Pomegranite AI policy

This will act as a “living document”, to be updated when necessary. This comes with an open invitation to interrogate it, as a team, at any point.

This document serves to outline Pomegranite’s approach, protocols and policy regarding the use of artificial intelligence (AI) in our work, particularly relating to research, concept and content creation and copywriting. The protocols below have been developed with an understanding of the setbacks and opportunities of AI usage, as outlined by OpenAI.

Introduction

AI literacy is becoming an essential skill, and there is an increasing need to equip Pomegranite’s team members with the necessary skills to navigate AI for use in our profession. Pomegranite aims to foster an environment that embraces and adapts to new technology and welcomes the learnings that come with it.

This involves developing an understanding of how the benefits of AI usage must be balanced with the potential drawbacks, threats and limitations of AI.

When AI is mentioned in this document, it is primarily in reference to OpenAI’s ChatGPT.

Pomegranite’s values

Our AI policy takes into account our values of honesty, transparency, efficiency, professionalism and providing a personal experience.

Adoption of AI tools

Generative AI is still a fairly new tool in professional spaces, but it has fast become mainstream. Therefore, the primary consideration for AI softwares now is whether to restrict, adapt, or adopt. Factually, because AI is mainstream, Pomegranite team members must be able to use AI tools to our advantage. However, AI must be used with skill, creativity, consideration for originality, and critical thinking. AI tools can help us offer our clients more value if used effectively and ethically.

The way that Pomegranite makes use of AI must centre on an understanding of  intellectual property, integrity and ethics in content production.

Bias

Pomegranite team members must understand that AI holds bias that will be reflected in the answers produced by ChatGPT. Content generated by AI must be interrogated from this angle.

Confidential, sensitive and proprietary information

Proprietary information and trade secrets are not intended for public disclosure and may not be fed to any AI software. This includes sensitive information such as client names, names of individuals, addresses, and any part of Pomegranite’s original digital communications strategies. Under no circumstances may the real names of minors be fed into AI software.

Before prompting ChatGPT, names of people must be changed to their initials, and names of organisations, businesses or clients must be changed to a descriptive phrase, such as “a youth-focused advocacy organisation”. Under no circumstances may the names of Pomegranite’s clients be fed into AI software.

Plagiarism, transparency and best use

Copying work directly from ChatGPT is considered plagiarism. As a Pomegranite team member, you may not submit work that is not your own. You are encouraged to use ChatGPT to make work more efficient, but this must be integrated meaningfully into the overall work produced.

While team members do not need to indicate each time ChatGPT is used, you are encouraged to share with the team when a new, helpful feature is discovered, or if you have found a way to use ChatGPT in a way that aligns with Pomegranite’s AI policy and adds value to client work.

Content produced by ChatGPT must be interrogated with critical thinking, and if used as part of client work, must show strong alignment with client goals, tone of voice and brand messaging.

Key uses

ChatGPT is currently limited to use for ideas generation, brainstorming, specific content generation, and to simplify content into concise ideas, for easier and more efficient work. This includes:

  • Image generation: This is allowed only if a suitable real photograph or graphic is unavailable. It must be made clear, both internally and externally, that the image has been created by generative AI. When the work of an artist or photographer is available, this must take preference.
  • Processing and simplifying client briefs: This is allowed only if sensitive information has been removed from the prompt.
  • Concept/ideas generation: This is allowed only if sensitive information has been removed from the prompt.
  • Writing: ChatGPT may be used to simplify a big idea into concise points, but the answers generated by AI must always be checked for grammar, spelling,accuracy and bias. Copy and pasting work directly as is, and submitting this as your own, is not permitted. Work must clearly show critical thinking and an understanding of the client brief and purpose of the copy. In this regard, ChatGPT may be used as an “assistant”.

Key restrictions

  • Processing confidential, sensitive or proprietary information.
  • Language and grammar checks: ChatGPT was not developed for this purpose, and therefore should not be used to check grammar. If ChatGPT is used for translation, this must be verified elsewhere.
  • Strategy: ChatGPT may not be used to develop strategy for clients. Marketing strategies must be authentic, original and effective, and generative AI cannot yet do this constructively. Prompting ChatGPT for brainstorming strategy ideas is permitted, but ideas must be relevant and not generic.
  • Blog writing: This is considered plagiarism. Prompting ChatGPT for ideas on the structure, layout or headings is permitted.
  • Post writing: This is considered plagiarism. Prompting ChatGPT in order to synthesise a brief for a post is permitted.

Policy updates

As AI is constantly developing, this policy will be revisited every six months and updated as necessary to reflect any new software developments on the platforms we use or should be using. We encourage all team members to open a conversation with the team if they have any ideas or thoughts on additions or revisions to the policy at any point.