AI and hyperautomation in the supply chain for robust software

AI in four phases of the supply chain 

You can deploy AI in all four phases of the software supply chain. 

1. Software design

The software development process always starts with a business idea: is there a way to automate this? It is up to business analysts to figure out how that should be done. AI can help with that in various ways. Firstly, ChatGPT can be used for brainstorming. For example, you can describe your process and ask ChatGPT where there are opportunities for improvement. AI and Machine Learning lend themselves perfectly to analysing large amounts of business data, which you sometimes need in order to support an idea and work out the business case. If developers are unsure which tool to use, they can simply ask ChatGPT, and no longer have to do the Google research themselves. And finally, business analysts can also use this generative AI to write user stories and acceptance criteria, and to check afterwards whether the user stories they have written meet the set criteria. 

2. Software development

You can also use AI in several ways in the development phase. Firstly, you can have AI generate code. GitHub Copilot, developed by OpenAI in collaboration with GitHub, has been around for several years now and works in a similar way to ChatGPT, but with a specific focus on writing code. The tool suggests lines or sometimes even whole blocks of code as you type. Other tooling, such as SonarCloud, uses AI to check code that has already been created for quality and security. You can also use these kinds of tools to check that the application developed does not use a lot of energy unnecessarily.  

If you’re already using generative AI anyway, it’s a small step to write your tests and documentation that way too.  

3. Deployment

New software can be rolled out more efficiently with test automation and hyperautomation. Especially if you want to put new software live several times a day, this will help you reduce the implementation time and minimise the risk of errors. 

4. Management

In the management phase, AI has been making a contribution for much longer with what is called AIOps. Today, this is already often a standard feature of monitoring tooling. AIOps brings together data that was traditionally held in different systems (IT service management, IT operations) and analyses it in context using machine learning. Combined with AutoHealing, this saves administrators a lot of time. You can also use AIOps to predict the amount of resources you will need at what times. And of course AI plays a large role in recognising security incidents. Finally, you can have a chatbot take on some service desk duties. It gets even more efficient if that chatbot combines an employee’s question with specific data about the employee, for example what device they are working on and what previous problems there have been. 

Software design in practice

Several business analysts at AMIS Conclusion use AI agents in the software design process. Software developer Steven Grond explains how this works in practice. “The AI agent starts an interactive question-answer game. This game starts with questions about the functional side of the software to be developed: what problem will it solve? Who will work with it in what situations? And it ends with technical questions about the tool’s position in the landscape and the applications with which it must exchange data. As a business analyst, the completeness and depth of the questions give you a better picture of the customer demand, and also of the technical requirements. So the tool helps you avoid overlooking anything.”

Patrick notes that there is another advantage. “The questions the AI agent asks force you to think a step deeper, and not to follow your ingrained way of thinking. This leads to more creative solutions that might not be immediately obvious.” 

Finally, he points out that this kind of tooling bridges the gap between state-of-the-art technology and business demand. “For example, the agent knows the latest Azure components. Sometimes technical developments move so fast that, as a business analyst, you simply can’t be aware of everything. The AI agent doesn’t suffer from that kind of limitation.”