Ethical best practices for using AI in marketing

AI has morphed from an abstract concept involving chess-playing computers to a scene-stealing disruptor on the marketing stage. Use cases for AI are proliferating like dandelions, as creative and innovative marketing teams harness the potential of these tools to hypercharge their marketing programs in creating and repurposing content with speed and efficiency that was unthinkable one short year ago.

But ethical concerns cropped up just as quickly. Used carelessly or improperly, generative AI tools can create big problems just as fast as it might be able to solve them, including risk of plagiarism, copyright infringement, or by introducing bias as a result of the corpus they were trained on.

As quickly as we identified the potential and the risks, we convened a study team, and charged them with creating a "Playbook," including guidelines for ethical use of AI in marketing.

"The AI space is growing very fast. We have collaborated cross-department in order to identify the best current and future use-cases for this technology across all of our disciplines. It is important to engage our entire team and get them comfortable with these technologies as their disruption potential is difficult to overestimate. It's an exciting time and we're applying new techniques and processes every day." –Jeff Turcotte, Director of Engineering

General do's & don'ts

Here are some high-level do's and don'ts generated by our internal AI team that are helping us harness the power of AI.

DO

1. Review the results of AI-generated content very carefully and thoughtfully, looking for accuracy, completeness, tone, bias. AI is very confident in being wrong.
2. Stay up-to-date on the use of the tools, and which ones are optimized for different tasks. The landscape is rapidly evolving, this is a very dynamic space.
3. Play and be creative! Try different prompts, or try refining your prompts so you can have alternatives to consider. Small variations in the prompts can create very different results in some cases. There is a whole field growing around this now: Prompt Engineering.
4. Collaborate with and correct the AI. If generated results are poor or incorrect, let the AI know about the issues and there is a good chance it can fix them.
5. Seed conversations with prompts for better results. For example: “You are an expert translator. You take words in X language and respond with the best possible translation for them in Y language.” In the case of ChatGPT, these will help set the context for the rest of the session.
6. Save the iterations of the prompts you try. Critical prompts and/or seeds should be stored with official project documentation. Share valuable prompts with your colleagues!
7. Check the content created for plagiarism using a tool like Grammarly Desktop, which is itself an AI-derived service. Also, check any images created or manipulated with AI for watermarks or other visible IP or copyright infringement.

DON’T

1. Don't expect tools like ChatGPT to tell the truth all the time. For example, If you ask about something that doesn’t exist, there is a good chance that ChatGPT will provide you with “what you want to hear” rather than deny you an answer. This is expected to get better with time.
2. Don't "grab-n-go" with AI-generated content without carefully reviewing and editing it, ensuring it is useful and accurate for the purpose intended. Generative AI can get creative and often takes liberties with facts. Consult other sources to validate information of great importance.
3. Don't use copy, images, or prompts to generate AI content that you do not have the rights to use.
4. Don't train a model on or create an embedding of anyone’s image, copy, artwork, or likeness without their explicit permission. Using these things as an input is allowed, but still recommended you ask for permission.
5. Don't input any confidential information into an AI system. Some tools use input to further train, fine-tune, or for analytics. Eventually, there should be clearer privacy policies with the popular cloud-based tools

Ethics of AI use

Some of the do's and don'ts we outlined are steeped in ethical concerns. We want to do the best work for our clients with the most cutting-edge technology, but we also know that there need to be limits to what we use as source material. Below you will find our bedrock ethical standards.

1. No use of external sources, images, content, or code that is not specifically licensed for reuse in the creation of prompts or source code.
2. No training or embedding of likenesses without explicit permission in writing, in the form of the approved Imarc model release.
3. Use anti-plagiarism tools like Grammarly to review generated content.

Departmental use of AI

Different departments at Imarc are utilizing AI varied options. As such we wanted to share some of the ways we've been experimenting with it by discipline as well as possible future use cases.

Engineering

• Current: Programming support
  
• Current: Troubleshooting and Debugging
• Current: Professional Development
• Future: Custom training on internal code

Creative & Strategy

• Current: Content writing support (final copy is not generated solely by AI)
• Current: Brainstorming / creative inspiration and support
• Current: Light image manipulation (background removal, color correction, upscaling, etc.)
• Current: Micro-copy for non-writers (CTA buttons, etc.)
• Current: Heat mapping without historical data or users
• Future: Converting wireframes and sketches into editable designs
• Current: Abstract image creation (within reason)
• Current: Generating mood boards
• Current: Rapid inspiration and visual exploration
• Current: Testing assumptions about audience and archetypes
• Current: Using AI to determine what not to do
• Current: Turning great writers into subject matter experts, quickly
• Current: “Image to Image” AI assistance
• Current: Improvements and automation around common workflows (subject isolation, color correction, etc.)
• Future: Using AI to fill out a smaller more strategic team for clients with ideas bigger than budgets
• Current: Initial research and drafting

Account management & business development

• Current: Recording transcript of meetings
• Future: Automating predictable task creation
• Current: Scope assistance for faster turnaround
• Current: Project cost estimation

AI can seem a bit intimidating at first, but you've likely already used it for several years, so don't be afraid!

• Are you a regular user of Grammarly?
• Is voice search part of your go-to tool set?
• Do you utilize predictive bidding strategies in your pay-per-click marketing efforts?
• Do you open your iPhone using face ID authentication?
• Are you on social media sites?
• Have you performed a Google search?

If you answered "yes" to any of the above, congratulations, you're already on board with AI use!

The dawn of the AI-era feels just the same as the "world wide web" did to me back in the 90s when I founded Imarc – a point of inflection, a moment when I felt 100% certain things were about to change in a very big way. We're all-in to explore and develop the potential to use AI to improve our work and the value we deliver to our clients.

AI is going to keep evolving and those who embrace it will ultimately win the day. We will proceed with caution, but for sure, we will proceed.

Would you like to discuss more about AI and how Imarc can help you benefit from it? Let's talk.