Using Rovo and JPD for formulating a task description

[Tatiana Torset]

Hello everyone,

I am a bit experienced with JDP, but have not yet started with Rovo, so forgive me if the question is somewhat naive. What I am currently exploring is a possibility of simplifying  the following workflow: 

1. Let's say, we agreed that we need a certain change in our platform behaviour, something like, for example "Allow an admin to upload a spreadsheet and the system to adjust settings for the selected user according this spreadsheet" (the real task will be more precise, but I am a bit vague here for obvious reasons).

2. The product designer formulates the core behaviour that she expects.

3. Then we normally have a walkthrough with developers and are discovering some questions that we need to clarify. "What if this field is empty?", "What should we expect the system to do if the values in the field A and the field B are not the same" etc - and these are not just random "what ifs", the constraints or ambiguities that we see but based on the current system behaviour and how exactly it is achieved in the code. 

4. Then we clarify what we want the systems to do in these cases, often involving the "domain side", or the "business side".

I wonder if it is possible to discover some (if not all) of the requirements that are not yet described in the ticket using a tool (AI agent or something like this) that has access to the source code? So that at least some parts of the step 3 are done earlier and/or easier so that the designer does not have to remember all the details of the systems behaviour and developers don't need to dived in and discovered these questions? 

I will be happy to hear from you how you solve this.

[Tatiana Torset]

Thank you, Walter!

Knowing that it is doable in principle is very encouraging.

I wonder if anyone has tried applying this in practice. I mean, really implemented something similar as a part of their regular workflow.

For example, when I was starting with JPD, one of the most helpful things that I saw was a webinar where Tanguy Crusson showed how they used JPD to design and develop JDP. Looks like now I am trying to research how to use JPD in combination with Rovo. Has anyone come across such resources or maybe can share own experience?

[Walter Buggenhout]

Hi @Tatiana Torset,

This is a great and well thought-out question. I rarely refer to AI tools to help out, but since your question is very specific as to what Rovo / AI might be able to do, I would suggest adding your question (with full context as you provided above) into Rovo chat and see what it returns.

I did it myself and it came back with pretty decent information and suggestions as how you could approach this and what you can expect. I am not going to return the full AI generated answer, but in short:

• an AI agent with access to a codebase is pretty good at recognising patterns and should be able to perform some form of gap analysis up front;
• given a designers prepared descriptions, it should be able to come up with things like:

  • Fields or states the description doesn't mention but the code handles

  • Permissions or role checks that exist in related features but aren't mentioned

  • Side effects (notifications, audit logs, cache invalidation) that similar operations trigger

• It gets harder when complexity increases, e.g. when you want to address behaviour across multiple systems, want to include desired behaviour or when you are interested in things the team knows, but has not documented ...

Keep in mind that you should scope your agent and questions narrowly. Don't try to let it scroll through your entire codebase, but try to ask targeted questions instead. And see it as your assistant that gets you started, but remain responsible of your own decisions and analyse the agent's feedback with a critical mindset.

Hope this helps!