I’ll start by clarifying that I will not use any AI to generate the following text.
I want to share my current considerations and opinions about using AI in Scrum teams, and specifically for Product Owners.
The first questions that comes to mind is, should we use AI? As an Agile Coach or Scrum Master you are often providing guidance and suggestions to your teams on best practices. I believe we need to be AI aware and do some learning. I am on that path, learning, looking up at a steep learning curve, a new way of doing things, over saturation of information, and classic change resistance. Perhaps you can relate.
Tech over talk?
I don’t find current practices of Product Ownership in need of change per say, so this is an opportunity to be “more effective and efficient”. My first impressions from my learning is AI can speed up the process of creating data output (user stories, estimates, roadmaps, vision statements, priorities etc.) to potentially making decisions faster and more accurate. But there is an investment to be made in choosing and learning tools, writing and editing prompts and analyzing output.
My people centric approach is resisting this idea of tech over talk. The Agile manifesto’s first two words are “Individuals and Interaction”, followed by (we value over) “Processes and Tools”. AI is a tool. Let’s see if it enables human interaction or inhibits it.
AI Transparency
When deciding if you should use AI, after you have learned a little bit about it, I think there are ethical and safety considerations. What are the ethics of using AI at work? Should we tell people when a document or product was created with AI? What parts and to what extent? One of Scrum’s three pillars is “Transparency”, so I vote ‘yes’ – tell people when you are using it. I recently filled in a form where it specifically asked you not to use AI to fill in the answer, and that they could tell if you did.
There are environmental impacts to using AI due to power and water usage for the servers. There are safety/security concerns about where the data is going, what country is it stored in, who could access it. And the ethics about where human checking should be present. By now I am sure you’ve seen something and said “oh ya that’s AI.” For a LinkedIn post the impact is low, but for a contract or customer perhaps much higher.
AI Tools
My early take on AI in Scrum Teams is it’s all about the tools. There are lots of tools, for almost everything a Product Owner or Scrum Team does. It feels like companies are trying to sell you their tool for every flavour of work output. I will list some at the end. Here is a claim by a tool – “AI suggests story point estimates based on historical data and complexity analysis, helping teams make more accurate forecasts.” So this would theoretically make estimation easier, take less time and be more accurate. I am interested because most of my clients don’t have a fully sized backlog, the top reason being that it takes too long. But I wonder what would happen if we tried it? Scrum does say “Inspect and Adapt”, run experiments, do retrospectives, learn, try, adapt. Just because we get the computer to spit out estimates, doesn’t mean we’ve been able to shift from a plan driven organization to a value driven one. So it’s not going to fix everything.
The key element to making most of these tools work, is “prompts”. Ever heard of a “prompt engineer”? It’s the human who writes the instruction to the computer, aka the AI agent. It’s the instructions for the work you don’t want to do yourself. A general template for a prompt you would write could have the following sections; goal for using AI (I want you to prioritize by backlog), roles of AI (you are a Product Owner), context (what data, audience, tone), action you want taken, format for output, provide examples, plus do and do nots). We used to say regarding testing, “garbage in, garbage out”. The key unlock is crafting and adjusting your prompts while being able to analyze the results it produces. These are human elements. If you want to get fancy, you can add agents to check your other agents work.
In conclusion, this is just the beginning, I am starting my steep learning curve, with curiosity and some suspicion. My next steps are to pick a handful of tools and try them out, but not as a replacements for individuals and interactions, but rather as a conversation helper. I will update you later once I am higher up my learning curve. If you want to share your experience let’s connect, I’d love to hear it!
Some tools to consider
- Thematic uses natural language processing to analyze open-ended customer feedback to identify themes and key insights.
- Beam AI uses AI to guide you through a series of questions to ensure your PRD is comprehensive and well-structured.
- Aqua cloud has the ability to generate product requirements from drafts, custom formats, and voice prompts. It also integrates with Jira.
- Productboard has AI-powered features for customer feedback analysis, prioritizing requirements, and planning a roadmap.
- Craft.io enables product teams to collaborate on product requirements, roadmaps, and strategy, along with user story generation and ordering.
- ClickUp Brain – Replace all your software. Every app, AI agent, and human in one place. uses AI to generate PRDs quickly and easily. Build your own agent
- Productroadmap.ai Ensure product ROI with the first AI roadmapping suite that instantly aligns roadmaps with Sales. Identify and prioritize feature gaps based on revenue impact, generate roadmaps, and easily close the loop with Sales.
- HubSpot – Make My Persona
- StoriesOnBoard.com automating user stories, acceptance criteria, documentation
- Otter – meeting assistant, note taker, transcripts
- Jira Rovo – order your backlog