A Generative AI Assistant for the Office of the CFO
Year
2023
Duration
4 months
Role
Product Designer
Platform
Web
Team
VP - Product Design, Engineering Director, 2 Front End Engineers
Tool stack
Figma, Jira, Miro
My Role
Pioneered the exploration of Generative AI concepts and prototypes for HighRadius' FinTech SaaS solutions, focusing on user-centricity, efficiency, and a delightful user experience in the finance domain.
Challenge
To create an AI assistant that can integrate into any Highradius products and suggests actions the user needs to focus on and provide contextual help.
Status Quo
Highradius being a giant in fintech space wanted to leverage generative AI and internal large datasets to offer personalised software, which is more assistive in nature.
Since this was a 0 to 1 experience, I had to start by first understanding the user requirements that we are trying to fulfil and also look at how other digital companies are introducing generative AI powered featured to their customers.
Solution
I created various concepts and 3 of them have been shipped by Highradius:
concept 1 of 3
FreedaGPT Email Compose
No more Inbox Overwhelm: Collectors are flooded with numerous emails daily, including invoice queries, payment notifications, and overdue reminders.
Email Triage: The system categorizes and prioritizes these emails based on urgency and content, understanding the central context of each message.
Automated Responses: Draft tailored responses for each email, attaching relevant documents, missing invoices or account statements, and reminders for overdue accounts.
Collector’s Oversight: Before sending the email, collectors review and refine auto-composed emails for a personalized touch, streamlining the response process.
Generate replies for emails that need your attention
concept 2 of 3
FreedaGPT Insights
On-click Data Gathering: As month-end approaches, users are stuck in the vicious cycle of gathering past data, making consolidated reports, and generating real-time insights.
Leveraging FreedaGPT: Bypass the manual process of creating a new dashboard or applying multiple filters on the data set. An user could prompt it to show the top 5 anomalies with high risk.
Generate Reports in Seconds: FreedaGPT promptly generates summary reports, insights, and graphs within seconds for any input. Users can delve deeper into the data as needed.
Chat with your personal intelligent dashboard to get insights
concept 3 of 3
FreedaGPT Support
Auto Support Portal: HighRadius users raise a support case when they need professional assistance resolving an issue. In most cases, resolution could have been quicker via self-service channels.
Self-Service: Resolve common issues independently by directly prompting FreedaGPT. For instance, an admin user can ask FreedaGPT for real-time assistance with tasks like adding or creating a new user.
Raise tickets: Eliminate the need to manually raise support cases by prompting FreedaGPT directly from your Highradius application. FreedaGPT understands the context to automatically raise a support case.
The following video is the demo of FreedaGPT Support by the the CEO of HighRadius.
Sharing more concepts that I had explored
Task Automation
Generate formulas, pull data from sources and fill forms
Fill forms by prompting
Automate any task and save as as favourite
Generate formulas by typing in english
Edit formulas by converting them to english
FreedaGPT Automation
Live AI a collaborator. A second cursor on the screen.
FreedaGPT Copilot
Process
step 1 of 6
Have a conversation. Keep it simple.
To create concepts that offer utility and not overwhelm the user, and to offer an assistive experience, I had to treat this as a conversational UI. Clearly indicating different between user input and AI system output.
Also to account for and design for the time system takes to generate responses.
1 is to 1 communication
step 2 of 6
Create concepts based on the Highradius product suite
I identified use cases and touch-points across all products where generative AI could be useful. There were some suggestions from the generative AI team that they too wanted to try.
Various concepts were explored, created showcasing how generative AI to could make the user tasks efficient.
Identify use-cases and create concepts
step 3 of 6
Give demos to all stakeholders and get buy-ins
Post the demo of the various concept prototypes, we finalised a few concept directions based on the best way Highradius could position itself in the market.
step 4 of 6
Integrate stakeholder feedback. Iterate.
Working on the feedback in the demo, the selected concept prototypes were then fine tuned to pixel-perfection and made ready for hand-off to the engineering team.
Feedback and market position to help guide concept direction
step 5 of 6
Setup design walkthroughs with engineering team
A walkthrough session where I presented the click-through prototype to the engineering team and clarified any doubts by them. I also covered various states this conversation UI reflect to show it’s assistive nature.
step 6 of 6
Test and call out bugs or request for enhancements
Rigorous UX Testing was conducted of all the concepts. Attention to details being extremely important, as the system would now be more responsive. Several bugs and defects were raised for the engineering team to fix before the public release.
Impact
Reduction in Turn around Time for emails.
Easy access to reports and ability to create custom 1-click dashboards
Reduction in number of tickets
Learnings
Understanding various ways generative Ai can help make the user experience better by providing contextual help.
The system can train based on user inputs and be designed to be more helpful.
This body is identified as Neelank Sachan, the mind is ignited with curiosity to always learn and the soul is searching for meaning through the growth of ecosystem it exists in.
