They could simplify the API integrations and allow us to automate certain tasks using our own server without just logging into the web page. They can also improve the scope of CoPilot.
One drawback is that the solution sometimes suggests unwanted code, especially if I accidentally press the tab. This doesn't happen often. Sometimes it seems to understand my code, but other times it doesn't. This inconsistency is confusing. I suggest adding a feature to write code directly in the IDE from chat responses instead of copying and pasting. GitHub Copilot can generate code based on method comments in the IDE, but chat responses require manual copying. A feature flag to allow direct writing from chat to IDE would streamline the process. The product should also reduce its price so that it can improve the community engagement.
They could improve the product in terms of integration with other tools. Additionally, they should offer more robust support for highly regulated environments like finance or healthcare to enhance its applicability.
The problem arises when a bot is not well-designed, which frustrates customers. For example, with our local Comcast cable company, if you need to repair something, you go on the app. They ask, "What's wrong?" You say, "My power cord melted. I need a new power cord." But the bot responds, "Oh, your Internet's not working?" So, it doesn't recognize unusual issues. Then, they make you go through troubleshooting, saying, "Your Internet looks good, but your device isn't plugged in." It's like, "Yes, I know. I want to speak to an agent." And it takes another ten minutes of trying different things before you finally get to an agent. This is an example of a poorly designed bot.
Speaking about the code suggestion capabilities of the product, I would say that though it is good, there is a need to develop it further. In our company, when we have more dependent classes opened, we realize that GitHub CoPilot needs to be improved. GitHub CoPilot needs to analyze the whole package and try to give more relevant codes. If the product plans to provide more extensions, then adding Eclipse would be helpful. There are no extensions provided by the product for Eclipse. I wish the context window gets expanded, as it would be good for the product, especially since I am currently using some custom packages and custom classes. GitHub CoPilot needs to give some generic code generation instead of analyzing my classes and giving relevant codes. Apart from the aforementioned area, there are no major challenges in the tool.
I have not used the product much to be able to comment on what needs improvement in it. With the passage of time, I believe that there will be improvements in its prompt response, like ChatGPT, considering that GitHub will get more data. With time, the product will mature. The product's price is an area of concern where improvements are required. ChatGPT is free, and people would rather prefer it over GitHub CoPilot. Considering the way GitHub CoPilot works, it can improve its connection with technologies like SQL. Right now, I can't write an AWS Glue job in GitHub CoPilot. In the future, if the tool can improve and include multiple technologies and languages, I see that they have a scope to offer better scalability.
In certain instances, OpenAI didn't respond in the expected way. It was in the Power Automate. The responses were more general and didn't address the specific point. There were significant limitations. Typically, the queries I've used it for are relatively small in scope.
In a few cases, the results aren't correct, so that needs improvement. Also, it would be great if the results could be presented in different formats, not just text. As engineers, it takes time to read through text-based results. If they could provide diagrams or basic design architectures, it would be easier to understand and take less time to review. So, as of now, image functionality is not provided by GitHub Copilot, and it would be great if they could add that. So, I would like to see improvements in the result generation, more results, different result formats, and image functionality.
Technical Program Manager at a healthcare company with 10,001+ employees
Real User
Top 5
2023-11-07T09:45:48Z
Nov 7, 2023
There's room for improvement to ensure that suggestions align more precisely with the context of what I'm seeking, minimizing instances of unrelated or inaccurate code suggestions. It could benefit from enhancements in several key areas. Firstly, there's a need for improved integration across different Integrated Development Environments to ensure a consistent user experience. Secondly, providing users with more granular control over when Copilot is active, preferably on a per-project basis, would enhance flexibility. Also, broadening Copilot's contextual awareness beyond open files to consider the entire project or codebase context could lead to more accurate and relevant suggestions, particularly in scenarios where opening all files is impractical. The current state lacks maturity, and reporting is not readily available out of the box. Users have to request specific reports, involving collaboration with the data team, indicating a need for more streamlined and accessible reporting features.
They could simplify the API integrations and allow us to automate certain tasks using our own server without just logging into the web page. They can also improve the scope of CoPilot.
One drawback is that the solution sometimes suggests unwanted code, especially if I accidentally press the tab. This doesn't happen often. Sometimes it seems to understand my code, but other times it doesn't. This inconsistency is confusing. I suggest adding a feature to write code directly in the IDE from chat responses instead of copying and pasting. GitHub Copilot can generate code based on method comments in the IDE, but chat responses require manual copying. A feature flag to allow direct writing from chat to IDE would streamline the process. The product should also reduce its price so that it can improve the community engagement.
They could improve the product in terms of integration with other tools. Additionally, they should offer more robust support for highly regulated environments like finance or healthcare to enhance its applicability.
The problem arises when a bot is not well-designed, which frustrates customers. For example, with our local Comcast cable company, if you need to repair something, you go on the app. They ask, "What's wrong?" You say, "My power cord melted. I need a new power cord." But the bot responds, "Oh, your Internet's not working?" So, it doesn't recognize unusual issues. Then, they make you go through troubleshooting, saying, "Your Internet looks good, but your device isn't plugged in." It's like, "Yes, I know. I want to speak to an agent." And it takes another ten minutes of trying different things before you finally get to an agent. This is an example of a poorly designed bot.
Speaking about the code suggestion capabilities of the product, I would say that though it is good, there is a need to develop it further. In our company, when we have more dependent classes opened, we realize that GitHub CoPilot needs to be improved. GitHub CoPilot needs to analyze the whole package and try to give more relevant codes. If the product plans to provide more extensions, then adding Eclipse would be helpful. There are no extensions provided by the product for Eclipse. I wish the context window gets expanded, as it would be good for the product, especially since I am currently using some custom packages and custom classes. GitHub CoPilot needs to give some generic code generation instead of analyzing my classes and giving relevant codes. Apart from the aforementioned area, there are no major challenges in the tool.
The tool needs to focus on integration, as it is the most important aspect. I would like to see some pre-designed modules included in my projects.
I have not used the product much to be able to comment on what needs improvement in it. With the passage of time, I believe that there will be improvements in its prompt response, like ChatGPT, considering that GitHub will get more data. With time, the product will mature. The product's price is an area of concern where improvements are required. ChatGPT is free, and people would rather prefer it over GitHub CoPilot. Considering the way GitHub CoPilot works, it can improve its connection with technologies like SQL. Right now, I can't write an AWS Glue job in GitHub CoPilot. In the future, if the tool can improve and include multiple technologies and languages, I see that they have a scope to offer better scalability.
In certain instances, OpenAI didn't respond in the expected way. It was in the Power Automate. The responses were more general and didn't address the specific point. There were significant limitations. Typically, the queries I've used it for are relatively small in scope.
In a few cases, the results aren't correct, so that needs improvement. Also, it would be great if the results could be presented in different formats, not just text. As engineers, it takes time to read through text-based results. If they could provide diagrams or basic design architectures, it would be easier to understand and take less time to review. So, as of now, image functionality is not provided by GitHub Copilot, and it would be great if they could add that. So, I would like to see improvements in the result generation, more results, different result formats, and image functionality.
There's room for improvement to ensure that suggestions align more precisely with the context of what I'm seeking, minimizing instances of unrelated or inaccurate code suggestions. It could benefit from enhancements in several key areas. Firstly, there's a need for improved integration across different Integrated Development Environments to ensure a consistent user experience. Secondly, providing users with more granular control over when Copilot is active, preferably on a per-project basis, would enhance flexibility. Also, broadening Copilot's contextual awareness beyond open files to consider the entire project or codebase context could lead to more accurate and relevant suggestions, particularly in scenarios where opening all files is impractical. The current state lacks maturity, and reporting is not readily available out of the box. Users have to request specific reports, involving collaboration with the data team, indicating a need for more streamlined and accessible reporting features.
GitHub CoPilot's stability is an area of concern where improvements are required.
Some of the suggestions provided by GitHub CoPilot are not accurate, making it an area of concern where improvements are required.