Areas of application of ChatGPT
Use Case 1: Knowledge management and information procurement
One of the supreme disciplines of ChatGPT is explaining concepts. This can be very helpful for IT consultants when gathering knowledge. When answering customer questions or creating an IT architecture, complex IT concepts such as software and hardware components, their functionality and interrelationships can be explained.
- You can summarize your own input if your own text seems too lengthy. Furthermore, long, complicated texts can be summarized for your own understanding.
- If PDFs or entire web pages are to be summarized, Bing Chat can be used, which is also based on the ChatGPT language model. [1]
- A bullet point structure can be specified, which ChatGPT then converts into a continuous text.
- You can even request a bullet-point structure, which you can then convert into a continuous text yourself
As an IT consultant, you may need to familiarize yourself with and understand code for certain tasks. You can either enter code snippets in ChatGPT itself or use IDE extensions such as GitHub Copilot, which integrate ChatGPT directly into development environments such as VSCode. [2]
Use Case 2: Support with modeling
ChatGPT can help with the creation of BPMN processes. Although these are often not detailed enough, they provide a good structure to build on and make it easier to get started with the task.
Here are a few prompt examples:
- What could the tasks for the process of a [process example] look like?
- Which lanes should be included in a BPMN for [process example]?
- Sort the various tasks into the lanes, add further tasks if appropriate.
The same applies to the creation of various UML diagrams, such as the UML class diagram.
Here are a few examples:
- A UML class diagram is to be created for [example]. Which classes should be included?
- The class diagram for [Example] contains the following classes: [classes]. Name possible attributes for the classes.
- What are the relationships between these classes? Also name the relationship types.
Depending on the situation, information that is already known should also be included. As with BPMN processes, it provides a good basis rather than a comprehensive final solution.
Use case 3: Optimization of writing tasks
As an IT consultant, you often have to write descriptive texts, which is where ChatGPT can be particularly helpful as a text-based AI:
- Develop a possible template for user stories
- Specify possible titles of user stories for a situation
- Specify a template and have user stories created based on it
- Formulate user stories for which you only specify the titles of the stories and a rough description
This is also possible in the same way for the creation of use cases or requirements.
Tips for working with ChatGPT
For the first chat request, it is helpful to let ChatGPT take on a specific role and define a target group to which the result should be directed. Furthermore, you should assume that one prompt alone is not enough to get a perfect answer. Prompts should be tactically chained together.
With recursive prompts, you can even leave the work with ChatGPT to the AI itself. In this case, you first divide your task into individual steps, which you then hand over to ChatGPT piece by piece.
Here is a prompt example:
- Prompt:
Create the required steps in a logical sequence for the collection and formulation of all use cases for [project example]. The use cases are intended for the business area and should therefore contain fewer technical details.
- Prompt:
Here is the necessary information for the first step: [Information]
Carry out this step for me.
It is best to ask every few prompts what information is still needed to provide a more concrete context. The additional information provides a better basis for the language model.
What many people don't know is that you can instruct ChatGPT to render the answer in a markup language format in order to insert it directly into HTML, Word or Confluence pages with the correct formatting. Even creating tables is possible.
Here is an interesting internet article for creating tables
Limits and challenges
OpenAI itself is also of the opinion that there are limits to the use of its AI. Therefore, the limits and possible pitfalls are listed here again if ChatGPT is to automate your work. [3]
OpenAI warns of these limits, among other things:
- ChatGPT 3.5 only has information for the period up to the beginning of 2022
- Although ChatGPT sounds very plausible and correct, it can deliver incorrect and nonsensical results. The answers should always be checked again yourself.
- The AI is "prompt sensitive", which means that even small changes to the prompts or identical prompts sent several times can deliver different results. However, this can also be used as an advantage for detailed prompt engineering.
- Depending on the prompt, ChatGPT may think that it does not know the answer, but after small prompt changes, it may still return a supposedly correct answer
- The language style of ChatGPT is often the same in the answers, very verbose and certain phrases are repeated again and again.
- ChatGPT itself does not ask for further required information, the answers are based on assumptions of the prompt. You should therefore take care to define the context well enough yourself.
Conclusion: ChatGPT as support for IT consultants
In general, ChatGPT is more suitable for routine tasks. Creative tasks that require completely new approaches are less likely to fall within its scope, as the answers are based on what is already known. When using such tools, the results should of course be regularly reviewed and adapted. Overall, however, ChatGPT can provide a good basis and support for the work of an IT consultant.
Learn more about AI technologies
Sources:
[1] Bing Chat | Microsoft Edge
[2] https://github.com/features/copilot
[3] openai.com/blog/chatgpt
