Introduction
At Udhyam, The Youth Business Project (YBP) is one of the flagship programs, that is dedicated to nurturing an entrepreneurial mindset in students by helping them turn their ideas into real business, while helping with the overall mindset development when they are going through the journey.To expand this initiative effectively, we utilize one of the most accessible platforms available to students today: the WhatsApp Chatbot.
WhatsApp Chatbot provides a simple, scalable way to engage with and support students throughout their journey. With just a keyword, students can easily begin the process, making it accessible to all. The chatbot’s flows are carefully created with the student persona in mind : easy, conversational, intuitive, and broken down into clear, manageable steps.
Bot Submissions: Part of Students’ Learning Journey
At the start of the program, teachers form teams of students who then work together throughout the program journey. The program runs in two phases – Phase 1 and Phase 2.To be eligible for seed money,the team must complete the required submissions in Phase 1 by uploading them on the bot.
Phase 1: What Team Needs to Submit
Let’s walk through the different submissions teams were expected to complete in Phase 1.
- Idea Submission : As part of this submission, teams need to submit their business idea. Once a team submits their idea through the bot, it goes through an AI validation to check and improve the quality of submission. In MP, teams had to submit at least one idea. If they had more than one, they could share a second idea later.In Punjab, teams needed to submit only one idea.
- Sample Submission/ Market Survey Submission : In case of MP,teams need to submit a sample image of their product or service. Once the image is submitted, it goes through an AI validation process to ensure that the sample image is clear and represents a business idea.This was an optional submission, but teams could submit it if they had a sample image of their product or service.
In the case of Punjab, instead of a sample submission, teams are required to submit an image of the market survey they’ve conducted. This image undergoes an AI validation check to ensure it is clear. - Pitch Video Submission : As part of this submission, teams are required to create a pitch video where they present and explain their business idea.The submitted pitch video goes through an AI validation check to ensure it genuinely explains a business idea.
In MP, once submissions went live, teams were free to begin with any submission they preferred. And if a team submitted a pitch video, they became eligible for seed money selection.
In Punjab, teams had to complete all three submissions to be eligible for seed money selection.The submissions were opened in a specific order, even though all of them were live.
Phase 2: What Team Needs to Submit
Let’s take a look at the various submissions required from teams in Phase 2.
- Improved Sample Submission / Sample Submission : In both states, teams need to submit a sample image of their product or service.Once the image is submitted, it goes through an AI validation process to ensure that the sample image is clear and represents a business idea. In MP, we’re calling it the ‘Improved Sample Submission‘ because in Phase 1 it was optional, and some teams had already shared their sample images earlier.
- Final Tejasvi Submission / Final BB Submission : In the final submission, teams need to share: how many customers they got, how much they earned in total, and what they learned from the program.
From Chaos to Clarity: A Tech Team’s Story of Building YBP (2024–2025)
Building Once, Using Everywhere: Reusability and the Shift to Modular Flows
We’ve been using Glific as the chatbot builder platform for a while to build our YBP submission chat flows. But in the early days, even though the flows were quite similar across states, we found ourselves recreating them separately for each one. It felt like we were doing the same work over and over again.
Eventually, we realized there had to be a better way. So we started designing the flows in a modular and configurable manner. The goal was to make them reusable across different YBP states with minimal changes. This not only saved us time but also made the flows easier to understand and maintain.
While redesigning the flows for reusability, we tried to build them in a way that could work for both MP and Punjab. And honestly, that worked out quite well in many cases. But as we got deeper into it, we realized that each state had its own small variation, so we had to tweak a few things here and there. Still, it felt good to see that a big part of the work could be reused across both states.
This experience showed us that learning from the past helps us build better, more flexible solutions that save time and effort.
Easy Flow Access for Everyone: Boosting Submissions
One major challenge was that many students didn’t have regular access to a mobile phone.This became a big hurdle. Students worked hard on their business ideas but couldn’t submit them.As a result, several teams struggled to complete their submissions,simply due to device unavailability.
To address this, we restructured our submission flows and optimized how submission data is stored and accessed. Now, if students don’t have access to a phone, they can simply reach out to their teacher or the state operations team, who can submit on the student team’s behalf.There’s no restriction on how many teams they can help submit for.They just need to type the keyword, enter a student ID from the team, and confirm the team. This then displays a menu showing the team’s progress.
What’s more, students themselves can step in to help others. If a team doesn’t have access to a device, another student can submit on their behalf, and many students are already doing this.
We’re happy to see this approach working in action, MP operations teams have been stepping in, using their own phones to submit for teams.
This experience showed us that flexible systems help students participate and complete their journey successfully.
Milestone-Driven Submission Flow : Enabling Smarter Nudging
The turning point came when the product team asked us — “Can we make sure users return to the exact point they left in the submission flow?”
Until then, we only knew whether a team had completed their submission or not. Although we had data on individual user actions within the submission flow, it wasn’t easy to understand or use effectively mainly because our focus was more on the team’s overall progress rather than each user’s actions.
That question made us pause. Until then, our flows weren’t designed to handle this. We only tracked whether a team had submitted or not nothing in between.
That question got us thinking differently: What if we broke the submission journey into smaller, trackable steps?
That’s how the idea of milestones came into the picture. By dividing the flow into clear milestones, we could easily track progress, pick up right where a team left off, and understand exactly where teams were dropping off.
To accommodate this, we maintain a separate table in BigQuery that tracks the milestones each team has reached. Using this data, we can see how many teams are at each milestone, send targeted nudges to help them move forward, and resume their submission flow from the last completed milestone.
This experience taught us that asking the right questions opens the door to new ideas and helps us build better, more effective solutions.
Seamless Sync: Chatbot ↔ PWA
Our Interaction with teachers and students happens through two platforms: a Chatbot and a PWA.
Earlier, users had to register separately on both platforms, which made the process time-consuming and repetitive.To make this smoother and less tedious, we worked with the product team to sync the registration process. Now, when a teacher or student registers on one platform, their profile is automatically created on the other as well.To achieve this seamless integration, we utilized the Glific API along with several other custom external APIs, which have been integrated into both the Chatbot and the PWA.
This experience taught us the value of closely observing users’ challenges and building better solutions.
Launching MP Idea Submission Phase 1: A Quick Story (2024-2025)
When we first launched the MP Idea submission flow, everything looked smooth,until the MP team noticed that for some teams, submissions weren’t showing as “complete” on the dashboard, even though they had submitted their Idea. After digging in, we found out that teams had submitted only 1 Idea, but the process was expecting 2 Ideas to mark it as done. That’s when we realised the mistake in the design.
So we quickly had a discussion, reopened the flow logic, and made a change: now just 1 idea was enough to complete the submission, and if teams had a 2ⁿᵈ Idea, they had to submit it right then and there.
But, of course, a few days later, another issue popped up. Some teams actually wanted to submit their 2ⁿᵈ Idea later but they couldn’t. The flow just wouldn’t allow it. What we heard was that some teachers were deleting teams and making new ones so they could submit a 2ⁿᵈ Idea for the teams!
Seeing this, we made another change to truly make the 2ⁿᵈ Idea optional and allow teams to submit it later if they wished.
This experience taught us the importance of staying flexible and listening closely to user behavior.Even with good planning, unexpected user behavior can surprise us. The key is to stay flexible and fix things fast.
Market Survey Twist: Punjab Phase 1 (2024-2025)
We were all set to kick off Punjab Phase 1, thinking the submissions would be just like MP’s. Plans were in place, and the Idea submission was already live, marking the beginning of the Phase 1 journey. Then, out of the blue, the state team told us that they actually needed a Market Survey submission instead of the Sample Submission as it was a request from the government department.We didn’t have a choice other than change the flow logic. With a tight deadline, all teams from product to tech immediately jumped into action. Everyone collaborated closely, working together to create, test, and launch the flow on time.
This experience taught us that even well-laid plans can shift unexpectedly & collaboration across teams accelerates success.
How AI Became Our Submission Checker – An Unexpected Journey (2024-2025)
So this time, with the help of the Product team and Consuma, we experimented with AI-based validation in our YBP submission flows.The goal was simple but ambitious: to validate student submissions using AI. But the journey? Not quite so simple.
From countless rounds of testing and tweaks to dealing with platform limitations on Glific, we experienced it all. There were moments of confusion, late-night bug hunts, and multiple iterations.
But all the effort paid off, we successfully integrated AI validation into our submission flows, and it went live. The best part? The submissions we received were clearer, more meaningful, and overall better in quality.
While it was a big milestone, going live also came with its own set of challenges. We faced issues like validation breaking under increased traffic and certain OpenAI restrictions.
Whenever validation issues came up, the MP and Punjab teams stood by us.They patiently handled student and teacher queries, often in high-pressure situations, and gave us the breathing room we needed to work on fixes. Their support made a big difference, and we truly appreciate their calm and collaborative approach throughout the process.
This experience showed us that trying new things is only possible when teams collaborate, stay patient through the ups and downs, and trust the process together.
MP Phase 2 & Punjab Phase 1 Media Submission Mishap: A Small Slip, A Big Learning (2024-2025)
During the middle of Punjab Phase 1, we encountered an unexpected issue resulting in the loss of over 4,000+ team pitch videos, around 9,000+ team market survey images, and more than 1,000+ improved sample images in MP Phase 2. It happened because of a small slip that went unnoticed from both our side and Glific’s.
What actually happened was that the connection between Glific and our Google Cloud Storage (GCS) broke. A notification email was sent when the disconnection happened, but no one from our tech team had access to that email. So we missed the alert, and the issue went unnoticed for a while.
Glific had resolved a similar issue last year, and during MP Phase 1, we had actively checked that GCS links were being generated correctly for the Gupshup media files. We were quite confident that this problem wouldn’t happen. That made this slip even more unexpected.
We discussed the issue with the Glific team and identified some action items from their end to handle such cases better in the future.On our side, we’ve set up a weekly alert to flag any Gupshup media files missing a GCS URL, so we can catch issues early and avoid similar problems going forward.
This experience taught us that even a small oversight can lead to major issues, making careful attention and proactive monitoring absolutely essential.
MP Phase 2 & Punjab Phase 1 Media Recovery: Our Effort to Restore Lost Data (2024-2025)
Since the volume of lost data was significant, we adopted a different strategy to nudge teachers and students.Instead of mass nudging, we studied the existing submission data closely. We figured out exactly which teams were affected and who needed to be reminded. This helped us reach out more effectively, nudging the right teachers and students.
To support recovery, we updated the submission menu from the backend to show missing submissions as “pending”, which effectively signaled teams to complete their submissions.Most importantly, we designed the recovery flow in a way that minimized effort on the students’ end. Our goal was to make it as easy as possible for them to reshare what was lost without starting from scratch.
We’re grateful to both the MP and Punjab teams for their immense support in recovering the lost submissions. They patiently addressed all queries and guided students and teachers throughout the process.
Thanks to these collective efforts, we were able to recover nearly 90% of the lost data.
The difficult setback showed us how crucial open communication and teamwork are in overcoming challenges.
Growing Together, One Step at a Time
As I reflect on the journey, one thing stands out above all – collaboration. Whether it was building modular flows, experimenting with AI validation, or recovering from unexpected setbacks, every challenge became an opportunity because of the people who came together to solve it.
I’m deeply grateful to the Product and Tech teams, whose constant iteration and problem-solving mindset helped us adapt quickly and build thoughtfully. A huge shout out to the MP and Punjab State teams for your unwavering support, deep ground insights, and patience during high-pressure moments made all the difference.Whether answering questions or guiding students and teachers, your dedication has driven our success.
Here’s to more ideas, more learning, and even stronger collaboration in the year to come. 🚀
