Top stories for technology initiatives. June 2023.
As the spring conference season comes to an end, something amazing is happening as we witness the hype surrounding artificial intelligence (AI) and its applicability to the field. As I mentioned in my April column, AI has many transformative uses, and client data comes with privacy risks. Recently, Google announced SAIF (Secure AI Framework) regulations based on his six principles that enable the creation of high-quality and secure artificial intelligence.
We believe there is a high privacy risk to customer data, but some consultants seem to ignore this exposure of customer confidentiality. Thought leaders and novices are working to convince you that AI understands. AI applications are evolving so quickly that it’s hard to imagine all the possibilities.
If you’d like to develop and share your use of AI, please email us at randy@k2e.com. This idea is yours and I will add it to my AI Idea Warehouse. I’m going to develop a cookbook of AI prompts and applications that you can actually apply in your company.
As a reminder, my colleague Brian Tankersley and I have written five papers on AI using ChatGPT4, Microsoft AI, DALL-E and AI competitors, truth and hallucinations in AI, and Large Language Model (LLM) considerations. I recorded a podcast. Be sure to check out these AI podcasts, as well as podcast discussions on various other products on The Technology Lab.
What are the AI trends that have recently surfaced in the accounting space?
Remember, imagination, a methodical approach, and application of AI engine prompts are essential to success. Last month, a product vendor suggested these uses on their platform to me.
- Applying tax planning strategies
- Analyzing Audit Data (Fieldguide.io has released AI on its platform)
- CAS trend analysis of cash flow, inventory and other metrics
- Explain concepts and services in the client’s language and style
- Lease analysis
- Form creation and extraction
- Data interpretation
- Document analysis and search
- Workflow optimization
- Create a macro
- Create RPA
- draft procedure
- draft engagement letter
So far, the best way I’ve found to structure AI prompts is with the accounting maul: a spreadsheet. I created prompts to generate bulleted lists, presentation slides, and other items described below. Remember that most Large Language Model (LLM) generation AI platforms like ChatGPT are like assistants that need to be trained. Providing feedback and guidance helps us get better results.
What are your suggestions for training AI models?
First, please note that my methodology is evolving. However, I created a prompt model that includes a user input control, a ChatGPT control, and a ChatGPT command. Each of these areas was developed to minimize typing. Frankly, I want to talk to ChatGPT, but tools like Speak to ChatGPT, TalkBerry, or Speechify just don’t seem to work for me. But most of the employees can’t understand my voice over the intercom, so I can’t even order from the drive-thru window.
Additionally, we recognize that our K2 organization tends to create CPE materials, so ChatGPT prompts tend to support those efforts, eliminating author block and overcoming blank slate syndrome. please. Each ChatGPT command (prompt) is generated after naming the Input and ChatGPT controls. User input controls include:
Specific slide control
Overall presentation control
The ChatGPT control contains the following controls that can change the specified length.
- Output quantity –> 5-6
- Body length –> 7-10
- Number of characters in title –> 4-6
…ChatGPT commands include:
bullet command
- Long Bullets – Slide Titles Only
- Short Bullets – Titles and Idea Points
- Short bullet points – practical examples
- Short bullet points – facts and statistics
- Give an idea – company/product – features, functions, etc.
- provide an example
- Rewrite and paraphrase existing slides
- Rewrite the last item as a bullet point
Check question command
- Simple review question – title only
- Quick Review Questions – Titles and Ideas
- Review questions about recent conversations
- Review questions on specific topics
generate a table of commands
- Generate tables and headings
- Compare and contrast.headline
- Generate a table.idea
command to generate the graph
Overview slide
- slide title
- Subtitles for slides
- title and subtitle
- Summary/Summary Slides
Presentation overview command
- Introducing Seed Data to Chat GPT
- How to bullet point
- ask for ideas
- Upcoming trends
The purpose of these prompts is to allow ChatGPT to provide results in an immediately useful format. We found that we spent 25% (yes, 1/4) of the time creating content this year compared to last year. In addition, I think that the content is also excellent.
So what can we do to learn AI?
start. Be methodical. Please be specific. Teaches ChatGPT as an assistant. Develop company-wide templates for consistency across teams. Learn from each other.
Everyone’s journey to AI is different. There is no right or wrong way to use tools, but over time you may discover better ways. Creating your own prompt model saves time and gives better and more consistent results. Finally, you may be surprised at how much more efficient your work is and how quickly you can complete items. If you find a better way, please share it with me as I share with you the best ways to use AI.
