AI Agent Builder
AI Agent Builder is designed for business transformation and enables agentic AI integration into apps through a user-friendly, low-code approach.
It allows businesses to build, configure, and deploy AI agents without deep technical expertise, bringing the power of large language models and external AI providers directly into the OutSystems ecosystem.
Role
User Experience Lead
Led the AI Experience team at OutSystems, responsible for multiple projects including AI Agent Builder. Reporting to UX management, I managed a multidisciplinary design team from concept to release, overseeing design direction, team wellbeing, and the quality of output throughout the full product lifecycle.
Worked on the strategic direction with product management, architecture, and engineering, translating business objectives and technical constraints into user-centred design decisions. I presented the product design and strategy multiple times to the CTO and CEO, communicating design rationale in terms of user impact and business outcomes.
Team
Reporting to
UX Manager
Working with
Product Manager, Lead Architect, Engineering Manager
Managing
Product Designers (3), Content Designer, UX/UI Designer, UX Researchers (2)
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
The product direction was agreed upon between a leadership quad consisting of UX, Architecture, PM, and Engineering. This required ongoing stakeholder management across functions with different priorities, and the ability to build consensus around a shared product vision without sacrificing user needs.
From the very beginning, the pressure of time to release was always present. After release, we shifted focus to understanding how customers and users actually worked with AI Agent Builder. To meet that goal, I led a continuous customer engagement programme that turned real usage into actionable insight.
Between extended leaves and internal moves, the team went through 5 major changes throughout this project. Maintaining alignment and a consistent design direction through that level of organisational change required deliberate communication practices and strong documentation of design decisions and rationale.
Designing for AI (Agent Builder)
The unpredictable and rapidly evolving nature of AI makes long-term strategic planning harder than in conventional product work. Decisions that seemed sound one quarter could be overtaken by new model capabilities or shifting industry norms.
Testing an AI-powered product without real-world usage exposes the limits of controlled research methods. We had to build tolerance for ambiguity into our design process, validating assumptions through progressive exposure rather than upfront certainty.
Many design patterns for agentic AI were being established industry-wide at the same time we were building. We were often navigating genuinely new territory, which required both confidence in foundational UX principles and openness to iteration.
Adjusting complexity
Our target user base, business developers and IT professionals in enterprise settings, was not yet fully comfortable with advanced AI concepts such as agent orchestration, LLM configuration, and third-party model integration.
We had to make active tradeoffs between technical accuracy and accessibility: using the right terminology to maintain credibility with technical stakeholders, while reducing cognitive load for users who were adopting agentic AI for the first time.
Making integrations with external AI providers feel less opaque and more trustworthy was a recurring design challenge. This meant rethinking how we communicated system behaviour, data flow, and configuration options in a way that built user confidence rather than creating friction.
The Experience of AI Agent Builder
There are always multiple possibilities, but this is the intended happy path.
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 more internal users on board and accelerate real-world product testing before broader rollout, a key part of building internal advocacy for a product that required cross-functional buy-in.
Beyond practitioner-level presentations, I regularly communicated the rationale behind design decisions to senior leadership, including the CEO. This meant framing design choices in terms of product strategy and business impact, not just interaction quality.
Main insights split per product stage based on the customer engagement program.
Supported by real user insights
Understanding how users and customers actually worked with the product was central to how we developed it post-launch. Partnering with PM and Strategic Research, I designed and ran a first round of more than 20 user interviews to capture structured feedback across different user segments and use cases.
The insights gathered fed directly into a leadership report covering key findings, identified friction points, and recommended action items, giving the wider organisation a clear evidence base for roadmap prioritisation. This continuous engagement model kept user needs visible at the strategic level, not just within the design team.
Other projects
AI Agent Builder
AI Agent Builder is designed for business transformation and enables agentic AI integration into apps through a user-friendly, low-code approach.
It allows businesses to build, configure, and deploy AI agents without deep technical expertise, bringing the power of large language models and external AI providers directly into the OutSystems ecosystem.
Role
User Experience Lead
Led the AI Experience team at OutSystems, responsible for multiple projects including AI Agent Builder. Reporting to UX management, I managed a multidisciplinary design team from concept to release, overseeing design direction, team wellbeing, and the quality of output throughout the full product lifecycle.
Worked on the strategic direction with product management, architecture, and engineering, translating business objectives and technical constraints into user-centred design decisions. I presented the product design and strategy multiple times to the CTO and CEO, communicating design rationale in terms of user impact and business outcomes.
Team
Reporting to
UX Manager
Working with
Product Manager, Lead Architect, Engineering Manager
Managing
Product Designers (3), Content Designer, UX/UI Designer, UX Researchers (2)
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
The product direction was agreed upon between a leadership quad consisting of UX, Architecture, PM, and Engineering. This required ongoing stakeholder management across functions with different priorities, and the ability to build consensus around a shared product vision without sacrificing user needs.
From the very beginning, the pressure of time to release was always present. After release, we shifted focus to understanding how customers and users actually worked with AI Agent Builder. To meet that goal, I led a continuous customer engagement programme that turned real usage into actionable insight.
Between extended leaves and internal moves, the team went through 5 major changes throughout this project. Maintaining alignment and a consistent design direction through that level of organisational change required deliberate communication practices and strong documentation of design decisions and rationale.
Designing for AI (Agent Builder)
The unpredictable and rapidly evolving nature of AI makes long-term strategic planning harder than in conventional product work. Decisions that seemed sound one quarter could be overtaken by new model capabilities or shifting industry norms.
Testing an AI-powered product without real-world usage exposes the limits of controlled research methods. We had to build tolerance for ambiguity into our design process, validating assumptions through progressive exposure rather than upfront certainty.
Many design patterns for agentic AI were being established industry-wide at the same time we were building. We were often navigating genuinely new territory, which required both confidence in foundational UX principles and openness to iteration.
Adjusting complexity
Our target user base, business developers and IT professionals in enterprise settings, was not yet fully comfortable with advanced AI concepts such as agent orchestration, LLM configuration, and third-party model integration.
We had to make active tradeoffs between technical accuracy and accessibility: using the right terminology to maintain credibility with technical stakeholders, while reducing cognitive load for users who were adopting agentic AI for the first time.
Making integrations with external AI providers feel less opaque and more trustworthy was a recurring design challenge. This meant rethinking how we communicated system behaviour, data flow, and configuration options in a way that built user confidence rather than creating friction.
The Experience of AI Agent Builder
There are always multiple possibilities, but this is the intended happy path.