Agile Methods

Before an AI agent writes a single line of code or drafts a blog post, it has to be clear what should actually be built and what shouldn’t. This is where agile methods come in: Scrum with the Product Owner role, the Lean Canvas and the principles of Design Thinking.

Scrum and Product Owner: hands-on experience

I have worked with Scrum for many years. I earned the Professional Scrum Master I certification from Scrum.org in 2012, and the Professional Scrum Product Owner I followed in 2020. Most of what I learned came from practice, though. Since 2018 I’ve been at MAN Truck & Bus, currently as D&A IT Spoke Lead. I started there as Product Owner for a data platform. During that time I saw that the role is often misunderstood. If you think product ownership means managing tickets, you miss what the job is really about.

The real work of a Product Owner is saying no and slicing requirements cleanly. Every new idea sounds reasonable at first, which is exactly why someone has to decide what delivers real value now and what can wait. It is just as important to break large initiatives down so that every step ends with a result that can be reviewed and accepted. A well-sliced backlog item is small enough to stay manageable and large enough to deliver visible value.

Product Owner thinking for teams of AI agents

Today I mainly use this mindset in the Ideate phase of my multi-agent systems projects. Before an agent gets started, I validate the idea, break it down into verifiable increments and set priorities. This works well because in one respect AI agents are hardly different from human developers: they also need clear assignments and unambiguous acceptance criteria. A vague assignment produces an unusable result from an agent just as it does from a team.

My experience as a Product Owner helps me write assignments that have a clearly defined goal and can be assessed objectively at the end. Saying no is still the most important discipline: not every task an agent could do is a task it should do.

Lean Canvas in practice: business planning for ketovavaria.de

I use the Lean Canvas for the business planning of ketovavaria.de, a blog run by my team of AI agents. The method was created by Ash Maurya and is derived from the Business Model Canvas. It captures nine fields on a single page: problem, customer segments, unique value proposition, solution, channels, revenue streams, cost structure, key metrics and unfair advantage.

Its strength is how much it condenses. The canvas shows at a glance what problem the blog solves, who it is for and what value it offers. When several agents split up the work, this shared foundation is essential. Each agent can align its task with the same description of the target audience and value proposition instead of making its own assumptions. I make a point of not treating the Lean Canvas as a one-off document. It is a working draft that I revise whenever new insights come up.

Design Thinking: principles in practice, formal format as the next step

I have applied the core principles of Design Thinking for years, just not under that name. As a Product Owner, every good requirement starts with genuine user understanding and a precise problem definition, exactly the first two Design Thinking phases. In 2020 I formalized that further with the openHPI course “Mastering Design Thinking in Organizations”. What is still missing is the formal workshop format as a standalone project approach, which is why it sits at “Assess” on my technology radar.

Design Thinking is divided into the phases understand, define the problem, ideate, prototype and test. These phases are not worked through in a straight line but iteratively: findings from a test can lead back to understanding or to a new problem definition. The first two phases in particular overlap with my practice as a Product Owner: if you have truly understood a problem and defined it precisely, you can write much better assignments, for development teams and for AI agents alike. The next step is turning that existing understanding into a structured workshop format once a suitable project comes along.