Q&A with Paolo Sammicheli

Exclusive Trusted Magazine Q&A with Paolo Sammicheli, Entrepreneur, Author, Speaker, Registered Scrum Trainer and Coach.

Scrum in Hardware Development:

What unique challenges do you face when implementing Scrum in hardware development?

Hardware Development differs from software development, where most teams use Scrum, mainly in the "compile" step. For a software team, writing software consists of describing an algorithm with a language that can be transformed into something understandable by the machine through a compiler. Then, you compile the source code, run it, and get feedback on whether it works or not. This coding, compiling, and testing sequence happens multiple times during a single hour

To "compile" a mechanical or electronic design into something testable in Hardware development, you must build it. This requires moving atoms, while in the previous case, you move electrons. Moving atoms is harder and more expensive than moving electrons, so to get a short feedback loop, we need to imagine something more clever than the theoretical linear process design-build-test

How do you handle the longer iteration cycles in hardware development within the Scrum framework?

You handle it by writing a user story about reducing this time and prioritizing it into a Product Backlog. In Scrum, we use the collective intelligence of a cross-functional, self-managing team to solve complex problems. So, it's not Scrum solving this problem; it's a team using Scrum to find the solution. My role when I coach a Hardware team is to let their potential be expressed freely

Can you share a specific example where Scrum helped overcome a significant challenge in a hardware project?

There are so many examples out there. Around 70% of my clients are under NDA, yet I have published a case study every year for the last seven years (https://paolo.sammiche.li/case-studies), so what I can share is the tip of the iceberg. We have public information on teams in IoT, Oil&Gas, Aeronautics, Civil Engineer, Industrial Machinery, Construction, and others. They all reduced time-to-market, improved communication, increased creativity, and talent retention

Scrum in AI Projects:

Artificial Intelligence projects often involve a lot of unpredictability and experimentation.

How do you adjust Scrum practices to accommodate the needs of AI development?

There's no need to adjust Scrum. What people commonly call Scrum is, in practice, the Scrum Guide. This short document contains around 30 Scrum Patterns, such as the artifacts and events. There's a more extensive library of Scrum Patterns, more than 100 (https://www.scrumplop.org). I've used additional patterns to address the specific AI challenges

What role does data play in managing AI projects under Scrum, and how do you integrate continuous learning and adaptation into the Scrum process?

I coined Murphy's Law for Data in many presentations: "Anything that can go wrong about data will go wrong." We treat data as an increment and write stories about the data the Team addresses during the Sprint. Then, we review the data together with the models we train during the Sprint. Ultimately, if Data is the new electricity, it contains value, so we inspect the value earned every Sprint

Integration of Hardware and AI:

How do you synchronize the software and hardware development teams under a unified Scrum framework in projects that integrate hardware and AI, such as robotics or IoT devices?

I used a specific Scrum Pattern called "Scrum of Scrums" that forms a Team of Teams. If a Team is a set of individuals working together and helping each other to reach the Team's goal, a Team of Teams is a set of Teams working together and helping each other to teach the Team of teams' goals. It's fractal. The same patterns we use to coordinate individuals work also for teams. We have scaled patterns such as the Product Owner Team and MetaScrum

Do you recommend any specific tools or techniques for managing the integration of hardware and AI in a Scrum setting?

The tool scenario is still very poor. State-of-the-art solutions are still far from reaching, and big software companies offering off-the-shelf solutions cause more harm than good, in my honest opinion. I suggest evaluating tools based on openness, customization, and easy extension rather than features. Post-its and markers work better in many scenarios, even though this is challenging with distributed teams