Tech Recipe: Growing your business with ChatGPT

Background:

ChatGPT has served over 100 million people faster than any service. Chances are you have at least heard of ChatGPT at this point, but what is it? ChatGPT is an example of?generative AI?that uses a?Large Language Model. Generative AI means AI software that is designed to create something?new?by taking your prompt, or query, and the training data within the software.

“Large Language Model” refers to a method of predicting the most likely next letter or word. ChatGPT uses an LLM to create its answers. Training data is all of the input that the AI uses to build the model it uses to understand inputs and return the output.

Did you ever make a sentence diagram in grammar class? It’s an oversimplification, but the software is doing that in every possible combination to your input to determine what you’re asking.

In other words, the software takes each word in your input along with the structure of your input and makes a really good guess at what words go together and how the entire query fits together. Now that it “understands” your query, it runs that against its model and training data to find all the combinations of words and the best way those words fit together.

AI trained on lots of quality, human generated data like books and articles will create quality output. While generative AI might be able to write a term paper, or even a blog post for you, it doesn’t know everything. It can be factually incorrect and it’s up to you, the human, to verify.?

Additionally, no AI released today meets the bar for “Artificial General Intelligence” or AGI. One of the criteria for intelligence is to use reason or logic to work through uncertainty and develop a plan or solution. Humans deal with this type of uncertainty all the time. The AI of today (and the near future) can’t.

It’s still very useful and very powerful. So let’s talk about how to get the most out of it.

Let’s pretend we inherited Bluth’s Original Frozen Banana shop. We know where it is and what it sells, but we want to grow the business. We could read some business books, hire some market research firms, ask our friends, or even hire business consultants, but let's see howChatGPT can help. AI can’t run your business for you, but it can be a fantastic resource!

Shopping List:

In this recipe we’re going to focus on ChatGPT which you can?access here. Let’s start cooking.

Cooking Instructions:

Off the bat we know we need to get more customers to the banana stand. ChatGPT has a few answers:

We can ask it to create content for any of these ads, for our example I went with Google Ads:

We can ask ChatGPT how to accept credit cards:

We don’t have a phone line nor internet at the banana stand. If we give ChatGPT this context it provides a more detailed answer:

You can continue to follow up with ChatGPT to ask about features or pricing, you can even have it make a table for you:

Advertising and credit cards are challenges that we identified. We can ask ChatGPT what it would do:

What ideas do you have for the banana stand?

Follow up questions you could ask:

Analyze the current state of dessert stands on beaches, trends, challenges, and opportunities, including relevant data and statistics. Provide a list of key players and a short and long-term industry forecast, and explain any potential impact of current events or future developments.

Offer an in-depth analysis of the current state of small business legislation and regulations and their impact on entrepreneurship.

What are the US and CA tax implications of inheriting a business?

Additional tips for using ChatGPT:

The most common pitfall with text generating AI is the input prompt. You need to be as descriptive as possible, but don’t go overboard with adjectives or requests.?

These tools will?feel?like they are?talking?to you but behind the scenes remember it’s just doing a lot of complex math. Finding that balance can require many tries and patience. You can always create a new chat if you need to start over.

Once you are on the right path, be ready to go back and forth to further refine the answer or as we did in our recipe, to dig deeper. Be very specific in your follow up. Highlight and direct the software to the parts you want to change or parts that are incorrect. The software needs context.?

It’s up to you to determine the accuracy of the answers. You shouldn’t trust historical or other facts. Remember the software is regurgitating words and phrases that “go together structurally.”?