Startup Mantra: Pivoting from service-based company to product firm

Ashutosh Saitwal realised that the industry badly needs a solution that could end their pain of manual document review and compliance workflows. And so, he built KlearStack

After spending 25 years with different software product/ SaaS (Software as a Service) companies, Ashutosh Saitwal, 53, decided to go on his own. He had experience with SaaS, enterprise software, finance and saw that document processing was a big headache for organisations.

Ashutosh knew that building an IP (intellectual property) meant a lot of money, time and effort. He hired people to develop his product - KlearStack. “Initially I hired six tech people. I spent a few crores until we had a minimum viable product (MVP).” (REPRESENTATIVE PHOTO)

“Being bootstrapped, we needed to generate revenue while we worked on various product ideas. I started with robotic process automation (RPA) services. In parallel, we were researching a few product ideas,” said Ashutosh, who completed his management accounting studies from Brihan Maharashtra College of Commerce (BMCC) in Pune

By late 2019, he saw that there was a real problem that craved an elegant solution. Almost 70% of the RPA project requests were about extracting and interpreting data from unstructured documents like invoices, PoS (Point of Sale), contracts and trade documents.

“While we automated mechanical steps for computer interaction, getting data from invoices, PoS was a different ball game. The RPA could not do that very well. Every invoice comes in a different format. RPA products attempted to solve this problem with template-based approach, which is very costly, since it means for every new variation of a document type (say invoice), you would have to create a new template, before the RPA tool can start extracting data from it. Alternatively, several companies employed large numbers of data entry and quality check (QC) operators to extract data from such documents e.g. supplier name, GSTIN (Goods and Services Tax Identification Number), customer details, invoice date, invoice number, tax amounts, invoice value, each line item details from the invoices, verify the data against supplier contracts and enter it into ERP (enterprise resource planning) systems before payments can be made. Both these alternatives (template-based RPA automation or manual data entry and QC) are time consuming, expensive and most importantly error prone,” he said.

Ashutosh quickly realised that the opportunity was big if someone could provide template-free automation. He knew that this problem could be solved using artificial intelligence (AI) and machine learning (ML). “We knew that we needed to come up with a solution that could read any document in any format and capture and audit the structured data correctly. AI can figure out on its own if the invoice number is on the left, the right or in the middle. ML models can be trained to read documents that companies receive in several different formats. I could have used someone’s ML API (application programming interface), but then I would not have my own IP (intellectual property). Also, I always wanted to create a product in India, for India and for rest of the world. So, I got down to build a team that could develop this AI/ML-based document reading and auditing solution,” he said.

In late 2019, Ashutosh pivoted from a service-based company to a product-based company.

Building a product

Ashutosh knew that building an IP (intellectual property) meant a lot of money, time and effort. He hired people to develop his product - KlearStack. “Initially I hired six tech people. I spent a few crores until we had a minimum viable product (MVP).”

In about one-and-a-half years’ time Ashutosh had an MVP ready. “We launched the first version of our MVP in December 2021. We tried it with our beta clients, and they were so happy with what it could do that they became our paying clients. The reason they liked our product was because of the high accuracy KlearStack delivered while providing a clean template-less experience. In a blind test, KlearStack delivered over 85%-90% accuracy. That was huge for our clients. They also knew that with KlearStack’s self-learning tech, the solution accuracy would further improve with usage.”

In the early days, Ashutosh managed to convert all of his beta customers to paying ones. But the road lay ahead. What was the plan to sell more?

“While our engineers keep working on our product making improvements constantly, we saw the data burden that banks had to tackle, during their loan journeys. Secure and unsecured consumer lending is big in India and a bank/NBFC needs to collect a ton of documents from borrowers for every loan. So, let’s say a person goes to a dealer to buy a car on loan. He approaches the bank representative for a loan. At several points in this loan journey, a borrower needs to provide many documents to the bank, such as his bank statement, Aadhaar card, pay slip, address proof etc. for onboarding.

“Once the onboarding is done, he will also need to submit the dealer invoice to bank, his car insurance papers, proof of his own downpayment, government tax receipt, RC (registration certificate) card and so on. The sales representative of the bank needs to take pictures of these documents and upload those to the bank’s system. In the backend, the banks/NBFCs have a team of data entry and QC operators or the sales team itself does the data entry. This manual data entry and verification is a big headache for the bank. It’s not as if all documents come in one simple format.

“So, with KlearStack, all that a bank sales representative has to do is upload pictures of the document and KlearStack does the rest in real time. All document pictures go to KlearStack via our API (application programming interface) integration and in real time KlearStack takes care of all the rest such as data capture, verification, document class identification, document image quality checking, fraud or suspicious document detection like the same vehicle that has been bought by the consumer is also the same one that has been insured and all the details mentioned in all documents are authentic.

“One of our clients is a well-known bank. Earlier this bank had 200+ people working to collate this data from consumer loan documents and KlearStack has reduced manual effort by 80% and improved productivity by 500%. Now, not just vehicle loans but the bank uses KlearStack for consumer durables and loans against property as well,” he said.

After working on a solution for loan document compliance, KlearStack set its sights on a solution for supply chain document compliance for import and export.

Ashutosh said, “All data across all documents for a specific import or export transaction has to be consistent and compliant. For example, you can’t have ABC as a seller in one document and ABC P Ltd in another. That will be a big operational and legal challenge. A mistake might go unnoticed in manual processing. Data integrity is vital for such kind of documents and that is where KlearStack comes in.”

Over the years KlearStack has several customers across India, Middle East and the US, processing millions of documents per month.

Competition

There is aplenty. Ashutosh said, “Most of our competitors are big names – Microsoft, Google. But the thing is that they have a platform approach. They provide their customers a platform and the customer has to then employ developers to work on that platform to create the solution they need. Plus, after development, you have to pay these solution providers for every page processed. So, the total cost of ownership is multifold.”

Unlike these horizontal platform-based competitors, Ashutosh said, “We focused on complete solutions for two use cases - loans and supply chain. Our clients do not need to hire developers, they simply use KlearStack for their document compliance automation needs.”

The future

“Our plan is to double down on these two solutions — loan and supply chain documents compliance — and increase our depth and breadth. While we do have these solutions that are loved by our clients, building a product is a continuous innovation journey. No product is ever totally complete. You need to keep getting better.

“We need to have more feet on the street. Currently, we have 15 people in development and a few in sales and marketing. We see a need to hire sales leadership and keep aggressively innovating on the product front. The big goal is to be a top player in loan document compliance and supply chain document compliance,” said Ashutosh, adding that “impossible is nothing!”