Introduction
Everything you've ever ordered online passed through a fulfillment warehouse before it reached your door. Behind those warehouses is a massive industry of logistics companies: 3rd party logistics providers (3PLs), freight brokers, forwarders, and in-house operations at retailers and manufacturers that store, pack, ship, and coordinate the movement of everything from e-commerce orders to industrial B2B distribution, perishables, pharmaceuticals, and beyond.
Ever since the inception of this business, every major increment in volume has meant a proportional increase in back-office work. And for just as long, there's been only one way to solve this problem: hire more back-office workers. This has resulted in a situation no logistics operator likes: labor costs eat everything and result in single-digit profit margins.
Prox's premise is that for the first time in history, it's possible to build a company that gives logistics companies a digital workforce to handle back-office coordination. Agents that first work alongside human workers, making them incredibly efficient, and then take on entire processes end to end.
That's what Prox is, and that's what we're doing. We're starting with 3PLs, where the pain is most acute and deployment is fastest. We're staying laser-focused on 3PLs: first mid-market (where our current traction and inbound are), then enterprise, eventually growth with self-serve.
Founders and why now
We know how tedious warehouse back-office work is because we both basically grew up inside a warehouse. Both of our families own small warehousing businesses in Ukraine, so we spent all of our free time working at the warehouses, helping to do things like coordinate carrier pickups, resolve inventory discrepancies, and process inbound receipts. We first met in 2016 at a startup summer camp in the Carpathian Mountains and bonded immediately over the fact that we could both nerd out about warehouse operations with each other.
Both being startup and tech minded (Greg was participating in physics olympiads and Dima was coding websites as freelancer) we often jammed possible warehouse automation we can build, but the tech of late 2010s was just, frankly, too limited.
We then set off to do a lot of other things. Right after turning 18, Dima moved to the US, got into MIT for CS, and started building agents almost immediately after the GPT-4 API came out - working with Pulley's (W20) founders to build 15+ internal tools for their team, then working on robotics foundation models at MIT CSAIL before graduating in May 2025.
Greg moved to Canada at 17, studied CS for a year, then dropped out to build Rektoff, a blockchain security firm that developed frameworks and provided enterprise security training across multiple ecosystems including Solana. His security standards were adopted by teams managing over $100B in assets. He then led DevRel at Runtime Verification, winning contracts with NASA, Jump Trading, and the Ethereum Foundation.
We both stayed in touch and frequently discussed agents for logistics. We understood that for agents to perform the multistep coordination work like the kind performed in 3PLs, the underlying models needed to become incredibly good at long-horizon tool calling and codegen.
We'd been tracking this for years, since GPT-4, but the models performed poorly on our internal benchmarks of real 3PL work. That changed when Claude Sonnet 4.5 came out this fall, blowing away our expectations. We realized it was finally possible to build digital workers that could handle coordination work in logistics.
We knew the technology was ready. What we learned after joining YC in Fall 2025 was that enterprises are even more ready. And we know this because we deployed agents at ShipBob (one of America's biggest 3PLs, 100m packages processed in 2025) going from first contact to agents working live in production in just 5 weeks (2 weeks of actual on-site work).
Traction
When we first made contact with ShipBob, they asked us to deploy agents to handle lost package claims. When a package is lost, ShipBob credits the customer immediately, then files a claim with the carrier (e.g. FedEx, UPS, USPS) to recoup that cost. With 160,000+ lost packages per year and only 6 people handling claims, they can't keep up, which results in millions in unrecovered claims.
We expected a typical enterprise sales cycle - months of pilots, committees, procurement. Instead, Dhruv (ShipBob's CEO) told the team: "We need this yesterday. Ship it live as fast as possible." That's not how enterprise logistics usually works, but the C-suite sees what's coming - AI will either be their competitive advantage or their extinction event.
In 2 weeks, we deployed claims processing agents in production. We competed against Pallet and Augment for this contract - ShipBob chose us because we moved faster and understood their workflows. The agents now own the claims filing workflow, file the majority of claims, and flag edge cases for human review. This generates $5K MRR for Prox. The 6-person claims team has transitioned from repetitive filing work to higher-value work of handling merchant relationships.
Claims processing is one piece of a bigger challenge every 3PL faces: back-office operations. 3PLs operate massive back-office teams handling:
- Claims processing: Filing carrier claims for lost/damaged packages
- Merchant billing: Preparing accurate invoices, reconciling costs to revenue
- Account management: Customer support, retention, upselling
- Reporting & compliance: Order tracking, documentation, regulatory filings
These workflows are mostly US-based, completely manual, with zero visibility until something breaks. We've signed an LOI with ShipBob for $60K ARR to automate their billing, reporting, and merchant support workflows—deploying in December.
And it's not just ShipBob. Without any outbound sales, 5 major 3PLs reached out after seeing our ShipBob case study: GoBolt, Portless, ShipCube, ShipMonk, 247Fulfillment. That's $1M+ in pipeline ARR. Every 3PL benchmarks against ShipBob—when they see ShipBob using AI for back-office, they see it as existential.
We've talked to dozens of other 3PLs, and every single operator said the same thing: "Back-office is our biggest cost center and we have zero automation." At Portless (500K-800K orders/month), their CFO is still manually doing billing reconciliation. At GoBolt (Canada's largest 3PL), they have 30+ support agents handling merchant inquiries manually.
Moreover, it's not just 3PLs. Landstar, a publicly traded logistics company, launched AI initiatives just last week with their VPs mandating all teams to map workflows for automation. Their back-office teams told us what we're building is 'desperately needed'.
Principles Prox Follows
When we first contacted ShipBob, we weren't the only ones in the running. Two well-funded competitors had already been in conversations with them for weeks, yet we ended up winning the contract.
We won because we built Prox on three core principles that directly address what 3PLs actually need.
Principle 1: Relentless Laser Focus on Third-Party Logistics Companies and Their Problems
Our competitors contacted multiple departments at ShipBob (back office, warehouse operations, freight), positioning themselves as "agents for everything in logistics" who could automate any workflow.
This scattered approach backfired. When you're trying to be everything to everyone, no one can champion you. No department at ShipBob could relate to an "everything in logistics" AI agents company. No one became a champion. Decisions stalled.
We came in and said: "We only automate back-office work at 3PLs and nothing else. We grew up in this and we know how it works." The back office team became our champion immediately. They owned the decision, and it moved fast.
Principle 2: Fast Customer Deployment is an Engineering Problem. Engineers Build Tools to Go Fast, Not Throw Headcount at the Customer
Our competitors' approach is to send forward-deployed engineers on-site for a month to custom-build integrations for each customer. ShipBob's VP told us: "We can't commit months of VP time to babysit on-site engineers."
We deployed in 2 weeks on-site because we invest our engineering talent in building exceptional internal tools, not in throwing bodies at the problem. Our goal is to drive deployment time down to days, and we're getting there by making our engineers more effective through better tooling.
We, as a startup, can't afford to throw forward-deployed engineers at every customer. So we build internal tools that let us move fast. Each deployment teaches us how to make the next one faster. Competitors scale by hiring more FDEs. We scale by making each customer require less custom work.
Principle 3: Trust is Everything in Logistics. Hard-Earned Trust Comes From Being Radically Clear With Customers.
Logistics has a complicated relationship with Silicon Valley, and 3PLs privately tell us they're cautious of well-funded agentic startups. The pattern they see is a 3-6 month free pilot where the startup integrates deeply with all their data. Then after 6 months, the startup has so much leverage (as they're integrated into everything) that they can jack up the price whatever they want.
That's why when our competitors offered 6-mo free pilots, we decided to charge from day one.
ShipBob went with us because we were clear from the start: here's what we're charging, here's what the model is, here's what this costs. No surprises.
In logistics, relationships matter more than features. We know this because we grew up in this industry. They trusted us because we were upfront, clear, and transparent from day one. That's how you build trust that lasts.
How we think about market
We're laser-focused on mid-market 3PLs for the next 12 months—that's where our current traction and inbound are. But it's also the right strategic choice: mid-market 3PLs are tech-forward, move fast because they're actively competing for market share, and they're in the fastest-growing segment of logistics (ecommerce).
We'll move upmarket to enterprise 3PLs once we prove repeatability with 10-15 mid-market customers. The long tail of growth 3PLs (4,000 companies) is Phase 3. We'll go after them only when models are mature enough for stable self-serve and we can build on what we learned upmarket.
Mid-Market 3PLs (200-500 employees)
- 600 companies in the US
- Prox pricing: $500K ACV
Enterprise 3PLs (Phase 2: In 12 months, 10,000 employees average)
- 50 companies in the US
- Prox pricing: $5M-30M ACV
Growth 3PLs (Phase 3: Self-serve version, 50-150 employees)
- 4,000 companies in the US
- Prox pricing: $180K ACV
Total TAM: ~$2.5B (only in 3PLs)