In this second part of my discussion about Chatgpt and its small business, I want to share some small chatgpt tips of small businesses. Also some tricks. Therefore, this publication provides more advice of mechanical advice.
Note: In the first part of this discussion, using chatgpt in a small business, I talked about how we have been using chatgpt in our small CPA company. That publication amounted to a list of tasks and projects for which we have used Chatgpt.
Chatgpt Council #1: Start today
My first advice. Start today. In fact, before reading another paragraph, open your browser, enter chatgpt.com In the address box, click on the registration button and then follow the screen instructions. Register for the option of $ 20 per month to start.
Once you have done this, start using chatgpt instead of the Google or Bing search engine. And start asking the questions you could have asked in the past to a friend, a co -worker or an online forum.
Chatgpt Council #2: Use detailed indications
You know how to use a search engine like Google. And Chatgpt works in the same way. You can write a simple question.
But what do you want to do with Chatgpt? Provide many more details. My typical initial warning probably executes 100 to 200 words. I provide background information. I explain my current understanding. If appropriate, I can cite or cite sources as a propaganda of a statute of the internal income code or a treasure regulation. I also provide instructions for chatgpt. I could ask for it to provide links to your source, for example. Or for a table that summarizes your analysis or your answers.
Think about the instructions you can provide to an entry level employee. Or someone very intelligent but also very new in their work or paper. That is the level of detail or instruction that you and I want to use.
Chatgpt Council #3: ask tracking questions
You will want to ask follow -up questions. Many of them.
For example, you may give your first notice and the Chatgpt response, which did not express things correctly. Or that you were not precise enough in your instructions. Ask ChatgPT for more information. Or to redo your work. Or request clarification about a bit of your answer.
I have found Chatgpt so “human” in his response, that I am inclined to treat him as I would treat a human. But what counts as good manners when interacting with a co -worker or professional? How to not train almost the same question ten times in a row? That is not the right approach with something like chatgpt. Rather with chatgpt, you or I probably want to ask numerous monitoring questions, all based on slightly different words of the facts.
Chatgpt Council #4: Observe errors
Chatgpt makes mistakes. And it is difficult to detect them because he does such an articulated job when he writes his answers. Also, most of the time? It gives us a good answer. I assumed that in the previous blog publication Chatgpt gives the correct answer, perhaps 80 percent of the time. And maybe that’s too low.
But my point is that it makes mistakes. (People call these hallucinations errors, by the way. By the way. Because Chatgpt presents their answers or answers with apparently unwavering confidence). Therefore, you want to be aware and observe this.
By the way, often, when you detect an error, you can explain to Chatgpt why he thinks he made an error. And you will agree quickly. But not always.
This example of the world of fiscal accounting. Therefore, most small businesses can use a deduction of the 199A section to avoid paying federal income taxes in the last 20 percent of their commercial income. However, the Fiscal Law adjusts the deduction for the owners of small high -income companies if they do not pay enough in salaries or if they work in a professional service. Chatgpt cannot work with adjustment formulas. (I have tried to make this a lot. And unless it explicitly provides the detailed steps and calculations, stumble over and over again).
This chatgpt feature to make mistakes (and another large language model, or LLM, AIS to make mistakes) suggests a better practice: for really consistent chats? You and I probably want to continue with issues in which we have a fairly deep knowledge.
Chatgpt Council #5: Curre the chatgpt memory
While you or I work with chatgpt, collect memories of commercial or personal facts that can import for future discussions. Personal information such as their age, education, ethnicity, cultural history, marital status and net assets. Details about the type of business that operates, including its income and profitability.
These memories help Chatgpt do a better job in providing answers and reports. But you probably want to administer and edit the information that Chatgpt collects. Some information can be incorrectly incorrect. Other information may be outdated. And then, some other information can be too personal to allow Chatgpt to use that information to infuse your answers to your questions.
You can do this healing of two ways. First, you can tell ChatgPT to forget the details you don’t want to remember. (“Please ignore the information I provided about the expected income for next year. That information was incorrect”).
Secondly, in the past, although not currently, Chatgpt has alerted him when his memory is full. I mention this here because the characteristic could be available in the future. If you do, you can click on a button, check your list of memories and then remove some memories. Be attentive to this function if it appears. I think it is a good way to clean the memories that are not correct, updated or precise.
Chatgpt Council #6: Use attachments
Chatgpt allows you to attach documents to its indications. You click on the Plus button. Show a dialog box that allows you to point to the file. Select the Microsoft Office document, PDF or other file that should be part of the message.
The attachments allow you to provide a lot of information to ChatgPT, thus avoiding writing a lot of text. In addition, it can include many more data when you use attachments. (Chatgpt not at the time I write this provides any firm recommendation about how much text can you enter the text box. But it suggests that it works and works more easily with attached and formatted documents and spreadsheets that with things copied or admitted to the warning box).
And an advice here: those JavaScript calculators that I have been adding to our blog with the help of Chatgpt? The attachments make it possible. What I do is build a small Excel spreadsheet. Verify the spreadsheet formulas correctly at some calculation set. Then, attach the spreadsheet file to an indicator that asks ChatgPP to write JavaScript that collects the necessary inputs, performs the same calculations and shows the outputs. Then I copy and hit the JavaScript in a blog post.
Chatgpt Council #7: Chatgpt procedure use
I asked Chatgpt if I had advice on how to start using an AI like Chatgpt in a small business.
His answer was long, and I will copy it and hit here. But to generalize, he suggested to create standard procedures to use chatgpt (for example, two -step reviews and key fact verification), add explicit renunciations to deliverables when appropriate, recycle results when possible and standardize best practices.
All good ideas.
Chatgpt Council #8: Consider P (DOOM)
I finish with a brief discussion of P (DOOM), the probability that artificial intelligence creates an existential threat to humanity. However, I will slightly redefine the term here and focus on the possibility or probability that I will take you to make some kind of catastrophic commercial decision or take us on an absurdly dangerous path in our entrepreneurship or administration.
However, before doing that, let me first reveal my default bias about this kind of thing. I am almost always skeptical. Oh really. To give an example that is probably greater enough to not be polarizing, I did not think there was any possibility that the Error Y2K would make the power network fail and the society collapsed. (I heard that concern of numerous friends and very intelligent colleagues, well intentional when the 1990s came to an end).
However, I think we must consider P (DOOM). Given the hallucinations, an AI like chatgpt produces regularly, given the incredibly polished and hypnotically good responses that it provides, and given the supreme trust that shows continuously, we must be careful that it does not send us from a cliff.
A personal example of this: recently as part of a talk about the staff in the face of talent scarcity, private capital firms that face the profession and then adjust the billing rates for inflation, ChatgPT recommended a tactical approach to our company that seemed to be a brilliant gamechanger.
Following your plan, we could apparently improve our profitability. Two cups of coffee later, I was on board.
Another cup of coffee, and I ran into the terrible defects of the tactics. Reduced to its essence, Chatgpt suggested that we increase our prices by 50 percent more than the prices that our capable, intelligent and highly qualified competitors charge. That tactic would not have worked, I’m quite sure. In addition, the irreversibly implemented tactics would have caused our CPA company to implude. What he counts, I think, as AP (DOOM) threatens.
The food to carry then: consider the risk of P (DOOM).
#Small #Business #Chatgpt #tips #tricks
