AI Assisted Workflows for Multiple Stakeholders
At Alma, case filing speed depends on how quickly multiple stakeholders - hiring managers, mobility teams, and attorneys can complete their parts of the process, yet key workflows were slowed by heavy manual effort. Through a workflow audit, I identified two major bottlenecks: hiring managers struggled to write PERM-compliant job summaries, leading to repeated back-and-forth, while attorneys manually reviewed I-94 travel histories in spreadsheets to calculate visa max-out dates. I led the design of AI-powered experiences to eliminate this friction, rapidly prototyping and validating solutions using tools like Lovable to integrate AI directly into existing workflows. As a result, job summary turnaround time was reduced from ~10 days to under 24 hours, and max-out date calculations dropped from ~1.5 hours to 14 minutes, directly accelerating case readiness and filing velocity across teams.
MY ROLE
Product Designer and Product Manager at Alma
METHODS
User Interviews, Affinity Mapping, AI-Assisted Wireframing & Rapid Prototyping, Usability Testing, High-Fidelity Prototyping
Problem Statement
After conducting 1:1 meetings with key stakeholders and performing preliminary research, I refined the problem statement and reframed it using the extended Jobs to Be Done framework:
As a result, a task that should take minutes often takes days—and creates unnecessary friction across the entire case workflow.
Goal
System Approach
We designed an AI-assisted workflow that balances speed, control, and compliance. I worked as both Product Designer and Product Manager, partnering closely with AI engineers and immigration attorneys to define:
What the AI should access
How it should behave
Where humans should stay in control
Final Design
Below is a step-by-step breakdown of the final solution we arrived at through a series of iterations:
Step 1: Guided Setup via Preliminary Questions
We ask whether the PERM is for a current or future role, so the system knows which data is available and avoids incorrect assumptions.
Step 2: AI-Generated Job Summary
With minimal inputs, AI selects reliable data, applies a compliance layer, and generates a PERM-ready job summary in one click.
Step 3: Flexible Iteration (Manual or AI)
Hiring managers can refine the summary manually or ask AI to adjust sections, adding role-specific details while staying compliant.
Step 4: AI Compliance Check on Edits
Any manual edits trigger an automated compliance review, flagging potential issues before submission to attorneys.
Result
We received positive feedback from hiring managers on how the AI integration helped them generate job summaries faster and with greater confidence. Compared to the pre-AI workflow, the new feature significantly reduced the time required to complete role information, as shown below:
PROJECT -2
Problem Statement
At Alma, attorneys handle high case volumes under strict SLAs, so our product team focuses on identifying moments where we can help them complete tasks faster and more effectively.
One particularly time-consuming task was calculating visa max-out dates, which involved repetitive manual work.
Current Workflow:
The work was slow, error-prone, and mentally taxing.
Goal
Airbnb’s Beyond 5-Star Thinking
This project was guided by a design philosophy from Brian Chesky (CEO of Airbnb): imagine what a 5-star experience looks like, then push it to 6 stars and beyond. The goal isn’t to build an 11-star solution, but to explore widely enough to find a meaningful step-change.
We applied this thinking to how attorneys review and calculate visa max-out dates.
Progression of Ideas:
Final Design
Below is a step-by-step breakdown of the final solution we arrived at through a series of iterations:
Step 1: Clear AI Attribution
Parsed and calculated values are marked with a clear AI badge. Clicking the badge shows the parsed values and how it was used in calculations, helping attorneys quickly verify correctness and build trust.
Step 2: Contextual Document Zoom
Hovering over a parsed value zooms directly to the exact section of the source document, allowing attorneys to verify information instantly without manual scrolling or searching.
Result
Attorneys responded positively to the automated parsing and source mapping, which enabled them to review information quickly without repeatedly scanning source documents and manually transferring values into spreadsheets. By eliminating this rigorous manual work, the feature significantly reduced review time, as shown below.
OTHER PROJECTS
Out of all the pixels in all the screens, you ended up here. Grateful for your visit!
2024 Ravi Jangir. All Rights Reserved.
Let’s connect