From Chatbot to Workspace:
Scaling an AI Visa Platform
From Chatbot to Workspace:
Scaling an AI Visa Platform
From Chatbot to Workspace:
Scaling an AI Visa Platform
The MVP proved the concept in chat. To scale, I split the architecture into two dedicated environments: a structured form with generative AI for applicants, and a high-density operations dashboard for visa agencies.
The MVP proved the concept in chat. To scale, I split the architecture into two dedicated environments: a structured form with generative AI for applicants, and a high-density operations dashboard for visa agencies.
The MVP proved the concept in chat. To scale, I split the architecture into two dedicated environments: a structured form with generative AI for applicants, and a high-density operations dashboard for visa agencies.
ROLE
Founding Product Designer
TIMELINE
2024 — 2025
PLATFORMS
Web, Mobile
RESULT
–38% drop-off / 4.4x B2B throughput
ROLE
Founding Product Designer
TIMELINE
2024 — 2025
PLATFORMS
Web, Mobile
RESULT
–38% drop-off / 4.4x B2B throughput
ROLE
Founding Product Designer
TIMELINE
2024 — 2025
PLATFORMS
Web, Mobile
RESULT
–38% drop-off / 4.4x B2B throughput



01 CONTEXT
01 CONTEXT
01 CONTEXT
The Chat Interface Solved the Early Launch, but Reached Its Limits
The Chat Interface Solved the Early Launch, but Reached Its Limits
The Chat Interface Solved the Early Launch, but Reached Its Limits
A conversational assistant was the right choice for the MVP: it launched fast, validated demand, and extracted document data reliably.
A conversational assistant was the right choice for the MVP: it launched fast, validated demand, and extracted document data reliably.
A conversational assistant was the right choice for the MVP: it launched fast, validated demand, and extracted document data reliably.
However, as workflows expanded and commercial clients arrived, the interface reached its limit. A linear chat simply could not handle that volume of structured data.
However, as workflows expanded and commercial clients arrived, the interface reached its limit. A linear chat simply could not handle that volume of structured data.
However, as workflows expanded and commercial clients arrived, the interface reached its limit. A linear chat simply could not handle that volume of structured data.






02 TWO AUDIENCES, TWO SCALING PATHS
02 TWO AUDIENCES, TWO SCALING PATHS
02 TWO AUDIENCES, TWO SCALING PATHS
The same application looked different
from an agency’s desk
The same application looked different
from an agency’s desk
The same application looked different
from an agency’s desk
APPLICANT
Is my application ready?
Your application
IN PROGRESS
France · Tourism
Personal details
Complete
Travel dates
Complete
Accommodation · Address missing
Add where you’ll stay before submitting.
Review answers → Complete details → Submit
See every answer. Find what’s missing. Fix it before submission.
APPLICANT
Is my application ready?
Your application
IN PROGRESS
France · Tourism
Personal details
Complete
Travel dates
Complete
Accommodation · Address missing
Add where you’ll stay before submitting.
Review answers → Complete details → Submit
See every answer. Find what’s missing. Fix it before submission.
APPLICANT
Is my application ready?
Your application
IN PROGRESS
France · Tourism
Personal details
Complete
Travel dates
Complete
Accommodation · Address missing
Add where you’ll stay before submitting.
Review answers → Complete details → Submit
See every answer. Find what’s missing. Fix it before submission.
AGENCY OPERATOR
What needs my attention?
Application queue
Different travelers. Different next steps.
A. Sharma
Needs attention
Passport scan unclear → Request a new scan
M. Patel
Ready for review
Review →
S. Rao
Awaiting traveler
Follow up →
Spot blockers → Prioritize → Process
See every status. Find what’s blocked. Choose the next action.
AGENCY OPERATOR
What needs my attention?
Application queue
Different travelers. Different next steps.
A. Sharma
Needs attention
Passport scan unclear → Request a new scan
M. Patel
Ready for review
Review →
S. Rao
Awaiting traveler
Follow up →
Spot blockers → Prioritize → Process
See every status. Find what’s blocked. Choose the next action.
AGENCY OPERATOR
What needs my attention?
Application queue
Different travelers. Different next steps.
A. Sharma
Needs attention
Passport scan unclear → Request a new scan
M. Patel
Ready for review
Review →
S. Rao
Awaiting traveler
Follow up →
Spot blockers → Prioritize → Process
See every status. Find what’s blocked. Choose the next action.
03 PRODUCT SOLUTION
03 PRODUCT SOLUTION
03 PRODUCT SOLUTION
Splitting the Interface:
A Hybrid Screen for Travelers, a Dashboard for Agencies
Splitting the Interface:
A Hybrid Screen for Travelers, a Dashboard for Agencies
Splitting the Interface:
A Hybrid Screen for Travelers, a Dashboard for Agencies
Instead of adding more filters to a crowded chat, I retired conversational UI as the primary container and built two purpose-built tools.
Instead of adding more filters to a crowded chat, I retired conversational UI as the primary container and built two purpose-built tools.
Instead of adding more filters to a crowded chat, I retired conversational UI as the primary container and built two purpose-built tools.
For B2C: full visual control and generative AI.
For B2C: full visual control and generative AI.
For B2C: full visual control and generative AI.
I moved the application into a structured form and kept the copilot alongside it. Passport details fill in automatically; applicants can review answers, edit generated documents and resolve file issues without leaving the application.
I moved the application into a structured form and kept the copilot alongside it. Passport details fill in automatically; applicants can review answers, edit generated documents and resolve file issues without leaving the application.
I moved the application into a structured form and kept the copilot alongside it. Passport details fill in automatically; applicants can review answers, edit generated documents and resolve file issues without leaving the application.



Passport details fill in automatically on upload. Applicants review and correct them directly in the form.
Passport details fill in automatically on upload. Applicants review and correct them directly in the form.
Passport details fill in automatically on upload. Applicants review and correct them directly in the form.

AI prepares an editable draft and marks the details that still need confirmation.

AI prepares an editable draft and marks the details that still need confirmation.

AI prepares an editable draft and marks the details that still need confirmation.

File checks flag issues before submission and show when a replacement is needed.

File checks flag issues before submission and show when a replacement is needed.

File checks flag issues before submission and show when a replacement is needed.
For B2B: high-density operations dashboard.
For B2B: high-density operations dashboard.
For B2B: high-density operations dashboard.
I replaced chat threads with compact data tables, clear status tags, and batch actions. Agency operators can now manage group bookings and switch between travelers instantly without manual copying.
I replaced chat threads with compact data tables, clear status tags, and batch actions. Agency operators can now manage group bookings and switch between travelers instantly without manual copying.
I replaced chat threads with compact data tables, clear status tags, and batch actions. Agency operators can now manage group bookings and switch between travelers instantly without manual copying.



04 MONETIZATION
04 MONETIZATION
04 MONETIZATION
A Paywall at the Submit Step and Actionable Approval Recommendations
A Paywall at the Submit Step and Actionable Approval Recommendations
A Paywall at the Submit Step and Actionable Approval Recommendations
Early paywalls created friction: users refused to pay before knowing their approval odds.
Early paywalls created friction: users refused to pay before knowing their approval odds.
Early paywalls created friction: users refused to pay before knowing their approval odds.
I moved the paywall to the final Submit button, after all 40 fields were verified. Instead of a passive score (e.g. 65%), the AI gave prescriptive advice: «Attach a sponsorship letter to boost approval from 65% to 90%+».
Applicants could check their details and resolve missing information before being asked to pay.
I moved the paywall to the final Submit button, after all 40 fields were verified. Instead of a passive score (e.g. 65%), the AI gave prescriptive advice: «Attach a sponsorship letter to boost approval from 65% to 90%+».
Applicants could check their details and resolve missing information before being asked to pay.
I moved the paywall to the final Submit button, after all 40 fields were verified. Instead of a passive score (e.g. 65%), the AI gave prescriptive advice: «Attach a sponsorship letter to boost approval from 65% to 90%+».
Applicants could check their details and resolve missing information before being asked to pay.
BEFORE PAYMENT
BEFORE PAYMENT
01
Complete
the application
Add details and
upload documents.
Passport details
Filled from your passport
01
Complete
the application
Add details and
upload documents.
Passport details
Filled from your passport
01
Complete
the application
Add details and
upload documents.
Passport details
Filled from your passport
02 →
Review
flagged issues
See what still
needs attention.
Accommodation
Address missing
02 →
Review
flagged issues
See what still
needs attention.
Accommodation
Address missing
02 →
Review
flagged issues
See what still
needs attention.
Accommodation
Address missing
03 →
Fix missing
details
Make changes
inside the application.
Accommodation
Address added
03 →
Fix missing
details
Make changes
inside the application.
Accommodation
Address added
03 →
Fix missing
details
Make changes
inside the application.
Accommodation
Address added
PAYMENT AT SUBMIT
04 →
Pay to
submit
Payment appears
after the review.
Application reviewed
Continue to payment →
PAYMENT AT SUBMIT
04 →
Pay to
submit
Payment appears
after the review.
Application reviewed
Continue to payment →
PAYMENT AT SUBMIT
04 →
Pay to
submit
Payment appears
after the review.
Application reviewed
Continue to payment →
INSIDE STEP 03
A missing detail.
A clear next step.
Your accommodation details are missing.
Add where you’ll stay to complete this section.
Add accommodation →
INSIDE STEP 03
A missing detail.
A clear next step.
Your accommodation details are missing.
Add where you’ll stay to complete this section.
Add accommodation →
INSIDE STEP 03
A missing detail.
A clear next step.
Your accommodation details are missing.
Add where you’ll stay to complete this section.
Add accommodation →
05 OUTCOMES
05 OUTCOMES
05 OUTCOMES
From a Conversational Bot
to a Scalable Platform
From a Conversational Bot
to a Scalable Platform
From a Conversational Bot
to a Scalable Platform
–38%
form drop-off reduction
(B2C)
–38%
form drop-off reduction
(B2C)
–38%
form drop-off reduction
(B2C)
16 min
average application time
(down from 45 min)
16 min
average application time
(down from 45 min)
16 min
average application time
(down from 45 min)
4.4x
increase in daily operator throughput
(B2B)
4.4x
increase in daily operator throughput
(B2B)
4.4x
increase in daily operator throughput
(B2B)
+24%
payment conversion
at the final step
+24%
payment conversion
at the final step
+24%
payment conversion
at the final step
Replacing a monolithic chat with specialized interfaces preserved our core AI strengths while providing clarity for applicants, viable monetization for the business, and scalable B2B expansion across India and MENA.
Replacing a monolithic chat with specialized interfaces preserved our core AI strengths while providing clarity for applicants, viable monetization for the business, and scalable B2B expansion across India and MENA.
Replacing a monolithic chat with specialized interfaces preserved our core AI strengths while providing clarity for applicants, viable monetization for the business, and scalable B2B expansion across India and MENA.

