Stelian ROMANProject Manager| MicroSafetyCarlingford, New South Wales, Australia
Agile ways of working are evolving rapidly, and artificial intelligence is at the centre of this transformation. Teams are increasingly turning to AI-powered tools for estimation, backlog prioritization, and even code generation. While these capabilities promise efficiency and objectivity, they also introduce new tensions and ethical questions into the decision-making process.
How is your team integrating AI into Agile practices? What questions are you asking about trust, accountability, and ethics?
Program Manager| HARPER SRLSanto Domingo / Distrito Nacional, Dominican Republic
hello Stelian ROMAN Right now, my teams are mainly using AI for things like code generation, code validation, repository management support, security-related checks, and similar technical activities.
One thing we’ve tried to keep very clear is that there is always a human in the loop, especially for decisions that can impact quality, security, architecture, or business logic.
Some of the questions we’ve been discussing internally are:
How do we validate AI-generated code before it reaches production?
Where should accountability remain fully human?
How do we avoid overdependence on generated outputs?
How do we make sure security and ethical considerations are not bypassed for speed?
I'm still exploring this more at a personal workflow level than as a fully embedded team practice, but I already see how valuable AI can be in supporting Agile ways of working.
From my experience, AI is really helpful for summarizing discussion points, refining user stories or action items, preparing options before making a decision, and spotting risks or assumptions that might not be immediately obvious. In a project delivery setting, I also believe AI has great potential to support improved reporting, capturing lessons learned, and having early warning discussions.
However, I think the accountability aspect is very important. AI can assist with the thinking process, but it shouldn't replace the team’s judgment. Someone still needs to review the output, challenge the assumptions, and decide what best fits the real-world context.
My questions would be: What data are we relying on? Who checks the AI’s output? Which decisions are too critical to delegate? And are we using AI to enhance collaboration or just to move faster? Saving Changes...
Stelian ROMANProject Manager| MicroSafetyCarlingford, New South Wales, Australia
Lissette Indhira Pimentel Sosa , thank you for responding to my question. Decades ago, when I managed software development teams, I was against code-generation tools because we spent more time fixing the code than we saved by using them. There was also the issue of losing our development skills and using patterns defined by others. Times and tools are different, and I agree that AI has a place in our toolset. As you said, AI is a tool, not a decision-maker, and not even a replacement for human developers. What looks easy when the code is new will become a nightmare for maintenance. Teams also miss the opportunity to learn, to 'uncover new ways' of building a product. Saving Changes...
Stelian ROMANProject Manager| MicroSafetyCarlingford, New South Wales, Australia
Aung Sint, thank you. Quality and volume of data are crucial when we introduce AI. Some people believe that AI is about automation; it is not. AI should generate recommendations based on the available information. Generative AI is still in an infant stage; it can create interesting pictures, although in my experience, AI doesn't understand the message, and the result will be a distortion of what we initially intended to communicate. The risk is that we will pass the AI results as they were, either because of time or workload or because we don't spot the problems. Saving Changes...
Ming YeungAdjunct Professor| Various academic institutesToronto, Ontario, Canada
Dan, your piece is timely and relevant. AI is becoming an increasingly natural part of Agile delivery, but integrating it responsibly requires more than simply adopting new tools. In my experience, the most important shift has been treating AI as a collaborator, but not a decision‑maker. We use AI to support estimation, refine backlog items, and generate early drafts of technical solutions, but the team remains accountable for every decision. That distinction matters, because AI can accelerate work, yet it can also introduce bias, hallucinations, or false confidence if left unchecked. One of the biggest questions we’ve been wrestling with is trust. Not “Can we trust the AI?” but “Under what conditions should we trust it?” We’ve learned to validate AI‑generated insights the same way we validate any other input: through peer review, experimentation, and customer feedback. When AI suggests a prioritization change or flags a dependency, it becomes a conversation starter, not a directive. Ethically, we are also asking harder questions: -- Who is accountable when AI‑generated code introduces defects? -- How do we ensure transparency when AI influences decisions that affect scope, cost, or risk? -- How do we prevent AI from reinforcing existing biases in our data or processes? The teams that navigate this well treat AI as a tool that enhances human judgment, not replaces it. For us, success means using AI to improve flow, reduce cognitive load, and spark creativity while keeping responsibility, critical thinking, and ethical awareness firmly in human hands. Saving Changes...
Community Champion
