Failed project comes back to life: Why did a16z write a $35 million check to this "debt collection" company?

Byline: Leo

Imagine this scenario: you’re the CFO of a well-known consumer brand, whose products are sold at big retailers like Target, Walmart, Amazon, and more. On paper, everything looks normal. But every month during settlement, you notice a strange pattern: the retailers always pay about 20% less than what the invoices say. It’s not a one-off or an occasional thing—it happens every single month. You want to prove that they’re underpaying, but to do that, your team has to comb through hundreds of pages of shipping records, log into dozens of different retailer portals, and reconcile thousands upon thousands of invoice line items. How big a workload is this? Your finance team can’t possibly handle it. In the end, they can only give up on pursuing those smaller deductions, watching millions of dollars slip through your fingers.

This isn’t a fictional plot. It’s a real story that plays out in the consumer goods industry every day. I recently dug into a company called Glimpse. They’ve just raised a $35 million Series A, led by Andreessen Horowitz. This Y Combinator–backed company is using AI to solve a pain point in an industry worth tens of billions of dollars: disputes over retail chargebacks. When I saw their data, I was shocked: a $1 billion consumer products company had Glimpse’s AI agent review 17,000 chargeback records in under 24 hours and identify millions of dollars in recoverable revenue. And if done manually, that same workload would take nearly two years.

The most expensive invisible cost in retail

Before I dive into Glimpse’s solution, I want to explain just how serious the retail chargeback problem really is. Many people might not realize that transactions between consumer brands and retailers aren’t as straightforward as most people think. Brands bill retailers, retailers pay—seemingly simple. But in reality, retailers almost always deduct part of the amount when paying, then provide a reason—for example, goods damaged, short shipments, packaging not meeting requirements, and so on.

Some of these deductions are legitimate and, yes, it can be the brand’s fault. But a significant portion are invalid chargebacks, meaning the brand didn’t do anything wrong, yet the retailer still took the money. The problem is that proving these chargebacks are invalid requires an extremely tedious process. The finance team needs to log into multiple retailer systems, pull scattered documents, review line items one by one, match them against internal records, and manage the entire dispute workflow. The process is so complex and time-consuming that most brands can only handle the larger deductions selectively, treating the rest as a loss.

One data point really stuck with me: industry analysts estimate that consumer goods companies have $8 billion in valid disputes each year that go unrecovered due to insufficient operational capacity. This isn’t a small number. For a mid-sized consumer brand, invalid chargebacks can make up 5% of retail revenue or even more. Imagine that your annual retail revenue is $100 million—then $5 million evaporates in this process, and you can’t recover it because you don’t have enough people and systems to handle it.

To make matters worse, the complexity keeps increasing. Take Amazon Vendor Central as an example: it has more than 30 different chargeback categories, from shipment delays to packaging violations—each with its own rules and dispute workflow. A finance team at a mid-sized consumer goods company typically consists of only a few people, and they simply don’t have the manpower to handle even half of the chargeback disputes. That’s why this problem has persisted for so long—until now, when AI technology has matured enough to make solving it possible.

How strong is Glimpse’s AI solution?

When I learned how Glimpse works, I realized they found a very clever entry point. They didn’t try to build a generic financial software product. Instead, they focused on solving a specific problem that’s both concrete and high-impact: automating the review and dispute process for retail chargebacks. Their platform uses AI agents to carry out the entire workflow, from data collection to dispute resolution, fully automatically end to end.

Specifically, Glimpse’s system first automatically logs into each retailer’s portal, finds all relevant documents, and centralizes them. This sounds simple, but in practice it’s incredibly complex, because every retailer’s systems are different and the data formats are completely different. Some use EDI (electronic data interchange), some are PDF documents, some come via email, and some are buried deep in the website. Glimpse’s AI needs to understand all these different data sources and integrate them into a single unified view.

Next, the system categorizes each chargeback. This step sounds straightforward, but it requires deep understanding of business logic. The AI needs to know what type of chargeback it is, which products are involved, when it occurred, and which order it corresponds to. Then it verifies these chargebacks against the brand’s internal data—for example, supply chain records, promotional calendars, shipping manifests, and more. Through this cross-validation, the AI can determine which chargebacks are legitimate and which ones are invalid.

Most importantly, once the system identifies invalid chargebacks, it doesn’t stop there. It automatically submits the dispute, follows through on the entire process, tracks the progress of cash recovery, and syncs all information back to the brand’s ERP system. The whole process is automated from start to finish, with no need for human intervention. Of course, Glimpse also keeps human-in-the-loop parts, primarily to ensure result quality—such as following up on disputes to drive resolution and cash recovery, and performing quality assurance at key steps like categorization and data extraction.

What I find most impressive is that this system gets smarter the more it’s used. Each time it handles a chargeback, it learns and improves, continuously optimizing its abilities in categorization, verification, and resolution. Over time, this creates a compounding data advantage: every new integration, every new customer, makes the network more intelligent and more effective. That’s why Glimpse can achieve a 91% dispute win rate while reducing up to 80% of manual labor time.

A customer case I saw makes the point especially well. Evermark is the parent company of the Suave brand and Chapstick. Their Senior Director of FP&A, Sean Quinn, said: “Like most major consumer brands, Evermark used to have to set a minimum threshold amount for chargebacks that could be reviewed, because we simply didn’t have enough time or headcount to review every single chargeback. By using Glimpse’s AI to automate review and reconciliation workflows, we not only removed that threshold—we unlocked a new source of cash flow, bringing in millions of dollars in revenue that used to be considered ‘write-offs’ or just the cost of doing business.” The key phrase here is “remove the threshold”—in the past, they could only handle chargebacks above a certain number. Now every chargeback is reviewed, meaning a large volume of previously ignored small deductions can now be recovered.

From failure to success: the transformation story of three Purdue University friends

Glimpse’s founder story is interesting in its own right. It reflects the most important thing about entrepreneurship: the ability to learn fast through trial and error and to pivot decisively. Founders Akash Raju, Anuj Mehta, and Kushal Negi were classmates at Purdue University. Their original project was completely different from what they do today: a company that did product placement for Airbnb. The project started in 2020, but by 2024, the founders realized the product-market fit wasn’t strong enough and decided to completely pivot.

In Akash Raju’s own words: “In the end, we felt we lacked product-market fit, so we decided to do a hard pivot. During this process, we got access to brands’ back-office operations and the chaotic situation surrounding what’s sold in retail—ultimately leading us to create Glimpse as it exists today.” Such a pivot requires tremendous courage, because it means giving up all the work you were doing before and starting from scratch. But it was exactly that decision that led them to a truly valuable problem.

Even more impressive to me is that, during the pivot, the founding team sometimes didn’t even pay themselves—fully fueled by passion and belief in the product. That “we won’t stop until we succeed” stubbornness carried through everything they did. And that spirit was recognized by investors. They connected with Andreessen Horowitz investors through a shared founder friend. As the business expanded, they built a strong relationship that ultimately led to this $35 million funding round.

Interestingly, there’s also a bit of story behind how this round was named. After their business pivot last year, Glimpse raised a $10 million round led by 8 VC, which they called a Series A. Now this $35 million round is also called a Series A, while the earlier $10 million has been redefined as the seed round. Add in funding before the pivot, and the company has raised a total of $52 million. Flexible naming of funding rounds like this isn’t uncommon in startup circles, especially for companies that have undergone major pivots.

You can also see the team’s execution ability from their performance in 2025. When entering 2025, they set a clear strategy: hire great talent and work together, deeply embed into customer workflows, and adopt an in-person go-to-market strategy. Their internal slogan was “everywhere”—building trust by showing up consistently and providing help. The strategy worked. In 2025, they achieved 10x revenue growth, increased revenue recovered for customers by 10x, grew the volume of invoices processed by 5x to reach $1 billion, expanded the team size by 5x to more than 25 people, and grew the number of customers by 3x to over 150 consumer brands.

The real value of AI agents in financial automation

Glimpse’s case made me understand more deeply the value of AI agents in enterprise applications. Over the past year, everyone has been discussing AI agents, but many times it stays at the concept level or in demos. Glimpse, however, shows the real business value that AI agents can create in real-world workflows: directly impacting an organization’s profit margins.

I think Glimpse’s success comes down to choosing the perfect entry point. The chargeback dispute problem has several characteristics that make it especially suited to AI. It’s a highly repetitive task that generates tens of thousands of events every month. It involves processing a large amount of unstructured data, from PDF documents to web data to email. It requires data validation and matching across multiple systems. It also has clear success criteria: did the dispute succeed, and was the money recovered? Put together, these factors allow AI agents to deliver their biggest advantage.

More importantly, this problem has an immediate ROI. One of Glimpse’s investors once said they were looking for “software that recovers its costs within the first quarter”—and chargeback recovery tools completely fit that standard. When a brand can recover millions of dollars per year through Glimpse, the software subscription cost seems comparatively trivial. This clear value proposition helps Glimpse acquire customers quickly and maintain very high customer retention.

I also noticed Glimpse hasn’t stopped at chargeback disputes. In 2025, they rolled out several important platform capability expansions. In addition to the initial KeHE and UNFI, they now support multiple retailers including Target, Walmart, Amazon, and Sam’s Club. They launched end-to-end AI revenue recovery agents that can handle the full workflow of chargeback retrieval, encoding, verification, and claim submission. They also developed automated cash application capabilities—automating one of the most painful workflows for finance teams during month-end close.

Especially worth mentioning is the AI chargeback line-item detail functionality they launched. Each chargeback comes with supporting documents—typically more than 100 pages—filled with a jumble of retailer details, SKUs, brokers, and unstructured information. Most brands don’t use this data—not because it lacks value, but because manually processing it at scale is simply not feasible. Glimpse’s AI can extract every relevant detail into structured tables, unlocking a whole new layer of intelligence: accurate broker commission calculations, profitability analysis by retailer, trade analytics, promotion performance evaluation, margin improvement strategies, and more.

This also makes me think about a deeper question: what is Glimpse truly building? On the surface, they’re an automated chargeback dispute tool. But in reality, they’re building AI infrastructure for CPG brands. Their CEO, Akash Raju, said: “Our vision is to become the AI infrastructure for CPG and retail brands.” That positioning is very smart. Chargeback disputes are just an entry point, a wedge that can quickly prove value. But by solving this problem, Glimpse gains deep access to brand retail operations data—enabling expansion into broader retail compliance automation.

Reportedly, their roadmap includes modules like promotion reconciliation, trade spend optimization, and retailer payment behavior prediction analysis. A person familiar with the deal said the company could ultimately build a complete “retail finance operations platform,” sitting between the ERP system and retailer portals and automating the entire order-to-cash cycle for CPG brands. If that vision comes to fruition, Glimpse won’t be just a tool—it will become core infrastructure for CPG brand operations.

What this means for the entire industry

I believe Glimpse’s rapid rise and successful funding mark a new phase in enterprise AI adoption. In 2025, consumer AI applications dominated the headlines, but now investors are starting to heavily back AI tools that solve unglamorous yet expensive business problems. Chargeback tracking, invoice reconciliation, compliance monitoring—these won’t produce flashy demos, but they directly impact EBITDA. That’s exactly the kind of value proposition that can survive during economic downturns, and it’s also why Andreessen Horowitz is willing to pay high enterprise SaaS multiples.

I’ve noticed an interesting trend: the competitive landscape is heating up quickly. Claimify raised a $12 million Series A last year for similar automation of retail disputes. Meanwhile, traditional players like HighRadius and Billtrust are adding AI modules to their accounts receivable platforms. But Glimpse’s Y Combinator background and early traction with mid-market CPG brands gave it an advantage during the fundraising process. Reportedly, the company’s revenue grew 14x year over year, although the specific ARR figures weren’t disclosed.

The continued involvement of 8 VC says a lot as well. They led Glimpse’s 2024 seed round and continued to participate in this Series A. 8 VC has a track record of investing in vertical SaaS that automates manual financial processes. Co-founder Alex Kolicich previously told Forbes that what 8 VC looks for is “software that recovers its costs within the first quarter”—and when brands can recover six- or seven-figure amounts each year, chargeback recovery tools fit that ROI model perfectly.

From a more macro perspective, Glimpse’s success validates a simple argument: there are big businesses hiding inside the “back-office middle layer” work that causes CPG brands to lose millions of dollars every year. With Andreessen Horowitz’s support and a product that can deliver measurable ROI from day one, the company is well positioned to lead the category of retail dispute resolution.

The real test will come in the next 12 months: whether Glimpse can scale beyond its initial customer base and prove that the platform can handle the operational complexity of enterprise-level CPG brands managing tens of thousands of SKUs across dozens of retail partners. If the product delivers on its margin recovery promises, this Series A round could look like a bargain by the time the company raises its next round.

I particularly agree with Andreessen Horowitz partner Joe Schmidt’s view: “For decades, retail back-office operations have relied on spreadsheets and fragmented workflows. What impresses us is customer referrals—Glimpse is delivering clear, measurable ROI. By embedding AI directly into core financial and operational workflows, they’re turning this market from incremental tools into infrastructure for modern brands.” This captures why Glimpse matters so precisely: it isn’t just improving existing processes at the edges—it’s redefining how those processes should work with AI.

My thoughts on AI transforming traditional industries

Glimpse’s story has given me a deeper understanding of how AI is transforming traditional industries. The consumer goods industry is one of the largest markets in the world, but it’s barely been touched by modern software. When brands sell to major retailers, they typically have to deal with fragmented, unstructured data scattered across dozens of retailer portals and legacy systems. Analysts spend countless hours pulling data from portals, extracting line items from documents, and working in spreadsheets to drive workflows like reconciliations, disputing invalid fees, and manually applying cash—work that directly affects profit margins but has almost no strategic leverage.

The entire industry spends more than $100 billion annually on back-office labor, and the productivity gains those tasks have received from previous waves of enterprise software have been very limited. AI is the first technology that makes end-to-end automation of this kind of complexity possible. I think that’s the most important insight: not every problem can be solved with traditional software. Some problems only become solvable effectively after technology advances to a certain threshold.

I’m also thinking about why now is the best time to transform these traditional industries with AI. Technically, large language models are now strong enough to understand and process unstructured data. Commercially, enterprises are facing margin pressure and need to protect profit margins—especially as retailer consolidation increases power and adds stricter compliance requirements. Amazon Vendor Central alone has more than 30 different chargeback categories, from delayed shipments to packaging violations. Finance teams at mid-sized CPG companies often lack the manpower to dispute even half of them. That’s why AI-driven platforms like Glimpse become essential infrastructure rather than optional tools.

I believe we’ll see more and more companies like Glimpse emerge—focused on using AI to solve specific, specific pain points within specific industries. These companies won’t try to build generic AI. Instead, they’ll go deep into a vertical, truly understand the business workflow, and then redesign those workflows using AI. This approach is harder than building general tools because it requires deep industry knowledge, but once it succeeds, the barriers are higher and the value is greater.

Glimpse’s $35 million Series A is just the beginning. I expect that over the next few years, we’ll see a flood of capital into this space, accelerating the adoption of AI in traditional industry back-office operations. Companies that can find high-value entry points like Glimpse, prove ROI quickly, and then expand platform capabilities will have the opportunity to become infrastructure-level players in their respective domains. And for CPG brands, embracing these AI tools is no longer a choice—it’s a survival requirement. Brands that adopt earlier and use AI to optimize operations more effectively will gain a significant competitive advantage.