Gone are the days of poring over online forums and blogs to understand product features, question the authenticity of five-star marketplace reviews, and compare dozens of products before zeroing in on one. Agents powered by AI are trained to do it all for you.
As ecommerce entered the age of agentic commerce, Ranjith Boyanapalli took the plunge, with his Flash AI last September. The former Flipkart executive launched it as an AI-powered shopping assistant that helps in personalised product discovery and research.
Bengaluru-based Flash had earlier raised $12.5 Mn across two rounds. The company’s cap table features the likes of Blume Ventures and PeerCapital as well as high-profile angel investors such as Binny Bansal, Lalit Keshre, and Kunal Shah. Its earlier avatar – a centralised inbox for power shoppers – was entirely different from the latest flagship AI venture.
The market opportunity for AI-powered ecommerce is huge. Agentic commerce has the potential to contribute $3-5 Tn in revenue to the global B2C retail market, as 64% Indian consumers reportedly use AI tools in their purchase journey, beating the stats of many other countries.
Boyanapalli designed his Flash AI not for closing a transaction, but for the pre-purchase stage, when the buyer decides which product to pick up. Inc42 sought to look at the fine print behind the making of Flash AI to find out what the company is trying to solve and how much of a departure it is from what it was doing earlier.
What Prompted A Different Approach
After it was rolled out in 2022, power shoppers used the Flash email service as a one-stop shop to track orders, give them insights on their spending, and earn rewards for shopping. In other words, it tailored the user’s post-purchase stage to give them a more personalised experience.
But, as it turned out, it wasn’t exactly the right solution to the problem. “People loved the product, but a certain portion of users was using it just to keep their Gmail free. They were using it on BharatMatrimony, gaming websites, etc, where it was not meant to be used,” Boyanapalli told Inc42.
Flash wanted to go deeper into the user’s purchasing journey and bring hyper-personalisation to the pre-purchase stage as well, he said. “From day one, we meant to get into the discovery phase as a way to personalise commerce for people. I need to know what you bought to give you better options in future.”
The rise of AI offered the right opportunity to Boyanapalli to rethink how Flash could make the shopping experience more tailored while taking a step up the funnel. “We have seen the era of catalogue-led shopping, influencer-led shopping and now the next era is research-led shopping,” he said, adding that 50 Mn ChatGPT queries a day are related to shopping.
Flash AI has a simple working strategy. A shopper who wants to research a particular product merely has to append ‘flash.co/’ before the URL of the product page. Then, the AI assistant works in the background to prepare its analysis. Within a minute, it presents a summary of all the features, pros, and cons of the product.
The aim isn’t just to give a fact-based evaluation, but a personalised one. Flash AI isn’t just trying to tell users whether a product is good or not, but whether it is right for their requirements and use case. That may be why it has seen the most traction in categories like BPC and electronics, which make up a third of the products users research on its platform, Boyanapalli explained.
What Influenced The AI Pivot
Switching paths wasn’t a small feat for Flash. Tough calls had to be taken like ending support for the email service and cutting back on rewards in order to prioritise the new direction Flash was aiming at. The company had to build new infrastructure for the venture. Flash’s inbox service did use AI to process emails and extract information, but only to a limited extent as per a template, whereas its new service called for more advanced capabilities.
“Now, when you research a product, there are 6-7 LLM calls, some pre-processing and some post-processing, some real-time and some from our cache, going through Youtube videos and Reddit forums and expert blogs, transcribing it, classifying the sources basis the quality, and finally creating a summary. All of this in less than 40 seconds,” Boyanapalli said.
In the era of its inbox product, Flash had to handle 3-4 Mn orders a month. Flash handles 750,000 users today who research nearly 2 Mn products every month. While the number looks smaller in face value, the type of queries it deals with are more technically complex. Flash had three years to build the capacity it needed to reach the scale it had achieved, but the AI assistant picked up the expected traction in less than six months, he said.
Flash was not the first one to get the whiff of the potential in agentic commerce. Some major ecommerce platforms like Amazon Rufus, Flipkart SLAP and Zomato AI offer similar AI-powered product research and discovery features integrated into their marketplaces.
Boyanapalli banks on the trust Flash has earned. “What we’re offering is consumer-first, unbiased reviews. Our job is to be a research platform that helps shoppers make the right decision, as against optimising for conversion and sales. This problem exists because people feel that marketplace reviews are fake or they may find the same product for a better price later,” he said.
Flash AI was also influenced by the emerging trend of shoppers turning to AI chatbots as an option for shopping. Giants like OpenAI and Google have embedded agent communication protocols in their interfaces for businesses and to help carry out transactions, which means users could soon be making purchases merely by entering prompts without ever leaving the interface of ChatGPT or any other tool of their choice.
But Boyanapalli doesn’t see this as a threat. He claims that Flash will aim to beat them in depth rather than in breadth. Building this category-level personalisation – starting with beauty and personal care and electronics – will be a moat, he said.
“We are working with dermatologists to build a skin analyser where you take a few selfies of your face and we create your profile and show you the products that you should be buying given that. I am not going to dump 2,000 moisturisers on you, but research 2,000 products and show you the top 20 moisturisers for oily skin.”
When Trust and Monetisation Conflict
Flash’s AI pivot calls for a change in its business model. It had earlier monetised its email service through a freemium model, where users were charged subscription fees for more advanced insights and better rewards. But the AI avatar seems more suited to a B2B monetisation model.
For the long run, Boyanapalli envisions Flash monetising the user data by providing brands with granular competitive intelligence. “Many brands don’t have product-level insights of their own products, let alone their competitors. We can help them by telling what other competing products people are searching for, where they are available, how much they are priced, and so on,” he said.
This would be implemented further down the line, he clarified. Currently, Flash is earning through brand partnerships and affiliate links. “I can offer direct demand. A user who is doing research at this level of depth is the perfect customer with high intent. If you help the brand close this order, you can earn a much higher share of the order value,” Boyanapalli said. He, however, declined to share the specifics on the company’s current revenue and business performance.
The Flash team is also working on ad monetisation, though it won’t be implemented until the product gets more traction and attains a scale of 3-4 Mn product research page-views per month, he added.
Monetisation comes with a trade-off. Users may lose trust in Flash AI if they feel its product evaluations have been influenced by revenue considerations, rather than presenting an objective analysis. OpenAI, too, came under flak for trying to introduce ads into ChatGPT.
As AI turns into a disruptive technology across sectors and more and more businesses are jumping on the bandwagon, Flash will face a steeper competition with ecommerce staying the course to be a $400 Bn opportunity in India by 2030.
Curated by James Chen
