For a $XXB organization venturing into AI
Intro
Embarking on the journey to design tools that are reshaping our roles was both exciting and daunting. Curiosity drove me to this project, where I gained invaluable insights—ranging from entrepreneurship lessons to harnessing AI to empower product builders like myself.
5 min read
Overview
From vision to impact: driving scalable AI-powered productivity tools for the future of product development
Driven by entrepreneurial ambition, a group of tech leaders within the organization set out to solve productivity challenges across the software development cycle by developing innovative tools for their teams. This initiative grew into a larger vision: an integrated ecosystem of 12 AI-powered tools, each designed to streamline and optimize key phases of the product development cycle, from ideation to delivery.
I played a key role in prototyping the vision to secure funding, defining the user experience and design language, and leading the development of two AI-powered tools: 'Requirement to Code' and 'Code Translation.' Collaborating with idea authors, I drove execution, ensured alignment, and guided product decisions based on user insights.
Organizations adopting these tools have reported significant productivity gains and streamlined workflows. With a growing pipeline of clients eager to embrace this AI vision, these solutions are set to transform how teams build products and collaborate at scale.
What I did
Buy-in
Demoed the product idea to investors to secure funding
Design & Refinement
Conducted guerrilla user research with developers to refine AI tools
Designed a visual language to break the "AI-thinking" black-box paradigm
Deliver
Collaborated with developers to fine-tune designs for launch
Outcomes
70% faster
Context-aware requirement generation for a large-scale client
13 organizations
Onboarded integrating AI-powered workflows
200+ organizations
Expressed interest about integrating modules into their workflows in the next 6 to 12 months
7 partnerships
With leading platforms and tools in the developer ecosystem, including GitHub, Stack Overflow, Faros, and Codeium
4 AI Models Used
Poolside, StarCoder, OpenAI, Llama
Unify standalone AI tools into a seamless, scalable system to boost productivity tenfold while delivering and integrating them into clients' workflows in 3 months
When I joined the team, the vision was to tenfold developer productivity by infusing AI into the software development lifecycle. But what if we told you that achieving this would require seven separate tools—would you believe it? We needed to unify these tools into one cohesive user experience, with a shared design language and consistent patterns. At the same time, we had to stay responsive to client needs—building based on their priorities while setting the design direction for the ideal future-state experience.
Challenge 1
How might we design a first-of-its-kind GenAI experience with limited access to module owners and minimal time for user research and testing?
As a designer, research is core to my process. When timelines are tight, I get creative in how we gather insights—leveraging quick, targeted conversations with developers and product builders to surface high-impact opportunities.
Key questions I tackled:
What moments in your workflow feel the most cognitively demanding or time-consuming, where an AI-powered assistant could help you stay in flow?
When trying new tools, what signals or behaviors give you confidence that the tool understands your intent and is genuinely adding value to your workflow?
What types of decisions do you prefer to make yourself, versus the ones you'd be comfortable delegating to an AI—especially when you're under tight deadlines?
Approach
Ruthless prioritization, guerrilla user testing, and offline communication to gain alignment.
Despite limited time for user research, designing for developers meant I had direct access to end users. I used guerrilla methods—quick chats, lightweight prototypes, even GIFs shared on Slack—to rapidly validate key design patterns for our AI-powered tools.
This scrappy, fast-feedback approach accelerated the design process while laying the foundation for a scalable, cohesive design language.
Curious about the insight that shaped the design language?
Process Wins I Led:
15+ User testing touch-points
Ran informal Slack conversations with PMs and developers to quickly capture mental models and guide AI workflow integration.
AI ‘process-thinking’ tracker
Validated and standardized a UI pattern showing AI workflow steps—now applied across all GenAI modules.
Daily 20-Minute check-ins with product lead
Aligned on priorities, adapted to feedback, and pivoted fast—enabling responsive design and client-ready solutions like white labeling.
Asynchronous Collaboration Across Time Zones
Shared offline work during NA hours to keep development in India moving and stakeholders aligned around feasibility and direction.
Challenge 2
How might we quickly unify scattered GenAI modules into a cohesive design and user experience, ensuring they are scalable and adaptable to clients' ecosystems?
As new GenAI tools were being developed in parallel, each by different teams, the product experience risked becoming fragmented—making it harder to scale, harder to brand, and harder for users to navigate consistently. The challenge was to create a unified design system that could bring visual and UX consistency across all modules without slowing development velocity or limiting client flexibility.
Approach
Minimal, high-impact design system that was modular and white-label friendly, allowing clients to seamlessly apply their own branding.
To move fast and ensure consistency, I leveraged the McKinsey Design System for rapid prototyping, while exploring scalable UX/UI patterns for all GenAI modules.
The final design language was a modular and flexible, white-label-friendly design system that allows the platform to scale and adapt to different branding needs.
What worked?
Modular design
The design language enables the platform to scale and seamlessly integrate new modules as it evolves.
White labeling friendly
A minimal, white-label-friendly design with subtle accents—like customizable chips and logo highlights—makes it easy to adapt to different branding guidelines.
Challenge 3
How might we gain sponsor buy-in for the future state vision while aligning on the scope for the minimum viable product (MVP)?
When designing complex platforms with both near-term deliverables and long-term potential, it’s easy for sponsors to lose clarity on what’s shipping now versus what's aspirational. The challenge was twofold:
Excite and align stakeholders around the long-term vision so they see the bigger opportunity.
Ground discussions in MVP realities, so the team could move fast without scope creep or misaligned expectations.
Approach
Generating excitement and securing buy-in through rapid prototyping, while aligning scope with the platform's information architecture.
We tackled this through a dual-track strategy: build momentum through rapid prototyping, and anchor decisions using a clear information architecture (IA) map.
Every two weeks, we ran “show & tell” prototype reviews during sprint reviews—walking sponsors through evolving designs that ranged from proof-of-concept (POV) to dev-ready. These live sessions kept excitement high and gave sponsors space to provide feedback in real time, deepening their engagement and confidence in the product vision.
In parallel, we used a high-level IA map to draw a clear line between what was in scope for the MVP and what was part of the future state. This visual clarity helped us guide strategic conversations and manage sponsor expectations while staying focused on delivery.
What worked?
Rapid prototyping
Frequent, high-fidelity prototypes created a continuous feedback loop—sparking interest, validating ideas early, and unlocking more budget as sponsors saw progress and potential.
Information architecture
A simple, well-structured IA map served as a living artifact—visually anchoring the MVP scope while keeping the future state within reach, helping sponsors track progress without losing sight of the big picture.
Read me first
To maintain client confidentiality, wireframes have been modified by removing logos, identifiable branding elements such as colors, and assigning placeholder names. This content is privileged and intended solely to demonstrate my design expertise. All solutions belong to the client. Unauthorized sharing or distribution is strictly prohibited.
Solution
Explore
Some developers struggle to grasp how AI tools could fit into their workflow.
Browsing by role and phase helps them quickly identify where AI can drive impact.
Open prototype
Configure workspace
Developers lack control over key technical configurations when working with AI tools.
Allowing them to customize workspaces—such as selecting their preferred LLM, server, and management platform—empowers them to tailor the environment to their workflow needs and infrastructure.
Open prototype
Performance tracking
Product builders struggle to measure the real impact of AI tools on their productivity and workflows.
A telemetry dashboard helps them track adoption, usage patterns, and performance metrics, enabling data-driven decisions to optimize their workflows and drive adoption.
Open prototype
Collaborative space
Product builders feel disconnected when working with GenAI tools in silos, making it difficult to align on outputs and manage integrations across teams.
A collaborative workspace allows teams to co-create, review, and refine GenAI outputs in one place, accelerating iteration and improving decision-making.
Open prototype
Automating system architecture: "Requirement to Diagram" tool
Engineers and architects often spend hours manually visualizing system architecture from written requirements, leading to misalignment and inefficiencies.
The "Requirement to Diagram" module automates this process by generating clear, accurate diagrams from structured requirements, helping teams align faster and reduce rework.
Open prototype
“Rocío’s ownership of end-to-end design delivery for AI modules and her ability to manage stakeholder feedback effectively have been key to the success of the product.”
McKinsey Digital Partner
“Her expertise in AI and product design, combined with her ability to collaborate seamlessly with developers, has resulted in high-impact, development-ready designs.”
Product Lead for $XXM organization venturing into AI
4 Takeaways
01. Flexibility to build around client needs
03. Maintaining Human-in-the-Loop
01.
Flexibility to build around client needs
We successfully adapted to real client needs adopting the AI-power tools by staying closely aligned with the team deploying the first generative AI modules. Daily communication with the product lead allowed us to be tactical, pivoting from module to module as needed to shift development goals based on client requirements.
Organizations interested in the platform expressed the need for a solution tailored to their organization's branding. In response, we developed a minimal yet impactful design language featuring bubble gradients and a single font and accent color. The main title is easily adaptable to the organization's font and color palette, allowing for seamless integration with their branding.
02.
Visualizing AI’s ‘Black-Box’ Thinking Process
The “Requirement to Code” module was my first end-to-end design for this product. After quick research on tools like GitHub Copilot, we landed on a simple idea: show the AI’s step-by-step reasoning and gather developer feedback.
This transparency clicked—developers said it helped them understand the AI’s process, leading to better adoption and smarter use.
Key Insight:
Transparency is critical for developer trust
To move fast without compromising insight, I conducted high-impact, focused conversations with developers to uncover where AI could truly empower their workflows.
One key insight stood out: developers want to understand how AI is tackling challenges to better optimize their input and align with the tool's output.
Solution:
Clear, transparent AI workflow steps
Each module was designed to outline the steps the AI would take to tackle the workflow. For example, the "Requirement to Code" tool clearly breaks down the process into the following phases:
Import Requirements
Plan
Code
Pull Request
This transparency ensures developers can see how the AI will proceed, which empowers them to make better decisions and feel more in control of the process.
Developer Feedback:
Breaking the black box paradigm
When asked about the step-by-step view, developers shared positive feedback, noting that it breaks the "black-box" paradigm—where AI decisions are opaque and hard to trust.
“It helps me understand what the AI is doing and, more importantly, how I can better provide input to get the best results.”
Developer on the team
03.
Maintaining Human-in-the-Loop
A core design principle in developing these modules was to enhance the human element. By leveraging AI to handle repetitive tasks while keeping humans in the loop for reviewing outputs and providing feedback, we ensured faster learning and iteration, maintaining high quality and efficiency. Additionally, each module includes clear disclaimers, prompting users to carefully review the module's output.
04.
One of the most challenging aspects was developing modules that required 0 to 1 development while working with busy tech leaders. I quickly visualized that waiting for weekly feedback was insufficient to meet deadlines. To address this, I established quick offline channels with the development team, tech leaders, and product lead. This approach accelerated feedback cycles and iterations, ensuring that user flows were refined and ready for sign-off, thereby expediting the development process.
What could be next?
We proved that with the right design patterns, AI can become a true partner in software development—but to unlock its full potential, we must now make it safer, more human-centered, and better integrated with industry needs.
My personal take on what I would explore next:
Human-Centered Guardrails
Build feedback loops for flagging issues, role-based controls, and ethical guidelines to ensure safety, accountability, and trust.
Industry-Specific Integration
Co-create with clients to map workflows, customize generative modules by sector, and connect to enterprise data via APIs.
Continuous Research & Testing
Run user testing, ethnographic studies, and telemetry analysis to fine-tune adoption, experience, and performance.
Collaborative, Transparent Workflows
Enhance workspaces with validation flows, versioning, and real-time team collaboration to ensure alignment and auditability.
Responsible Metrics & Standards
Track misuse, define ethical KPIs beyond speed, and partner with regulators to shape responsible AI practices.
To unlock this potential, three key research questions emerge:
Feedback mechanisms
How do users perceive and interact with feedback loops for flagging incorrect outputs or edge cases?
High-risk spaces
Evolving AI-driven collaboration
Final Thought
By continuously evolving the platform with these strategies, we can not only deliver high-impact generative AI solutions that accelerate development but also build safer, more intelligent, and human-centered technology that adapts to the unique needs of each organization and industry.
To scale AI responsibly, we must design not just for performance—but for trust, collaboration, and adaptability across industries.
Let's talk :)
