AWS Builder Center
We Didn’t Build Another Dashboard. We Built an AI Team for Businesses That Live on WhatsApp.

We Didn’t Build Another Dashboard. We Built an AI Team for Businesses That Live on WhatsApp.

BizMate is a multi-agent AI business assistant for SMEs, retailers, wholesalers and store owners. Customers talk to a Sales Agent, while owners use a Manager Agent coordinating Finance, Social and Operation Agents to turn WhatsApp conversations into sales, operations, payments and growth.

AI Engineer & software Developer

We Didn’t Build Another Dashboard. We Built an AI Team for Businesses That Live on WhatsApp.

Every hackathon starts with excitement. Ideas are flying around, someone is opening Figma, someone is setting up the backend, and someone is already thinking about deployment before the first feature even exists. But then the clock starts moving faster. A few hours disappear, the pressure increases, and one question becomes unavoidable: are we building something that looks impressive for the judges, or are we building something that someone would actually use?
For us, that question became the foundation of BizMate.
We started by looking at how small and mid-sized businesses actually operate. Talk to a local store owner and you quickly realise that their business does not live inside a perfect ERP dashboard. It lives in WhatsApp conversations, phone calls, spreadsheets, notebooks, vendor messages, payment reminders and memory. A customer asks for a product on WhatsApp. Another customer wants a quote. A vendor sends a new price. Someone forgets to make a payment. Another customer needs a follow-up. At the same time, the owner is trying to manage stock, sales, finance, employees and social media. There is software for almost every individual task, but the real problem is that the owner has to remember and coordinate everything.
That led us to a much simpler question: What if we could reduce the amount of business an owner has to remember?
That is where BizMate began.  https://app.stockaware.vaaani.co.in/ 
The idea was not to build another complicated business dashboard. We wanted to build an AI team that could work around the conversations businesses already have. Customers could talk directly to a Sales Agent. Business owners could talk to a Manager Agent. Behind the Manager Agent, specialised agents could handle different parts of the business, including a Finance Agent, Social Agent, and Operation Agent.
The goal was to keep the experience simple on the surface while making the intelligence underneath much more powerful.
A customer could send a message like, “Bhai 10 black kurtis chahiye, size mix chalega?” The Sales Agent understands the product and quantity, checks available inventory, and responds with what can actually be supplied. If seven pieces are available, the agent can explain that, prepare a quote, and continue the workflow toward payment. The owner does not need to manually check the stock, calculate the quote and remember to follow up with the customer. The conversation itself becomes part of the business workflow.
We deliberately used a local women's clothing store as one of our primary examples because it reflects the complexity of real retail businesses. Customers discover products through Instagram and WhatsApp, products come in multiple sizes and variants, new stock arrives regularly, vendors send updates, payments have to be tracked, and customers often need follow-ups. It made the problem very real for us.
At the same time, the owner has a completely different experience. Instead of needing to open multiple dashboards, the owner can interact with the Manager Agent. They can ask something as simple as, “Kal ka business kaisa tha?” or “Aaj kya important hai?” The Manager Agent brings together the information that matters. Yesterday's sales, pending payments, expiring quotes, stock issues, vendor updates and customer follow-ups can all become part of one business briefing.
And behind that briefing, the specialist agents do the work.
The Finance Agent helps with bills, calculations, balances and financial summaries. The Operation Agent focuses on stock, products, categories and vendor updates. The Social Agent helps with captions, hashtags, content planning, offer creatives and publishing workflows. The Sales Agent stays focused on conversations with customers and moving those conversations through the sales process.
This became one of the most important ideas behind BizMate: the owner should not have to talk to five different systems. They talk to Manager Agent. Customers talk to Sales Agent. The specialists work behind the scenes.
Then came the AWS part of the project.
We didn't want BizMate to remain something that only worked on our laptops. We wanted to build it as a cloud-deployed system with infrastructure that could actually be reproduced. We used Terraform as Infrastructure as Code and built our deployment architecture around AWS services including Amazon ECS, AWS Fargate, Amazon ECR, Amazon VPC, Application Load Balancer, Amazon RDS, Amazon S3, AWS Certificate Manager and Amazon CloudWatch.
Amazon ECS with Fargate allowed us to run our containerised application without managing the underlying servers ourselves. Amazon ECR gave us a dedicated place to store and manage our container images. Amazon VPC provided the networking foundation for the application and database layers. Application Load Balancer handled incoming traffic and routed it to the deployed services. Amazon RDS gave us a managed relational database layer, while Amazon S3 provided object storage. AWS Certificate Manager helped with HTTPS and TLS certificates, and CloudWatch gave us logs and monitoring visibility. Terraform tied the infrastructure together so we could define and reproduce the environment as code instead of manually configuring every resource through the AWS console.
Working with these services during a hackathon also taught us that the difficult part is often not one individual AWS service. The real challenge is making everything work together. ECS, networking, IAM permissions, health checks, security groups, load balancing, database connectivity and monitoring all have to line up correctly. That learning became just as valuable as the product itself because we stopped thinking only about whether the code worked and started thinking about whether the system could actually run.
The hardest part of the project, however, was not AWS. It was making the product feel simple.
A request like “Black kurti hai?” looks trivial. But underneath it, a useful business assistant needs to understand the product, quantity and context, check inventory, work with pricing, prepare the right response, know when approval is required, continue the sales workflow, handle payment and eventually connect the transaction to invoicing. The AI shouldn't force the owner to understand all of those steps. That complexity should disappear behind the experience.
We also wanted BizMate to understand how real people actually communicate. Customers and store owners don't always write perfect sentences. Sometimes they use Hindi, sometimes English, sometimes Hinglish, and sometimes they simply send a voice note. So voice and multilingual communication became an important part of the experience. The idea is straightforward: you don't have to type, you can just talk. An owner should be able to send a voice note asking about yesterday's sales or today's important updates and receive a useful response in a natural conversational format.
Then something happened that changed the project for us.
We stopped talking only to judges and developers and started talking to actual business owners.
That changed the conversation immediately. People weren't interested in hearing about AI architecture for ten minutes. They were asking practical questions. Can it answer customers? Can it check stock? Can it handle voice messages? Can it follow up with customers? Can it help with Instagram? When can I actually use it?
That last question mattered the most.
We have now onboarded early businesses around BizMate, and some of those businesses have already pre-booked the product for ₹2,000 each for two months of early access. For us, that was a completely different form of validation. A hackathon judge can tell you that an idea is interesting. A real business owner telling you that they want to use it is a different signal. And when someone is willing to put money behind that interest, even at an early stage, the project starts becoming more than a hackathon prototype.
It starts becoming a business.
That experience changed how we think about the product. When we started, our goal was to build something impressive within the constraints of a hackathon. Now the question is bigger: How do we make running a small business less exhausting?
Because the ambition of business owners is rarely the problem. They want more customers. They want better sales. They want to grow. They want to expand. They want to spend more time on the parts of the business that actually matter. The problem is everything that constantly interrupts that growth: the payment that wasn't followed up, the stock that wasn't checked, the vendor message that was missed, the customer who never got a response, the quote that was supposed to go out two hours ago, or the social media post that never got created.
That's the work we want BizMate to absorb.
The biggest lesson we learned while building it wasn't really about AI, AWS or even hackathons. It was about usefulness.
People don't want more technology. They want less work.
An impressive AI demo can get attention. A useful AI system can get adopted. That difference completely changed our approach. Instead of constantly asking, “What else can our AI do?” we started asking, “What can the business owner stop doing manually?”
That leads to a very different kind of product.
The hackathon gave us the opportunity to build the first version. AWS gave us the infrastructure to deploy it. Our team built the AI workflows, system architecture, cloud infrastructure, backend and frontend experiences. But the conversations with business owners gave us the direction to keep going.
Today, BizMate is moving from being just a hackathon project toward becoming something that real SMEs, retailers, wholesalers, distributors and store owners can use every day.
Not another dashboard they have to learn.
Not another complicated system.
Not another chatbot that simply gives generic answers.
An AI team that works quietly behind the conversations they already have.
Customers talk to Sales Agent. Owners talk to Manager Agent. Finance, Social and Operation Agents work behind the scenes.
We started with a hackathon.
But now we're building for the people who don't have time for one.

Run your business. Not the busywork.

BizMate.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article