There's a quiet infrastructure layer underneath every meal you've ever eaten at a nice restaurant.

Before the fish hits the Nobu kitchen, before the organic milk reaches the Whole Foods cooler, before the neighborhood taqueria gets its produce - a food distributor moved it there. Coordinated the pricing. Absorbed the volatility. Kept the supply chain intact.

This is a trillion dollar industry. There are 40,000+ distributors in the U.S. alone. They move 99% of the food you eat.

And most of them are still running on ERPs built in the 1980s, stitched together with spreadsheets, phone calls, and institutional memory.

This is the market Tzar Taraporvala and Smayan Mehra decided to attack.

Their company is called Anchr. It's an AI-agent operating system for food distributors. It's backed by a16z Speedrun and leaders from Open AI. And it's one of the more interesting vertical AI bets I've come across - not because the market is glamorous, but precisely because it isn't.

Tzar Taraporvala grew up in India, met his now-co-founder and best friend at age five, and by his early twenties was on track for a PhD in metaphysics. Philosophy of time. Formal logic. The deep stuff.

Then McKinsey called.

In collaboration with QuantumBlack - the firm's elite AI/ML group - Tzar spent years inside Fortune 500 companies figuring out why most of them were failing to capture real value from AI. Not demos. Not slide decks. Actual operational leverage.

The pattern he kept seeing: the companies most ready for AI transformation were the ones where the pain was most obvious, the data was most fragmented, and the processes were most embarrassingly manual. Glamorous tech-forward industries had already been picked over. The real opportunity was in the unglamorous ones that everyone else had passed on.

The food supply chain kept rising to the top of his list. ERPs that looked and behaved like they were built in the Reagan administration. Back offices held together with sticky notes and tribal knowledge. Entire critical workflows - order intake, vendor management, invoice reconciliation - running through a chain of phone calls and spreadsheets. Labor scarce. Margins compressing. And nobody building anything serious for it.

Who better to take it on than two childhood best friends.

Smayan Mehra grew up spending every summer in Silicon Valley, but not as a tourist. His dad and uncle were serial founders who had raised hundreds of millions in venture capital. His childhood included movie studios, water purification warehouses, and random startup offices. The default professional path wasn't "get a job." It was "start a company."

He studied computer science and electrical engineering, interned at Bay Area startups, worked on early vision-models AI at Apple, and then went deep at LiveRamp - building big data infrastructure and AI-enabled products like data clean rooms, serving some of the largest companies on the planet.

He also, with a friend, spun up a crypto fund using a slice of their signing bonuses. Around $20K to start. They rode the 2020 NFT wave to a multi-hundred-thousand dollar portfolio. Outside capital started knocking at the door.

Then the market peaked. They held it all the way down.

That loop was painful. It was also clarifying. Smayan knew he wanted to build something durable - in a real economy, with real customers, with workflows that didn't evaporate in a bear market. Not speculative cycles. Real goods moving through real supply chains.

When Smayan and Tzar came across Wulf’s Fish, a Boston-based seafood distributor, the pain ceased to be a theory and became a reality. They spent months mapping workflows on the factory floor. Orders were manually keyed into ERPs at 3 a.m. Purchasing decisions relied on fragmented spreadsheets. Finance teams reconciled invoices across disconnected systems. The pain was structural, daily, and expensive.

Here's the thing about food distributor ERPs: they were designed to be books: systems of digital record to get organizations off pen and paper.

Everything around the ERP - the actual operational work - is still done by humans. Here's what that looks like in practice:

A large order comes in, probably via phone or email. Someone manually enters it into the ERP. If there's a pricing discrepancy or a credit issue, someone works that out via a separate email thread. The vendor sends an invoice; someone reconciles it against what was actually received. Inventory is tracked in a spreadsheet that may or may not sync with anything. The sales team has their own system. The purchasing team has theirs. None of them talk to each other cleanly.

And the products being distributed are perishables - fish, meat, dairy, poultry. They're price-volatile, demand-volatile, and time-sensitive. Every breakdown in the process has a real cost: wasted product, mispriced SKUs, missed orders, angry customers.

The system works - barely - because of decades of relationship capital and institutional knowledge. The 60-year-old sales rep who knows every buyer's preferences. The operations manager who can reconcile any invoice discrepancy from memory. But labor is getting scarcer and more expensive, and when these people retire, they don't leave behind a knowledge base. They leave behind chaos.

This is the gap Anchr is attacking.

Anchr calls itself an AI-agent operating system for food distributors, but the practical description is more precise: it automates the back-office workflows that live around the ERP and other disparate, archaic systems.

It plugs into existing systems - ERP, CRM, warehouse management, accounting tools, and yes, the spreadsheets that are actually running the business - and then it deploys AI agents to handle the repetitive, high-stakes, human-intensive workflows:

The ERP stays the book of record. Anchr becomes the system of action - handling everything that needs to happen between the entries in that book.

The key insight is what happens over time. As Anchr touches more workflows and sees more operational data, the question of who should own the system of record starts to shift. If you're already automating order intake, managing vendors, handling credits, and orchestrating cross-system workflows - owning the core ledger is less of a conceptual leap and more of a sequencing problem. When, not if.

Anchr sells across a wide band today - from $20M regional distributors up to a $6B+ publicly traded company. But their ideal customer is the $50M–$500M food distributor with a heavy focus on perishables.

Below $50M, the business often looks like an owner, a family, and a few employees running a handful of routes. Highly manual by design. The pain exists, but the process depth and data volume needed to make AI agents shine may not.

Above $500M, you're typically looking at a roll-up - multiple business lines, each operating like a $150–200M distributor with its own tech stack and stakeholder map. That's squarely in Anchr's wheelhouse, but winning there is a multi-unit, multi-stakeholder game, not a single linear sale.

Perishables are the gravitational center because that's where all the pain amplifies. More dynamic pricing. Higher demand and supply volatility. Greater SKU complexity. Much higher cost of waste. The sharper the pain, the clearer the ROI.

This is not a market you can crack with cold email sequences and a PLG funnel.

Food distribution runs on relationships. Buyers and sellers have known each other for 20 or 30 years. They attend each other's kids' weddings. They've been burned by tech vendors before - vendors who promised transformation and delivered headaches. The default posture toward a new software company is skepticism, not curiosity.

So Tzar and Smayan built a different motion entirely.

They created a partner network of industry insiders - former executives, people who've spent decades working with distributors, sometimes customers who've seen Anchr's value firsthand. These partners get on 30-minute calls with their contacts in the industry. They vouch for the founders and the product from a position of real credibility. Aligned incentives all the way through.

The results are striking: their referred-call close rate is north of 50%.

It's slower than high-volume outbound. It doesn't scale like a standard SaaS motion. But in a market where trust is the currency, borrowed trust is the most efficient path to the room - and to closed deals.

Anchr recently had what Tzar described as a "lumpy jump" - four or five customers closing in the same week. The breadth of that range, at this stage, is notable. It suggests the pain is real across the market, not just at one segment.

The long-term roadmap runs along two axes.

Going deep: As Anchr becomes the system of action for more and more workflows, the path to becoming the system of record becomes increasingly natural. This is the playbook - start with automation, earn the trust, own the data layer.

Going broad: The pain Anchr is solving isn't unique to food. Janitorial supplies. Chemicals. Any wholesale vertical with a complex catalog, thin margins, a legacy ERP, and manual back-office workflows. The long-term vision is the operating system for businesses that move atoms - not bits.

a16z Speedrun backed them because the combination is compelling: a massive, under-automated real-economy category; founders with deep AI, data, and ops backgrounds; early commercial traction across a wide revenue band; and a beachhead that logically extends into system-of-record and multi-vertical territory over time. It's exactly the kind of profile you'd expect them to lean into hard in this AI wave.

1. Unsexy markets often have the best economics. Food distribution isn't glamorous. It's also indispensable, enormous, and dramatically underserved by modern software. Founders willing to go deep into "hidden" operational categories often find less competition, more durable moats, and customers who are genuinely grateful someone showed up.

2. AI is most powerful where processes are messiest. The most overhyped AI applications tend to be the flashiest - copilots for tasks people were already doing fine. The highest-value applications are in workflows that are genuinely broken, fragmented across systems, and dependent on human coordination to function at all. That's exactly where Anchr is playing.

3. GTM must match the trust architecture of the market. Some markets reward speed and volume. Others reward credibility and relationships. The founders who figure out early which one they're in - and build their go-to-market accordingly - compress years of wasted effort into focused, efficient execution.

If you're building vertical AI for real-world operational industries, Anchr is worth studying closely. And if you want an intro to Tzar and Smayan, reply here - I'll connect the dots where I can.

foundersbrief. drops weekly. If someone forwarded this to you and you'd like to subscribe, click below:

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more