回顧會化身地城探索：團隊如何從便利貼白板走到像素酒館

While working on our project and going through N retrospectives, which, by the way, barely differed from one another, our Scrum Master came up with a neat trick: delegating the facilitation of the retro to one of the team members. The next facilitator was chosen at random. And when it was my turn, I teamed up with Cursor and prepared a dungeon for the team. I'll tell you about the dungeon itself and what came out of it in this article.

Let's start with how retrospectives used to be held. Usually, the team was greeted by three boards: "Our Achievements," "What Hindered Us," and "Action Items." With your permission, I won't describe what each board is for, if only because the scheme is trivial for Agile (deviations happen, but nothing critical), and the vast majority of readers (if not all) are already in context. Secondly, the dungeon, and the D&D setting in general, didn't bring anything fundamentally new to this scheme. Before filling out the sticky notes on the boards, the team went through an interactive section. As an activity, we used a board with memes, where each participant, based on their mood (reflecting on the sprint), chose a meme that reflected their feelings and thoughts about the sprint results. In my opinion, it's a fun activity that lets you take the team's temperature and see everyone's individual mindset.

Now let's move to the main narrative. For my turn to facilitate the retrospective, I decided to change things up, but in a way that wouldn't lose the core, main benefit of retrospectives in general. Thus, the idea was born to wrap the retro boards in a D&D theme.

Armed with Cursor and reckless enthusiasm, I started turning the idea into stone... I mean, code. I deliberately chose an AI tool because I wasn't aiming for perfection or technical elegance. The goal was to build a working MVP in a short time, which would allow me to run this experiment on the team and, of course, on myself.

What We Got

We ended up with a virtual tabletop for one evening, working title: Sprint & Dragons. The facilitator (me that evening) sat as the Dungeon Master, and the rest became Seekers. The DM opens the room, drops a four-letter code in the chat, participants join, choose a hero and HP reflecting their mood at the end of the sprint. The classic retro boards are broken down into rooms in the dungeon. There are 6 in total: the tavern, the hall of fame, the trap corridor, the secret crypt, the quest board, and the campfire.

The Tavern is the gathering room, the first place everyone ends up. This is where everyone chooses their class, name, and sets their HP (mood/state) on a scale of 1 to 5. HP was introduced intentionally, as previous retro experience showed that the "weather" in the team was very vague. You could only guess from the tone and content of the sticky notes, and even then, only if the author owned up to them. Here, HP is visible at every table for every participant. It seemed to me that having named scrolls in the room gives both the team and the facilitator a tool to hint at where the emotions are in the scrolls and where the work-related stuff is.

After the gathering phase, the DM can launch an interactive activity for the team to relieve tension and distract them from the tasks they stepped away from to attend the retro. With a click of a button, the DM lets an orc into the tavern. The participants' task is to defeat it. The orc doesn't hit back. You defeat it with whatever is found in the tavern. Everyone picks something from the suggested items in the window, and has half a minute to swap items. Items have their own damage and can give buffs. After the item discussion phase closes, participants take turns rolling a d20. A roll of 1 is always a miss; otherwise, the damage is whatever was rolled + item buffs. I agree that the orc idea might seem naive, silly, or even out of place. The goal of adding it was to replace the usual meme-start with something team-oriented. In my defense, the orc is an option, not a mandatory event. The DM can lead the team straight to the Hall of Fame without a tavern brawl.

The Hall of Fame is a direct analog of the achievements board from the classic setup. In the scrolls, we write exactly what we're used to writing there. Scrolls from here don't go to the quest board, since we don't create action items for such things.

The Trap Corridor replaces the problems board, but not all of it—only a part of it. Here you can write about problems that arose during the sprint, things that hinder moving forward. But you cannot write about issues requiring complex measures. For example: we hit a process that ran into legacy code, and now it's unclear how to allocate resources—whether to dive into refactoring or look for another solution. This is exactly the topic for a scroll in the next room—the Secret Crypt. This is where such questions should be concentrated.

Before moving to the next rooms, voting takes place from the trap corridor and the secret crypt. It's implemented simply: a participant can toss a coin onto the scroll they think needs to be discussed. The facilitator has the option to merge scrolls with similar meanings.

You might ask here—why make two rooms for questions and problems? During the design phase, I thought about how to shift the focus from immediately discussing every sticky note on the classic board to gathering specific problems. The transition between rooms creates a sort of fog of war. At each stage of the retro, we focus exactly on the stage we are at; it's easier for the facilitator to keep the pace, and it's easier for the team to gather what they want to discuss. Moreover, this allows for immediate decomposition of the problem at the application stage, highlighting the verbosity of both the problem itself and its potential solutions.

The Quest Board

Once all the scrolls of secrets and traps are gathered, it's time for discussions. Participants move to the quest board, where action items are assigned for the collected scrolls, and crucially, a responsible person is assigned to execute them. We didn't have this strict requirement before, and it seemed to me that the issues raised by the team would be resolved more effectively if the problem had a name attached to it.

The Campfire

To avoid ending the retro on a dry note of handing out tasks and to send the team off in a good mood, the DM leads everyone to the campfire. This is a place for gratitude; you can shout out a colleague or just say that we're all doing great. To wrap up the retro, the DM seals the chronicle, exporting it as a PDF file, and the dungeon closes for everyone. The next retro will start in the tavern by checking the action items from the previous chronicle.

Classes

I told you about the rooms, but didn't explain the character classes. Actually, they aren't there just to pick your favorite; they are meant to reflect the overall work of a specific participant in the sprint, as they see and evaluate it themselves. So, for example, the Warrior is for the one who fought an elusive bug, the Mage is for the one who researched and learned something new, the Bard is for the one who communicated a lot with the client, ran demos, etc. The character lens indirectly gives an idea of what kind of work was done in the team during the sprint and how it was distributed among the participants.

What We Learned from Running This Retro

Running the retro in Sprint & Dragons at least showed that the team was craving a breath of fresh air. The unusual format and fantasy setting helped shift everyone's approach to the retro from "just another task." Multi-layered problems started surfacing earlier. The discussion became more sequential and, by feel, more thorough. By the next retro, we know exactly who to ask about the progress of action items. We haven't yet been able to verify whether this format successfully shifted the focus from the person to the problem. One of the ideas was that the scrolls shouldn't become accusatory cards, though the facilitator keeps an eye on that anyway.

We will definitely try running retros with this tool again, going into dungeons with the team to improve our sprint-over-sprint results. We'll accumulate experience, improve the tool, and test hypotheses. As soon as new features and interesting observations appear, we'll meet in the next part of the story. Until then, all the best, and thank you for reading to the end. Feel free to leave your thoughts in the comments.