Artificial intelligence in requirements engineering
Hi, I work in a software company that has been living Scrum for a long time in different DevOps environments, mostly Azure DevOps. My role is that of a PO.
We are increasingly asking ourselves how we can bring artificial intelligence into the requirement engineering processes.
Can anyone here report from experience? I would also be interested to know which tools have been used to good effect and where risks have perhaps also been identified.
We are increasingly asking ourselves how we can bring artificial intelligence into the requirement engineering processes.
I have no experience with this but I would like to ask why you feel artificial intelligence would be better than actual conversations with the people asking for something to be built? What is it that you expect will improve by doing this? Especially since todays "artificial intelligence" is not actually intelligent. It is data analysis. Intelligence is the ability to learn and apply abstract thinking to finding solutions. Computers do not have that ability. Humans do.
We are increasingly asking ourselves how we can bring artificial intelligence into the requirement engineering processes.
In Scrum, requirements engineering is an empirical process through which experiments are conducted in the real world. Validated learning is the intelligent thing to do, and the less artificial it is the better.
My 2 pennies worth…AI is something my organisation, as I’m sure many others, is exploring. We will only know whether AI can contribute to the Scrum framework if we all experiment and make transparent our findings. This is quite an exciting time.
Personally, if it’s discovered that there are benefits AI within Scrum, these should only support the human interactions, but not replace them, after all, would an AI tool share the Scrum values of respect, courage, commitment, focus and openness?
Well, I don't agree that AI is better and can replace the empirical process. But as Richard described, we live in an exciting time where AI is gaining influence everywhere. And I can very well imagine that AI can also flank the role of a PO in a supportive way.
Maybe I'll rephrase my question a bit:
How could artificial intelligence support the activities in the environment of requirement engineering?
In Scrum requirements engineering *is* an empirical process. It is not a separate environment. We learn to build the right thing at the right time.
My hope is that AI will help with testing, and help assure that work is Done. Quality might then become less of a variable, and the validated learning loop would be correspondingly enhanced.
I see AI working better for Support or in development for things like code reviews or testing. But I don't see it playing a role in the Product Management aspects. It can't proxy for actual stakeholders. You *might* be able to use if for market research but I'd see that as an add-on to the work that is already done by humans.
It's an interesting topic. Like Ian mentioned above
In Scrum requirements engineering *is* an empirical process. It is not a separate environment. We learn to build the right thing at the right time.
I don't think AI can simply help on the "empirical" part, but there are places AI shines regarding to you question
How could artificial intelligence support the activities in the environment of requirement engineering?
Suppose you've gathered much requirement or information from stakeholders, users, partners, etc. You may utilize AI to do below things
- Analyze and extract key info so that it may inspire you to do some valuable and creative job.
- When you need to craft the requirement contents, you can use AI to check not only things like grammar error, whether the contents are ambiguous, but also whether some specific requirements are covered or explained well. That can improve the quality of the requirement artifact to some extent.
- If you want some suggestions on which measures you can monitor as the feedback from market, asking AI can be a starting point.
That's currently how we utilize AI in our projects.
One of my friends also mentioned
- Generate flow diagrams using AI based on requirement contents, and the diagrams help people to communicate.
But I never tried it so far. Probably I'll give a try later. Another thing I'd like to try later is to let AI generate test cases for the requirement, or at least to ensure the requirement is testable.
Last but not least, AI is a tool and keep in mind
Individuals and Interactions over Processes and Tools
My two cents worth :)
I think AI is only going to get more important, but in the near future at least will stay a tool, although a powerful tool. As already mentioned testing/QA and code reviews are the current strong points. The other ability of current AI is data analysis, finding trends and make possible predictions. Here the data can be sourced from the sprints like, velocity, issue duration, defects to name a few. AI data analysis might reveal patterns that humans find difficult to find. I agree, I don't see AI taking over PM or replaces jobs in IT scrum teams in it's current form, but it promises at least to become a powerful tool. (We will have to see 5 to 10 years from now if this is still true.)
i can see AI be used for analytics and pattern recognition. and as every one else is saying we have to see what it brings to the table. and it should never replace the human process.