonprofits of your staffing size and budget using AI, and they don't present these examples in a way that makes the benefits of use obvious, what they are saying is just theory and hype.