It began, as so many Amazon stories do, with a single unhappy customer. One book, out of maybe 50,000 shipped that day, arrived late. And somewhere in Seattle, Jeff Bezos hit forward, dropped the complaint into the logistics team's inbox, and added a single character: a question mark.
That question mark — the obsessive, almost unreasonable pursuit of the exception — is the through-line of Harish Abbott's career. It ran from the fulfillment centers he helped build at Amazon, to Deliverr, the e-commerce logistics company he co-founded and sold to Shopify for roughly $2.1 billion, and now to Augment, his AI-first bet to rewire an industry that accounts for nearly a tenth of the American economy. Speaking with 8VC founder and Palantir co-founder Joe Lonsdale on the American Optimist podcast, Abbott made the case that logistics — a $10 trillion global colossus running largely on emails, phone calls, and text messages — is finally ready to be transformed. And he is building the machine to do it.
▶ Watch on YouTubeAbbott is, by any measure, a man who could stop. He is a billionaire. He is, as Lonsdale put it, "rich enough" that he doesn't "need to work again." And yet here he is, already staffed with 50 engineers and hiring 50 more, chasing a problem that has humbled some of the best-funded companies in Silicon Valley. The reason, he explained, is simple: he saw an inflection point, and he couldn't look away.
From a Small Town in India to the Arena
The origin story is quintessentially American, even though it doesn't start in America. Abbott grew up in a small town in India, in what he describes as a middle-class household, before earning a place at the fiercely competitive Indian Institute of Technology — a school where, as he tells it, "literally millions of kids apply, a few thousand get in." From there came graduate work in the United States, at the University of Illinois Urbana-Champaign and Stanford, where he immersed himself in graph theory, mathematics, computer science, and logistics — a combination he describes as "graph theory applied to the physical world."
But he is careful not to credit only the pedigree. "If I look back at all my studies," he said, "growing up in a household — I'm just so thankful for my mom and dad — like just inculcating hard work, grit, perseverance. Those values, that's what I think at the end of it keeps you going." Lonsdale, unsurprisingly, reframed them as "very American values." Abbott didn't disagree.
What pulled him across an ocean was opportunity, and something more subtle: trust. In the India of 1996 and 1997, he recalled, there simply weren't that many young companies to go build. America was the place "where you can go, people will trust you, people will trust your ideas. Merit matters and merit speaks." It is a theme he would return to later, pointedly, when the conversation turned to the politics of immigration.
The Amazon Years: Question Marks and Six-Page Memos
Abbott landed at Amazon when its tagline was still "Earth's biggest bookstore" — a phrase that now reads like a quaint artifact. He arrived as the company was building out its fulfillment centers and logistics network, joining a small band of "maths, computer science people who also had interest in logistics." He wrote a lot of code — some of what became Fulfillment by Amazon — but the more durable inheritance was cultural.
He watched Bezos from a distance and worked closely with leaders like Jeff Wilke, who would go on to run all of Amazon retail. The lessons stuck. There was the near-religious customer obsession, embodied by that forwarded complaint and its lonely question mark. There was the operational rigor: 8 a.m. meetings where roughly 150 metrics across seven or eight warehouses were laid out on a single sheet of paper, and leaders arrived already knowing the five numbers they would be grilled on — and what had gone wrong the night before.
And there was the willingness to cannibalize your own success. Abbott pointed to the Kindle, incubated even as physical books were Amazon's beating heart. On paper it looked like self-sabotage — selling books at $9.99 when the average price was $20 to $25. But Bezos's logic was ruthless and clear: "If I don't do it, someone else will. Might as well let me lead it, let me shape this industry." Abbott has carried that instinct into everything he builds.
Perhaps the most portable lesson, though, was about writing. Amazon's culture demanded that before you called your first meeting, you wrote — sometimes five or six pages — and circulated the document so everyone arrived having read it. "It forced you to really distill your thoughts," Abbott said, "because you knew other people are going to truly read your stuff. You don't want to be looking stupid in front of really good people." He still enforces it. And he pushes his most talented friends to do the same, precisely because success breeds the temptation to skip it. "A lot of people, when they've already had success, they think, okay, this time I don't need to write it down because I already know what I'm doing." His verdict: no matter who you are, write it down. "We all have blind spots."
Deliverr: An AWS for the Physical World
The vision behind Deliverr was audacious in the way venture investors love: make the physical world programmable. Build, in effect, an AWS for atoms — a world where warehouse space, truck capacity, and shipping could be summoned and paid for in small increments, the way you rent a single cycle of compute in the cloud. In the physical economy, Abbott noted, "the leases are really big, the contracts are super long." He wanted to break them into pieces.
The wedge was e-commerce. A new generation of merchants was flourishing on Shopify, Walmart Marketplace, and eBay — but none of them could offer the one- and two-day shipping that Amazon had trained consumers to expect. The only way to match it, Abbott realized, was to build "a CDN for the physical world": a network of warehouses close to consumers. Amazon could afford its own. A "random merchant in Texas doing something pretty cool" could not afford 50 warehouses, nor the inventory to stock them, nor the data science to promise a customer a delivery date.
So Deliverr built it for them — and in doing so had to go "several levels deeper in data science," predicting whether a merchant's scarce 50 units should sit in New York, Los Angeles, or Seattle. As tens of thousands of merchants joined, the economics compounded into a network effect: pool 50 merchants' inventory onto one truck out of the Port of Los Angeles, and everyone rides cheaper. "The truck is going out full," Abbott said. "Everybody gets a benefit from it."
Shopify, whose merchants were increasingly leaning on Deliverr to make faster delivery promises, eventually bought the company. Abbott is candid about what acquisition by a giant actually feels like — a rare bit of unvarnished founder honesty. You gain scale and brand; a $10 billion company suddenly picks up the phone. But "you're no longer the main agenda. You are a sub-point, a sub-mission." There were culture clashes, too: Deliverr was a low-gross-margin, physical business grafted onto Shopify's high-margin software machine, and merging compensation structures and HR systems was, he admitted, "a very hard exercise I never appreciated."
The eventual conclusion he reached with Shopify CEO Tobi Lütke was clarifying: physical logistics was important to merchants, but it was a "side quest" for Shopify, whose main quest was the digital infrastructure of commerce. Better, they decided, to pursue it outside. The business moved to Flexport, where, Abbott reports, "the warehouses are full. They're adding more capacity." Lonsdale, still a Flexport shareholder, seemed pleased.
The Inflection Point: Why He Came Back
Abbott didn't need a new company. He needed a reason. AI gave him two. First, the arrival of genuine reasoning models — "the power of those reasoning models was incredible." Second, exploding context windows; he name-checked Gemini's roughly million-token window. Put them together, and suddenly you could hand a large language model "fairly complex business issues and cases and see what you can do with it."
For a logistics mind, that was the moment. Abbott offered Lonsdale a definition worth pinning to the wall: at its core, "logistics is about trading information so you can trade goods." Every warehouse, truck, shipper, broker, and freight forwarder is really just trading information — so that goods can move from point A to point B, which is what makes "civilizations happen, trade happen, markets happen."
The catch: that information trades over the lowest-bandwidth channels imaginable — emails, phone calls, texts — because the industry is almost incomprehensibly fragmented. Abbott's numbers are staggering. There are, he says, roughly one million truck-driving companies in America. ("It's not kidding — one million.") There are 40,000 to 50,000 brokerages, the average one a single person, though the biggest employ thousands. When Lonsdale pressed on why deep-pocketed challengers like Uber Freight and Convoy hadn't simply fixed it, Abbott's answer was structural.
Getting a million companies onto one platform is "a herculean task." Worse, it's a low-trust industry, and for good reason. Abbott's illustration is vivid: imagine a trucker whose home base is Austin, who hasn't slept in his own bed in three weeks, now stuck in Chicago and desperate to bid on a load heading home. If he reveals all of that, "they're going to give you a really bad price because they know how desperate you are." So everyone hoards information to protect their leverage. Force all of that onto a single transparent platform, and you're fighting human nature.
Enter Auggie, the AI That Doesn't Sleep
AI, Abbott argues, sidesteps the trap entirely. You don't have to force anyone onto a new platform. "We can get you an AI employee that can take on all this tedious work of coordinating." That employee is Auggie — and the way customers have adopted it borders on the uncanny. They rename it. They give it personas and pictures. They invite it to weekly and monthly town halls. An AI teammate, in other words, is becoming part of company culture.
Behind the anthropomorphizing is serious machinery. Auggie works 24/7 and is fully multimodal, handling email, text, and phone — but also Telegram, because many dispatchers are in Eastern Europe and Telegram is how they communicate, and WhatsApp, because a large community of drivers hail from Punjab, India, and that's where they live online. Auggie meets people on whatever channel they already use.
Its work is organized into "workflows" — written documents no different from a standard operating procedure, which Auggie "can follow religiously," including knowing exactly who to escalate to for a decision or an approval. Collecting an invoice, tracking a load, building a load, running collections: these get chunked into workflows Auggie executes at scale. And the scale is already real. Collectively, Abbott said, Auggie is deployed across businesses managing about $25 billion of freight. "It's up and running. It's doing it at scale."
When the AI Gets It Wrong
Abbott is refreshingly unromantic about AI's fallibility. "AI is just not perfect by any means." Augment sorts mistakes into low-consequence and high-consequence. Low-consequence: Auggie calls a contact a third time in a day when it was told to stop at two — annoying, correctable with a guardrail. High-consequence: Auggie once moved to book a load without confirming with a human first. Those get hunted down.
And yes, people try to hack it. With $25 billion flowing through the system, Abbott knows the stakes: "If someone's good at hacking it, you're in trouble." So Augment attacks itself, building adversarial agents whose only job is to try to break Auggie — because, as Lonsdale put it, "you have to be the best ones at hacking it to then find ways to make sure it can't happen."
The most human anecdote came from the day before the taping. Auggie was on an email chain that was meant to stay internal, but a company's customer had been accidentally copied on the list-serve. Not knowing an outsider was watching, Auggie answered a status question honestly: it computed the ETA, realized the truck would arrive 45 minutes late, and said so. The customer was startled. But here's the tell — Auggie's math was flawless; the truck did show up 45 minutes late. "It got all that right," Abbott said. "It's just the way you wanted to communicate to the customer may be different." Like any new employee, "you got to teach it sometimes."
The $300 Billion Nobody Is Talking About
The market, Abbott insists, is bifurcated into two opportunities. The first is obvious: productivity. The industry carries roughly $80 billion in payroll; make that 50% more productive and you've unlocked tens of billions of dollars. Employees freed from the deluge of emails and calls can focus on "relationships, negotiations, route building" — the work that actually moves the needle.
The second opportunity is the one he says nobody's discussing, and it's bigger. Because information trades over such low-bandwidth, asynchronous channels, the system hemorrhages waste. His example: a Nike warehouse with 50 dock doors expecting 50 trucks, seven of them running late. Someone has to notice, then call or log into a portal to rebook an appointment. Fine, if it's business hours. After 5 p.m.? "Good luck till 9 a.m. next day." So the truck arrives late, finds no appointment, and idles all night — a few hundred dollars wasted, per truck. Multiply by three and a half million trucks, at least 10% laying over on any given night, and the losses balloon. Meanwhile, on Nike's side, planned labor sits idle. Everyone loses.
Now put an Auggie on both sides, coordinating in real time — rebooking the appointment, redoing the labor plan, no human required — and the waste evaporates. Abbott's estimate: at least 10% of the industry is wasted this way. That's "a $300 billion opportunity in the US, almost a trillion dollars globally."
The Talent War, and a Clear-Eyed Take on H-1B
Building all this requires people, and the fight for them in San Francisco — where Abbott estimates 1,500 startups are competing — is brutal. His pitch to engineers is mission-first: this is one of the largest industries on earth, many have tried and failed to transform it, and AI is the moment it finally happens. Even "the hundredth engineer," he tells recruits, "is still super early in this innings." Augment is also diversifying geographically, having opened a Toronto office with 10 engineers to tap the University of Waterloo and University of Toronto talent pools.
Then Lonsdale steered into the controversial. On H-1B visas and immigration, Abbott — himself an immigrant who came on that path — was unequivocal: "For us, merit comes first. We're seeking talent no matter where it is." He'd prefer local, American talent where the economics dictate it, but he won't compromise the bar. His hope is that the administration makes the system "merit first," fast-tracking exceptional talent rather than lowering standards. He drew a sharp distinction, echoed by Lonsdale, between fast-growth companies desperate for the very best — willing to pay up for it — and "body shop" firms using visas to arbitrage lower salaries. The controversy, both agreed, belongs to the latter. For Augment, "the cost of us not raising the bar is so high" that arbitrage was never the point.
The Optimist's Case
Abbott closed on the note that gives the show its name. Cheaper, more efficient logistics is disinflationary and good for consumers — and essential if America wants to compete in a new industrial era. The country "will never have the lowest labor cost," he conceded, but it can compete on efficiency through data. And more logistics means more choice: walk into a Whole Foods and you'll see thousands of products, but "there are probably 10 times more products that you don't have access to because logistics is so broken."
His favorite analogy is the internet itself. Slow, restricted internet gave you a little content; broadband gave you YouTube and a world of it. He recalled arguing with a high-school teacher who insisted a 28k modem was fast enough to load the page you were reading — missing the point entirely. "It opens up new possibilities." Better logistics infrastructure, he believes, will do the same for the movement of physical goods. Beyond his own domain, he's bullish on robotics finally reaching mass use, and on autonomous vehicles — the Teslas and Waymos already rolling through Austin, and the autonomous trucks he expects to drive down costs further. The interesting problem, he noted, is the coordination between human and autonomous fleets — which is, of course, exactly the kind of information-trading problem Auggie was born to solve.
He doesn't have to do any of it. That's what makes it interesting. "You're one of the most successful entrepreneurs who already has all the money you need," Lonsdale observed, "but it's obviously a very fun job to be in the middle of this transformation." Abbott, the man who once answered a question mark from Jeff Bezos, seems to have found a new one worth answering.