AI Agent Builder is designed for business transformation and enables agentic AI integration into apps through a user-friendly, low-code approach. Project done at OutSystems, a B2B low-code software development company.

The main working area of AI Agent Builder and the different interactions possible around it to address concerns such as data privacy and learnability.

Main Challenges

Leading on different dimensions

• Between extended leaves and internal moves, the team went through 5 major changes throughout this project, putting more focus on guaranteeing alignment and direction over time.
  
• The product direction was agreed upon between a leadership quad consisting of UX, Architecture, PM, and Engineering, which required a strong capacity to manage and collaborate with stakeholders efficiently.
• From the very beginning, the pressure of time to release was always present. After release, we shifted to understanding how our customers and users worked with AI Agent Builder. To meet the latter goal, I led a continuous customer engagement program.

Designing for AI (Agent Builder)

• The unpredictable nature of AI makes it hard to plan long-term from a strategic perspective.
• Testing an AI model without real usage is tougher than optimizing interaction and experience.
  
• Worked around assumptions as we were doing some of these things for the first time in our field.
  

Adjusting complexity

• Our target user base isn’t yet comfortable with the more advanced AI concepts we were using.
  
• Tradeoff between sticking to technical terms and tuning down complexity and accuracy.
  
• Make integrations with external AI providers look and sound less scary and more accessible.

Part of a presentation I've done along with PM to R&D about the experience of AI Agent Builder.

Spreading the word

• The image above represents a slide from one of the many times I presented AI Agent Builder to the whole engineering group alongside the PM. The goal was to bring onboard more internal users and more quickly test the product with real users.
  
• Besides more practical presentations, I regularly had to explain the rationale for the product's experience as high up as to the CEO on a few occasions.

Main insights split per product stage based on the customer engagement program.

Supported by real user insights

• Understanding how our users and customers used our product was key to further developing it. Partnering with PM and Strategic Research, I've done a first round of more than 20 interviews to capture user feedback.
• Based on all the data gathered, we reported back to leadership all the insights and potential action items to generate further roadmap.

Other projects

Main Challenges

Leading on different dimensions

• Between extended leaves and internal moves, the team went through 5 major changes throughout this project, putting more focus on guaranteeing alignment and direction over time.
  
• The product direction was agreed upon between a leadership quad consisting of UX, Architecture, PM, and Engineering, which required a strong capacity to manage and collaborate with stakeholders efficiently.
• From the very beginning, the pressure of time to release was always present. After release, we shifted to understanding how our customers and users worked with AI Agent Builder. To meet the latter goal, I led a continuous customer engagement program.

Designing for AI (Agent Builder)

• The unpredictable nature of AI makes it hard to plan long-term from a strategic perspective.
• Testing an AI model without real usage is tougher than optimizing interaction and experience.
  
• Worked around assumptions as we were doing some of these things for the first time in our field.
  

Adjusting complexity

• Our target user base isn’t yet comfortable with the more advanced AI concepts we were using.
  
• Tradeoff between sticking to technical terms and tuning down complexity and accuracy.
  
• Make integrations with external AI providers look and sound less scary and more accessible.

Other projects