Transform their existing Figma plugin for streamlining design workflows by leveraging AI to instantly generate actionable tasks and design feedback.
Explored prompt engineering, chaining and fine-tuning of LLMs (LlaMA-v2, OpenAI’s GPT-4) on data from Figma, Slack and Google Docs to automate task management and design reviews.
Tintash’s engineering efficiency and AI expertise helped the DesignPro team develop a quick prototyping and research pipeline, allowing the startup to validate their product ideas with real customers.
DesignPro approached Tintash to help transform their design platform into an innovative AI-driven task generation system. A groundbreaking Figma plugin, DesignPro was created to streamline the way design teams handled tasks. The platform’s creators had recognized a common need among designers for a more efficient and organized workflow within the Figma and Figjam platforms.
Designers often grapple with scattered tasks, notes, and communications across various channels. This fragmentation poses a significant hurdle to maintaining a focused and organized design process. While the plugin they had created was able to streamline these workflows, the DesignPro team was now looking to leverage AI to automatically transform those comments, notes, documents, messages and video calls instantly into actionable tasks.
By partnering with Tintash, they sought to integrate AI into their existing platform, to efficiently prototype and test these new ideas that would make task organization 6-8 times faster.
I have worked with some of the best AI talent in the world at companies like Siri and Apple. When it came to building an AI team for my own startup DesignPro, I relied on Tintash to find me talent with excellent research, engineering and communication skills. I am beyond impressed with their talent pool and team’s ability to do exceptional work at an affordable price.
Mohammad Abdoolcarim
Co-Founder, DesignPro
Our cross-functional team comprising web frontend and backend developers, project managers, and AI engineers managed to rapidly prototype new features and deploy them for quick testing.
The Agile Process
We followed an agile process where we would explore new product ideas in every sprint. During the initial phase of the project, we would implement a new prototype every other week, test it with real customers, and then quickly move to the next iteration. This process allowed the client to diligently experiment different product ideas with target customer segments using live prototypes and demos.
After achieving the initial product-market fit, during the next phase, we developed and deployed polished products in a more scalable and flexible way. Post deployment, we again switched back to the prototyping phase, developing and changing many of the features based on customer feedback and requirements.
LLM Integration for Task Synthesis
The first feature we developed and deployed was a design workflow management system. Our team adeptly integrated Large Language Models (LLMs) to convert unstructured feedback into structured tasks, akin to Jira.
The feature allows designers to import feedback from their clients on the platform. Our solution then analyzes that feedback using LLMs to create a Kanban task board for designers that also gets updated automatically. In addition, it uses AI to figure out task priorities, groupings, labels and tags, assignees and assigners, and other helpful details to enable designers to seamlessly manage their workflows.
Since a lot of this required the use of generative AI, our team decided that the state of the art LLM offering by OpenAI was best suited to deliver the required results. In order to work efficiently with ChatGPT via the OpenAI API, we utilized Python’s LangChain framework so as to design effective prompt templates.
By leveraging this cutting-edge technology, we were able to dramatically reduce task generation time from minutes to seconds. The integration of LLMs impressed clients with its accuracy and speed, resulting in enhanced customer engagement and satisfaction and a rapid growth in DesignPro’s user base.
AI-Powered Design Feedback Generation
The second major feature we developed for DesignPro was a design reviewer which could autonomously analyze UI designs in Figma and provide actionable feedback. This allows designers to concentrate on creative work while the system automatically generates precise critiques.
This solution uses LLMs and AI trained on best-practice frameworks such as Nielsen Norman and IDEO to generate useful feedback for designers. Along with this proprietary dataset of design criteria, we use concepts like prompt chaining and retrieval-based LLMs to get more control over the quality and type of feedback the LLM models generate.
This AI feature gained DesignPro a lot of traction with customers and investors alike. Successful demonstrations of our AI's capability to generate precise feedback attracted immediate interest at a designers' conference.
Extensive and Secure Tool Integration
Our dedicated integration team continues to incorporate tools such as Figma, Slack, Zoom, Loom, and Google Docs into DesignPro. This adaptability is key to scaling the product and reaching a broader audience, and has been pivotal in convincing designers to incorporate DesignPro into their workflows. The team ensures a seamless integration of these tools together with robust data security to enhance the user experience.
DesignPro is able to analyze design requirements and feedback and use AI to assist designers manage their workflows. This OpenAI ChatGPT-enabled Figma plugin allows designers to optimize multiple aspects of their design journey:
- Streamline workflow: DesignPro can create tasks or use AI to transform comments, notes, docs, Slack threads and video calls into actionable tasks on a Kanban board. These tasks are linked to frames, making it easy to navigate designs.
- Organize work faster: DesignPro automatically assigns tasks with labels, priorities, frames and collaborators. It applies filters to focus on tasks for specific labels, pages or sections, ensuring that the most important work is finished first.
- Search designs with ease: DesignPro brings together tasks from multiple data streams, such as Figma, Slack, Google Meet and Google Docs into a central hub users can search. It can also search the design library by label, priority, and date.
The project’s success came down to the productive collaboration with the DesignPro leadership through all stages of the development lifecycle. This included the initial AI roadmap planning and development to designing a user-friendly and intuitive interface for the plugin. Our AI experts stepped in to provide the client with AI thought leadership, LLM fine-tuning and R&D, interfacing with OpenAI services via the LangChain framework, and prompt engineering, all this to go along with our backend development, UI and UX design, project management and quality assurance services.
