[L-R] Nūl co-founders Malini Kannan (CEO) and Raghav MS (CTO)
The fashion industry has long struggled with a costly and unsustainable flaw: overproduction. Billions of garments are made each year that never find a home—leading to lost revenue, wasted resources, and overflowing landfills. But what if there was a smarter way to align what brands produce with what customers actually buy?
Enter Nūl, a startup on a mission to fix fashion’s inventory problem using agentic AI. Founded by Malini Kannan and Raghav MS and launched recently by Wavemaker Impact, Nūl helps brands make sharper, faster decisions across their supply chains—reducing overstock, increasing sell-through, and cutting environmental waste along the way.
In this Q&A, Malini and Raghav walk us through how Nūl works, why agentic AI is a game-changer for retail, and what they’ve learned from working at the intersection of technology, sustainability, and fashion.
What inspired you to tackle the issue of overproduction in the fashion industry, and how does Nūl’s mission align with your personal values?
Malini Kannan (MK): At Nūl, we firmly believe that aligning profit with sustainability is possible, even as we drive for significant growth.
A major issue at the heart of this belief is overproduction, which is a common challenge in the fashion industry. When a fashion brand miscalculates, the consequences can be severe—not only affecting profitability but also hindering the ability to fund future collections.
It’s also wasteful, considering the resources, time, and labour involved in producing clothing. The situation becomes even worse when brands are unable to sell their items, even at discounted prices.
These unsold goods often end up in a cycle of being recycled into secondary markets or products. In the worst-case scenario, they are discarded in landfills.
Now imagine that for every 10 items a brand produces, it only sells six or seven at full price. How can we close this gap? This is the problem we aim to solve.
Can you walk us through a typical use case of Nūl’s technology for a medium-sized fashion brand, from data integration to actionable insights?
MK: Let’s take a typical example of a medium-sized fashion brand (with ~US$50 million in annual revenue) that has retail stores in two to three countries and sells online through its own store and a few regional marketplaces.
This firm has six seasons per year and some evergreen staples. Every season, it stocks about 200,000 to 250,000 units of clothes across 20-25 styles, allocated across its different stores and online channels. It is left with anywhere between 30-40 per cent of the inventory unsold each season. Today, it manages its inventory levels through a combination of ERP and Exel spreadsheets.
Also Read: Wavemaker Impact launches Nūl with US$500K investment to tackle fashion overproduction
Nūl ingests data from the ERP and Excel sheets used by various business teams without any change to its existing system. Through our simple web-based platform, we provide teams with real-time sales and inventory levels across all of the brand’s retail, online, and marketplace channels.
Furthermore, it provides smart recommendations on how to reallocate inventory. Assuming an item is selling really well at one location and expected to sell out within the next few days, Nūl will suggest reallocating from a store where the same stock isn’t moving as fast.
It identifies micro-trends around size, style, and colour at specific locations, allowing brands to anticipate and trigger reorders.
In addition, Nūl computes SKU-level performance data in real time and provides longer-term forecasts for planning & production.
Simultaneously, at the SKU level, for non-performing stock, it suggests timely shorter-term actions such as markdowns (minimising inventory holding periods or inventory sent for recycling)
Most importantly, it learns from the actions the brand took and customises its approach to the cycle.
What metrics do you use to measure Nūl’s success, both in terms of business growth and environmental impact?
MK:
Growth Metrics: We are still in our build phase, so right now, we are focused on getting customers to pilot with us to understand the variety of use cases they are using Nūl for in their operations and incorporating these into the core of our solution. More customers, more use cases, more variables – leading to a more robust solution.
Environmental impact: Currently, we are focused on helping brands reduce overproduction. Each apparel that isn’t likely to sell has embedded carbon emissions and water usage that varies based on the type of material, dye and process that went into making it.
A simple cotton t-shirt, say, would have about 6-8kgs of carbon emissions in its production and 2,700 litres of water consumed in the process. We provide brands with a baseline linked to their sell-through and the improvements delivered by using our solution to bring down the full-price unsold apparel’ number.
What are the biggest risks or challenges you foresee for Nūl in the coming years, and how are you preparing to address them?
MK: We are part of a bigger ecosystem of solutions required to truly support the fashion industry’s move toward more sustainable production. Keeping in mind our mission, some of the biggest risk factors would be the speed of development and adoption of a broader range of solutions that can get the industry there.
Specific to Nūl, it would be to build with momentum to capitalise on the global potential of our solution quickly.
Can you explain in simple terms how the term “agentic AI” applies to Nūl’s technology?
Raghav MS (RMS): Agentic AI is like a smart assistant that’s always learning and helping teams make faster, better decisions in real-time—based on what’s happening, not just what happened.
At Nūl, we’ve embedded agentic AI into the core of our inventory optimisation engine. Each agent has a specific role—whether forecasting SKU-level demand, evaluating stock imbalances, or optimising store-to-store transfers. These agents operate independently but coordinate using protocols we’ve built, such as a multi-agent coordination protocol and a model context protocol, to ensure they’re aligned and context-aware.
This allows our system to detect emerging patterns—like a sudden spike in demand for a product in one store—and proactively reallocate stock from slower-moving locations. It transforms inventory management from reactive to autonomous, dramatically reducing overproduction and lost sales.
What data sources does Nūl use to train its AI models, and how do you ensure the quality and relevance of that data?
RMS: Nūl’s AI models are trained on both internal and external data. Internally, we pull from POS and ERP systems like Shopify, SAP, NetSuite, and Oracle—tracking sales velocity, inventory, and sell-through at the SKU-store-week level.
Externally, we layer real-time signals such as weather data, regional holidays, footfall patterns, search trends, and campaign metadata. This is where our Model Context Protocol (MCP) comes in—it ensures every prediction is contextualised based on when, where, and why a trend is happening.
Also Read: A deep-dive into Wavemaker Impact’s decarbonisation strategies in SEA
To maintain accuracy, we use multi-stage validation and real-time feedback loops, allowing agents to self-correct and continuously improve.
How does Nūl’s platform handle the complexity of fashion inventory, with variables like size, style, colour, and location?
RMS: Fashion inventory is inherently multidimensional. A single product can have dozens of variants across size, colour, and style, and each performs differently depending on the store location, customer demographic, and time of year.
Nūl’s platform is architected to operate at the SKU–attribute–store–time level, meaning we don’t treat “a dress” as a single item—we model demand separately for the size M, black variant of that dress in, say, a downtown boutique in Singapore versus a suburban outlet.
Our AI agents understand that a size S beige jumpsuit might sell out quickly in a warm, urban Singapore store catering to younger professionals, while the same SKU in XL, the navy might lag in a different neighbourhood with different customer behaviour.
We integrate real-time POS, ERP, and e-commerce data with contextual signals like weather, foot traffic, and local events. This allows our system to dynamically balance inventory across the network—not just by style or category, but by the exact size and colour that’s needed, where it’s needed.
As a result, Nūl optimises inventory at a level of precision traditional systems can’t match—minimising overstock, maximising availability, and ensuring high sell-through across every variant.
What has been the most surprising or counterintuitive insight you’ve gained from implementing AI in the fashion supply chain?
RMS: One of the most surprising insights has been how micro-level decisions often outperform big seasonal strategies. We used to think the biggest gains would come from improving forecasts at the start of the season. In reality, we saw brands unlocking more value by making small, frequent adjustments during the season—like moving 10 units of a fast-selling SKU from one store to another or tweaking restock timing by a few days.
Another counterintuitive finding? High-performing stores often become overstocked simply because they’re performing well—not because they need more stock. Without AI, brands tend to overcorrect by overfeeding their best stores. Our system shows that sometimes the smarter move is to let a product sell out and redirect those units to where growth potential is higher.
It flipped how we thought about “winning” in fashion retail—it’s less about big bets and more about small, fast moves at the right time.
The post Fixing fashion’s inventory crisis: How Nūl uses agentic AI to stop overproduction appeared first on e27.
