At Magnolia Communications, we embrace rapid advances in technology and equip our team to understand the pitfalls and possibilities of integrating new technology into our workflows. Generative Artificial Intelligence (AI) has seen immense growth in recent years and we acknowledge its many uses in the marketing and public relations fields with regards to ideation and automation. But we recognise that GenAI technology is still in its infancy and must not be used without human intervention and oversight.
There is no doubt that GenAI is here to stay, especially considering the vast investments being made into AI by companies around the world. We understand that there is going to be continued use of AI in our industry, as well as our clients’ industries, which is why we have decided to place guardrails within Magnolia on how to use GenAI, and we have set this policy for such internal use, as well as how we communicate AI use to our clients.
01Purpose
This policy outlines Magnolia’s dedication to transparency around the ways GenAI is used within Magnolia’s operations. We aim to use AI technology safely and ethically, ensuring the highest level of confidentiality for our clients.
02Statement of Transparency
At Magnolia, we understand how valuable AI tools can be in kickstarting the creative process and as a tool for research. From a digital marketing standpoint, we use iterative AI tools solutions with our martech stacks. But we are committed to championing human creativity above all as we continue to deliver on our promise to produce the best quality for our clients.
To ensure transparency in all our GenAI use, we have implemented safeguards to protect the privacy of our clients and their intellectual property. All GenAI usage within Magnolia will always include steps for human oversight and verification before dissemination.
03AI Usage Best Practices
All members of the Magnolia Team are expected to adhere to the following usage and security practices when utilizing GenAI for any and all work, whether used for internal or external audiences:
- Original Work. We will never pass off AI content as created by our own team members. Any content that includes any portion of AI use will be disclosed as having been generated using AI.
- Reviewing Work. All content that has been partially generated through AI will be approved by human team members before deployment.
- Understanding AI Downfalls. AI tools will not be used as a search engine as information may be outdated, unverifiable, or hallucinated (made up) by the AI tool.
- Checking for Accuracy. Any data generated by an AI tool will be verified by checking at least three other sources.
- Leveraging AI; Not Relying on It. We understand that suggestions made by, and content generated by, an AI tool do not need to be followed. We stand by the statement that a task that cannot be done by a team member must not be commenced or completed by an AI tool. AI can only be used to aid in tasks and for efficiency, not as a replacement for staff capabilities. Any creative content generated by AI must still pass the human authenticity test and quality assurance via peer review.
- Subject to Review. Any content that is created using AI beyond the threshold of efficiency-management is subject to being checked using a plagiarism checker tool or a tool with a similar function.
- Confidentiality. No sensitive information, including but not limited to client data, Magnolia data, or intellectual property will be input into an AI tool for any reason, whether idea generation or content refinement, as Magnolia is committed to protecting both ours and our clients’ confidentiality.
04List of Appropriate AI Tools and Usage
The Magnolia team has been exploring AI tools and are currently using the below tools with limited capacities:
- ChatGPT is used only in combination with SEMRush to research keyword ideas. We do not use ChatGPT as a search engine or for content creation.
- Grammarly is used for grammar checks on content but without compromising on an individual’s communication style.
- Adobe’s embedded AI tool is used for image refinement only. All graphics are created and approved by in-house designers before being sent to clients.
- On occasion, CanvaAI is used to resize image dimensions.
- Team members use Perplexity for brainstorming and idea generation.
Magnolia will continue to explore new AI technology as it is launched but we will proceed with caution. We will only engage with AI tools that require the use of an official Magnolia email address to register. This step will ensure we have the ability to keep track of AI usage within the company.
05List of Blocked AI Tools
The Magnolia team understands the difference between iterative AI and Generative AI tools. Iterative AI is programmed with rules to make decisions and to make data-based predictions. GenerativeAI is trained on data input into its systems to create new data and content. Iterative AI has been in use for many years and affords users a greater level of control. Many existing tools within the martech suite that Magnolia uses employ iterative AI. GenAI is a newer form of AI that is still considered unreliable due to its creation of falsified data, as well as accusations and suspicions of plagiarism.
The following is a dynamic list of AI tools that have been blocked from use within Magnolia due to poor data privacy, cybersecurity issues, and bias:
- ChatGPT, guided use only. We only allow guided use of ChatGPT, as it is prone to hallucinations, built with inherent biases, and often has outdated information, or shares contradictory information. Additionally, ChatGPT uses proprietary company information as training data which might impact company and client confidentiality. Hence, Magnolia strictly avoids the use of ChatGPT in the following instances:
- As a search engine
- For client research
- For media list research
- Data analysis
- Text content generation
- Image content generation
- As a tool for verification
- Grok, the xAI chatbot, which can be manipulated by false data into generating disinformation in text and image form.
- Microsoft Copilot, which has an opt-in mode to take screenshots of user screens, putting client confidentiality at risk.
06Disclosure of AI Usage at Magnolia
The marketing and PR worlds have adopted widespread use of AI. While we understand the draw to use AI in one’s work, guidelines are required to protect our clients from misuse of GenAI, as well as to protect our employees who choose to use AI technology.
Thus, the use of AI at Magnolia must always be disclosed for complete transparency when employed for internal use, and with external parties, and the same is expected from our vendors and suppliers.
Internally
Any team member within Magnolia who uses AI for any public-facing content, graphics, or reports must inform team members of the following:
- Which AI tool(s) was used
- The prompts that were input
- Which areas of the content AI was used for
The AI content must be vetted and verified by two other members of the team before it can be deployed in public.
Externally
We will disclose AI use to clients in MSAs and SOWs, as well as on the Privacy Policy page of the Magnolia website.
If an employee is using AI to the point that the client needs to be informed, this information must be shared with the employee’s direct managers, and escalated to the President of Magnolia, who will share a statement with the clients.
Vendors
Magnolia’s suppliers, vendors, and subcontractors are requested to disclose any and all use of AI for data analysis, text or graphic content creation, or report generation. Only once AI use has been disclosed and verified by a Magnolia staff member, can the content be sent to clients. If AI is not disclosed by a vendor but discovered later, the vendor will be subject to financial penalties equal to the amount of the contract.
07Client Confidentiality
We are aware that data input into AI tools can be used for training large language models. While AI tools can expedite many tasks, Magnolia is deeply committed to keeping client information confidential at all times. Hence, when using AI tools, we will not input any information that directly mentions a client’s name, their product(s), service(s), or any information that we would not otherwise share in a public forum without prior consent.
Examples of prompts to avoid
- Create a news pitch for [company name] about their [product name]. Include the benefits of [product], [target vertical], and a request to interview [spokesperson name].
- Compose an email pitch to journalists in [target sector] [target country] about [client name] being available to share thought leadership.
- Write three versions of ad copy about [client name]’s [service] for LinkedIn sponsored ads. Personalize copy for [target vertical].
Examples of acceptable prompts
- Share three ideas for a pitch on financial planning in 2025 from a financial institution based in BC, Canada.
- Suggest a news pitch idea for why an organization specialising in foreign exchange is a better choice for financial institutions than traditional banks.
- Brainstorm three ad ideas to show the benefits of using a mobile top-up organization to send money.
08Repercussions on Repeat Violations
At Magnolia, we have complete trust in our team to use AI responsibly and with transparency. But in the instance of a team member repeatedly violating our policy for AI use, they will be subject to an internal inquiry and appropriate disciplinary action. If inappropriate use of AI continues following the inquiry, the team member will be subject to further disciplinary action, up to and including termination of employment.
09Review & Revisions
This AI policy will be regularly updated following new findings about AI tools, our experiences with AI use, and government regulations, as and when they come into place. Any revisions to the policy will be communicated to all members of the team.