Accern.Rhea Full Case | Lazarev.
About the project 🤗
Challenge:
Kumesh Arrogant, a co-founder of Accern set on a mission to solve the problem of unstructured financial data. Lazarev. engaged in the design of AI-powered solutions for Accern back in 2018.
Fast forward to 2023, Kumesh approached our team to develop a sophisticated tool for financial researchers, VC investors, ESG analysts and others, leveraging the capabilities of a pre-trained AI model that extended beyond conventional chatbot functionality.
Approach:
We followed the product-centered approach when prototyping Rhea, conducting UX research, and learning the daily workflows and problems of our TA.
The solution that emerged is a hybrid interfaces that combines a prompt-driven AI functionality with dynamic widgets, graphics, offering a seamless research-to-report workflow with a handy split screen mode.
Prompt-based functionality🎊
At the heart of Rhea's capabilities is an advanced prompt-based functionality. Often, AI-generated answers to users' questions may be inaccurate, or users struggle to articulate their queries effectively.
To address this issue, Rhea offers clarifying questions and provides hints to expedite users in getting more accurate answers. Following the response, Rhea provides related questions on the topic or specific commands to aid users in completing their tasks.
Designing the multi-purpose input field 🪡
At Lazarev, we were dedicated to making Rhea a multifunctional solution, enhancing its value for users. We transformed a standard prompt field into an advanced command line.
People can now effortlessly search for files, set up notifications, handle automated emailing, schedule calls, issue alerts, and perform various other actions integral to the day-to-day workflows.
Designing a hybrid interface 🤖
The users of Rhea often work with data in the form of graphics, tables, or other widgets, which can then need to be saved, transformed in another format or utilized for report generation. So, considering the specifics of users' jobs to be done, we intentionally developed a hybrid GUI/prompt-based interface.
Depending on the type of information displayed in a widget, there are corresponding CTAs provided to assist users in further processing the results 😎
