Navigating the AI Wilderness: A Guide for Advocacy Professionals

Don’t get stung!

Frameworks for internal and external advocacy & AI policies

Here’s how to do that:

Consider purpose and scope separately when defining your initial AI usage policy — and make sure you understand what both terms encapsulate. Purpose is the why, while scope is the how. To determine your purpose, pull together any potential benefits that AI could bring to your team. To decide on your scope, get a good understanding of exactly how your team can use these tools.

Remember that you’re building an adaptable policy. It’s meant to shift as things change and grow. With this initial plan, you should create something that you can adjust as your organizational needs move. It’s a good idea to start your policy at a relatively reserved level and establish a future roadmap for adjustments since it’s easier to expand a policy than restrict it. Moving through this progression will allow you to gather feedback and make new decisions about platforms and use cases as they arise.

Tips on defining purpose

Starting with purpose is usually best. Before jumping into your AI plans, you’ll want to get a good sense of your goals for the technology. Ask yourself why your team wants to use AI, and then go from there.

Answers to that question typically fall into a few categories:

1. We want to save time by automating tasks. 
2. We want to accomplish our goals using fewer resources. 
3. We want to micro-target messaging to audiences without manually rewriting it.

Defining your organization’s purpose for using AI will set your team’s expectations and ensure your next choices further those goals.

Tips on defining scope

After you’ve determined where AI fits into your overall strategy, the next step is defining how it will be applied. Consider these questions:

• How much should we allow AI to integrate into our workflow? 
• Who is eligible to use AI tools?
• Who is responsible for overseeing our AI policy and who are the stakeholders?
• When and where can our team access these tools?

These are difficult questions and they need to be considered carefully. To help you work through them, we’ve examined them in detail. Here’s our rundown:

We’d bet most people in your organization could benefit from some form of AI, but it’s up to you to decide when and where these tools are appropriate. There is plenty of ambiguity in this AI ecosystem, so a thoughtful and practical policy will get everyone on the same page.

Who is eligible to use AI tools?

Your AI policy shouldn’t necessarily focus on individual roles but, rather, on the inputs and outputs these roles handle. For example, you may want to allow anyone who writes membership emails to use an AI for that responsibility or not allow anyone who designs social media posts to rely on those tools. When making these decisions, you should specify what type of information is safe to be introduced into an AI prompt and how AI-generated content should be reviewed for accuracy and reliability.

You might run into situations where AI usage should be restricted for certain employees. It’s likely that, in those circumstances, you’ll want to provide restrictions around particular tasks. We’ll dig into this option in the fourth question of this section.

Who is responsible for overseeing our AI policy and who are the stakeholders?

You need to know who is running the AI show in your organization, and establishing an AI Governance Board early in the process can make that clear to everyone. Believe us, you’re going to want to institute a formal, transparent process for what could easily turn into an informal, vague practice. With a board, your team members will be able to monitor decisions and understand why they’re made.

An AI policy is only effective in the world in which it was written. When the technology changes, the policy needs to change with it. It will be the board’s responsibility to review any technology advancements and update the policy to include any new tools that could be acceptable.

Once you’ve figured out who is responsible for AI oversight, you can determine your AI stakeholders, both internal and external. These stakeholders will be people who are impacted by the AI decisions that you’re making for your team. For example, internal IT staff members will be concerned about security, while external members may be directly affected by AI-assisted content creation. Each one of these stakeholders offers insight into the appropriate role of AI in your organization.

When and where can our team access these tools?

It’s not just what you send into AI systems that can cause security or privacy risks; it’s also how you do it. To tackle that concern, you’ll need to delineate when and where your team can access these tools. You’re going to have to put some bumpers on this bowling lane because a free-for-all is only going to bring headaches.

For instance, even if you’ve determined which AI tools are acceptable and secure, if an employee were to access them from a home computer, using your organization’s proprietary data, they could unintentionally initiate a breach. To avoid this, you might determine that all AI use would be limited to organization-issued computers and networks.

How much should we allow AI to integrate into our workflow? 

The most pressing question when determining scope is how deeply you should allow AI to penetrate your work. You’re going to need to figure out the right balance, and that’s not a simple decision.

Let’s cover the bases before you make a choice that will set the tone for the rest of your AI policies. 

There are three primary levels of integration, each one with its own set of pros and cons. From least permissive to most permissive, here are those levels:

1. Restricted use – AI tools are prohibited; no questions asked.
2. Limited use – AI tools are generally prohibited except for specific, defined use cases. 
3. Open use – AI tools are freely allowed.

A restricted-use policy will completely prohibit AI use. It’s the most straightforward policy and easiest to enforce, but it’s not necessarily a practical solution for organizations seeking to navigate an advocacy ecosystem that is increasingly turning to AI tools. While this policy may not be pragmatic for your organization in the long run, there is one situation in which it makes excellent sense. If you’re currently detailing your AI policy, a restricted-use policy can provide some guidance until those details are settled. 

Meanwhile, limited-use policies provide flexibility and options for AI use. You’ll be able to choose when and how you turn to AI — and the impact that it will make on your efforts and goals. This policy is a good jumping-off point for organizations that haven’t yet implemented AI into their workflows. Examples of limited-use policies include allowing teams to use one specific tool, such as Google Gemini or ChatGPT, or limiting the use of AI to brainstorming ideas while restricting AI use for drafting messaging. After exploring these approvals, you could open up your policy to include additional tools and outputs.

Within a limited-use framework, you can also explore ONLY allowing locally installed AI tools. This could look like running an LLM (language learning model) on a laptop and not having it run on the internet to process things. It is a little more complicated to set up, but offers a lot more in terms of privacy. 

Finally, an open-use policy is the most permissive system, giving team members the complete go-ahead to determine which tools and uses are most appropriate for them. There is still an option in this seemingly free-rein plan to overlay some parameters. Even in an open-use policy, you could prohibit team members from uploading proprietary data into AI tools or from using AI-generated content in member communications.

After answering these four scope questions and considering the purpose of your AI use, you’ll have the answers you need to build the first layer of your initial AI policy. Your team will get clarity about how and why they’re relying on these tools — and you’ll feel like the plan-creating master you are. With that confidence, you can move on to the next layer.

With great power comes great responsibility. Yeah, we know, a too-on-the-nose movie line, but it’s true – and remember Peter Parker got a family member killed by messing around with powers he didn’t understand. AI is an impressive, dynamic, and impactful tool. It’s also a risky endeavor when attempted by untrained or unbound hands. That’s where you come in. You’re going to need to create boundaries for your team and their AI use.

In an advocacy organization, where trust is truly the foundation of every success, the use of AI must be governed by clear and strict ethical and responsibility guidelines. Luckily, there are standard considerations and expectations to which we can turn. 

It’s critical to quickly establish the ethical and responsibility guidelines for your organization, and you can do that by reviewing the possible impacts of your AI use, how you expect your team to behave with these tools, and the effects that your AI use could have on your staff. Once you’ve established those guidelines, you need to be able to communicate them clearly and effectively.

We’ve laid out some of these considerations here:

Set your ethical guidelines

AI can be great for your workflow productivity, but any thoughtful organization needs to consider the dilemmas that come with it. Life is all about balance, isn’t it? There are biases that need contending with, data security concerns, employee retention questions, and more. If you don’t grapple with this, every AI effort you make will be compromised.

We’ve compiled a list of some of these considerations, but you’ll need to turn inward and determine which guidelines you need to include in your policy to ensure your team’s ethical AI use. 

Here are our initial considerations:

1. Algorithmic bias and discrimination: AI tools are trained with information available on the internet. Since internet content is influenced by the biases and discrimination of those creating it, AI often inherits those biases. Once that happens, the algorithmic bias will impact the responses that come from the tool. To tackle this problem, you’ll need real people who can carefully review AI outputs to clear them of discriminatory and biased information.
2. Data security and privacy: AI uses a tremendous amount of data to generate outputs that seem like they could come from a human. With all that data comes ethical concerns about its collection and storage. To maintain secure processes, you’ll need to train your staff on methods for protecting security and privacy, especially as these tools get integrated into your systems and practices. This could also be another place for you to think about locally installed LLMs. 
3. Data manipulation and transparency: AI transparency is a top-of-the-list issue when it comes to trust. It can take many forms, but considering the tools you choose to use and your own labeling of AI-generated content is a good place to start. For the first, you should make sure any system you rely on has strong and transparent data collection, discrimination, bias, and security policies. For the second, you should clearly state which content is AI-generated across all platforms, including social media and messaging. It is also worth establishing a policy regarding what can be manipulated, and what cannot. 
4. Environmental sustainability: AI requires a tremendous amount of energy since the data centers needed run on massive computing power — and these data centers run on fossil fuels. For advocacy organizations grappling with environmental sustainability or concern for global climate issues, it would be imprudent to ignore these environmental impacts.
5. Job displacement: For many people, AI sounds like job loss. People are worried that tasks that once required humans to complete can now be done by these tools, and that translates to a small workforce. Advocacy organizations concerned with issues of employment, labor, poverty, and workforce rights should consider the impacts and impressions that could be made after AI use.

Set your responsibility guidelines

In addition to managing these crucial ethical considerations, you’ll need to deal with the issue of responsible AI use. While ethical considerations are a baseline for trustworthy AI-generated content, a reasonable advocacy organization will be sure to add extra responsibility measures to ensure the most productive, secure, and supportive practices. We’ve all got to try to be our best out here.

Here are some additional worthwhile considerations:

1. Human oversight: As helpful as today’s AI tools are, they’re useless to an ethical advocacy organization without thoughtful human verification, monitoring, and editing. As helpful as today’s AI tools can be, they can’t create content that’s above human review and monitoring. When turning to AI-generated content, you should be sure to clearly define your organization’s process for human review, including steps that should be taken to properly verify the accuracy of every aspect of the content.
2. Copyright, plagiarism, and vetting sources: It’s essential to properly vet and reference every bit of information used in AI-generated content. No one likes a thief, people. The AI tool will pull from sources across the internet, so ensuring that everything is both accurately and appropriately cited is key to maintaining the integrity of your organization’s work. This will make sure you avoid copyright and plagiarism issues from improperly referenced details or issues of veracity when it comes to sources chosen.
3. Erosion of fundamental skills: It’s irresponsible for an organization to ignore the impacts that AI use can have on its team members. Relying on AI tools too often can allow team members and advocates to begin to lose some of the fundamental skills that make them effective in their roles. It’s possible to see diminished written communication, verbal communication, and training retention abilities, but providing continuing education can help keep these skills sharp.

It’s important to stand behind what you decide when it comes to these ethical and responsibility guidelines. Your policy will only be as effective as its accountability requirements. After answering these questions and focusing on your ethical standards, you’ll need to determine how to communicate them and, then, how to enforce them. 

Speak up! Communicate your policy to stakeholders

Every stakeholder deserves to understand if and how you’re relying on AI. Once you create your policy, you’ll need to communicate that in writing to both internal and external stakeholders.

For internal team members, that’s as simple as keeping them appraised of the policy and any ongoing changes. For external stakeholders, including members, advocates, and legislators, that requires tactful messaging that reinforces trust and confidence. You could add a short statement to your external communications that directs readers to your AI policy, display that policy on your website, or include a line on specific communications, when appropriate.

In both situations, communication is key to maintaining trust and reliability — values that can’t be sacrificed as you walk toward improved efficiency and connection.

At this point, you’ve nearly finished building an AI policy that will work for your team members, support your mission, and maintain the trust you’ve built between you and your stakeholders. Now, you need to make sure this policy can shift with the needs of your organization and grow with the advancements of AI. You can do that by establishing a timely feedback and monitoring loop. Basically, you’re going to need to get a lot of other opinions from a lot of other people.

A productive feedback and monitoring loop depends on both the quantity and quality of the feedback. Collecting data from both your internal teams and your external stakeholders can bring both.

Internally, consider polling your AI-using team members to identify pain points. For example, you could ask them to consider which tasks are most time-consuming and then think about expanding your AI usage policy to assist with those responsibilities. As AI adoption continues and new methods become available, you could ask team members to share new tools they’ve encountered and consider adding them to your list of acceptable options.

There are a multitude of tools available to collect feedback from external stakeholders, and each one can provide useful insights into their experience with your AI-generated content. For example, you can run A/B testing in your advocate email communications to identify and quickly adapt your AI-generated messaging. You can also compare an AI-assisted email to one written by a team member to determine whether AI is an effective communication tool for your team. Comparing open and clickthrough rates provides a useful barometer for messaging impact.

Knowing what works is integral to the success of any system, especially one that is consistently growing and changing. Common policies currently in place regarding AI may not be appropriate or relevant after the landscape changes, so it’s imperative that you analyze and iterate when it comes to these decisions. Your established AI Governance Board should be meeting frequently to consider these feedback loops, advancements in technology, and the need for shifting guidance.

How do you adjust?