The AI Ethics Class - Using Microsoft Copilot as a Small Business Owner

This blog post is based on an essay originally submitted by the author as Assignment 1.2 for Emory's ECE Ethical AI Course

Microsoft Copilot for Web is a consumer facing generative AI system that completes everyday tasks.  However, unlike ChatGPT, as of February 2025, it does not have live access to the internet. 

Although it can be used for many things, within my business, I primarily use it to create marketing copy and social media posts for my business.  I don’t yet have M365 Copilot Pro for Business account but Microsoft recently added Copilot to my personal Microsoft Office subscription.

According to the Microsoft site, standard copilot allows you to generate and edit up to 15 AI images, use GPT4 and GPT4 Turbo, and use a special notebook to craft prompts.  GPT4 and GPT4 Turbo are both models provided by OpenAI.  GPT4 Turbo allows a context window of up to 128k tokens and a maximum output of 4,096 tokens.  GPT4 allows 8,192 tokens each for input and output.

This tool is subject to the laws of the United States and the business account includes data protection so that my input is not used to train the model.

To create my marketing copy, I usually upload a transcript of my class after it ends with a prompt identifying my target audience and the context of how I would like to use the output along with a role (e.g. “Imagine you are a master Twitter/X social media marketer.”)

Charge your phone or create an AI generated image.

Even when using the free version, AI models require a lot of power.  Power is used during both training and end user consumption.  In a Feb 2024 article in The Verge, James Vincent estimated to generate one image requires the same amount of energy as charging a telephone.  Text prompts are less energy intense, requiring electricity equivalent to watching a few minutes of your favorite Netflix show.

As more and more people use AI, it increases the volume of information available on the internet but not necessarily the quality.  My AI generated post may be effective at gaining impressions, but it is not the same quality as a post actually created by a master social media marketer, even if that marketer also used AI.  AI models then use this content to generate new content.  This feedback loop can contribute to the drift we read about in the unit on Risk.

Garbage In. Garbage Out.

As more small business owners like me rely on AI for marketing or other white-collar jobs, that subject matter expertise will begin to be lost.  The ability to find and hire a specialist will become increasingly difficult and expensive, in much the same way as fast fashion has reduced the availability of seamstresses and fabric stores. 

Although this may not make as much difference in industries such as marketing and fashion, there are real dangers of not having SMEs available to correct or at least sound an alarm when AI models produce low quality results.  Another comparable example might be how the closure of small airports has impacted the commercial airline industry because there is no longer a robust feeder system for pilots. It is up to us as consumers and business leaders to decide if this is the future we want.