Every restaurant order has a hidden second order behind it. Before the plate reaches the diner, somebody has to source the tomatoes, find the right case of cheese, confirm the pack size, check the price, enter the line items, load the truck and collect the money. For years, much of that work has traveled through a remarkably human tangle: the voicemail left after service, the text sent before dawn, the email with an attached spreadsheet, the photograph of a handwritten list. Bowie Cheung built his career around the first order, then started a company for the second.
Cheung is the co-founder and CEO of Pepper, a software company for independent food distributors. The company’s pitch can sound like a tidy exercise in vertical software: ecommerce, sales tools, marketing, order automation and payments in one system. The actual job is less tidy. Food distribution is a world of sprawling catalogs, customer-specific prices, perishable inventory, thin margins and relationships built over years. A useful product has to understand the chaos before it tries to organize it.
A habit built in Toronto
Cheung arrived at technology with an engineer’s training and an operator’s education. He studied Engineering Science and Electrical and Computer Engineering at the University of Toronto, worked at Bain & Company, and completed Kellogg’s one-year MBA program in 2014. He also carried an older memory into the work: early jobs in restaurants had shown him that distribution was foundational, even if technology builders rarely talked about it.
At Uber, he became part of the founding Uber Eats team. Toronto served as one of its important laboratories. In December 2015, with Cheung as the city’s general manager, Uber expanded from a short list of pre-made meals into full restaurant menus and introduced a standalone Eats app. The change sounds obvious from the future. At the time it meant replacing an experiment designed around sub-ten-minute delivery with a service that waited for a restaurant to cook the meal.
Cheung stayed for roughly four and a half years, eventually leading regional operations across the United States and Canada. He watched restaurants adopt digital tools at speed. He also saw the mismatch on the other side of the kitchen: the independent suppliers serving those restaurants did not have equivalent technical budgets or products built around their work.
“There are 25,000 food distributors in the U.S. and almost no technology built for them.”Bowie Cheung, 2021
Going upstream
In 2019, Cheung left Uber to start Pepper with Chetan Narain, an early Uber Eats product leader, and Ivana Tesanovic. Their first north star was broad: help restaurants and suppliers work better together. The insight beneath it was sharper. Large distributors could spend heavily on proprietary ordering systems. An independent distributor, even one moving substantial volume, could not easily make the same investment. The software gap was becoming a competitive gap.
Pepper first opened in Manhattan as a network connecting restaurants with smaller suppliers. Then the pandemic emptied dining rooms. Two-thirds of its restaurant partners temporarily furloughed staff or closed, according to Cheung at the time. The young company changed direction quickly, launching Pepper Pantry so households could buy from distributors whose restaurant routes had gone quiet. It expanded from New York into New Jersey, Boston and parts of Pennsylvania.
The pantry experiment brought Cheung back into consumer delivery, but with an important constraint: wholesale operations were designed for big cases and early commercial drops, not a single household’s bacon order. He spoke openly about those awkward economics and about lessons from Uber Eats, including the need to bring restaurant partners along during rapid change. Pepper eventually returned its attention to distributor software. The pivot became less a new identity than a field lesson in how the supply chain behaved under pressure.
One doorway becomes a hallway
Pepper’s first durable doorway was ecommerce: branded apps and websites through which a distributor’s customers could browse a large catalog, see their own products and pricing, and place orders. Once the company was inside the workflow, customers kept showing it more rooms. Sales reps needed alerts about buying patterns and churn. Order desks needed help translating messy messages into structured data. Marketing teams wanted placements that reached operators while they shopped. Accounting teams wanted digital payments, autopay and fewer hours spent chasing checks.
This is how a vertical product becomes a platform: not by collecting features at random, but by following the work. Cheung has described customer need as Pepper’s north star. The phrase earns its keep in the details. Pepper says it has integrated with more than 70 ERP systems, the old and varied systems that hold the data a distributor cannot simply abandon. Implementation is part of the product because a gleaming storefront attached to unreliable inventory is worse than a familiar phone call.
AI after the fog clears
Cheung’s version of AI is grounded in unpleasant inputs. Order Agent is designed to process a voicemail, text, photograph, PDF or handwritten note and turn it into a digital order. Sales Hub looks for patterns that may indicate an upsell or a customer drifting away. These are narrow decisions connected to the distributor’s own information. The standard is not whether the demonstration feels magical. It is whether the correct case reaches the correct customer.
Autonomy, with a thumb on the button
In a 2026 workshop, Cheung demonstrated an inbox agent that sorted messages and drafted replies. He still kept the final send in human hands. His practical sequence: let the system read and draft, judge its work, then widen its authority only when trust has been earned.
That caution is revealing. Cheung called “agentic” a buzzword before explaining the useful idea underneath it: software connected to real data can take scheduled action rather than wait for another prompt. He also said that strange machine errors deserve respect. For a food distributor, where a wrong quantity becomes spoiled inventory or a missed delivery, judgment is part of the infrastructure.
Returning time to the relationship
The emotional center of Pepper’s story is not automation. It is what the automation gives back. A distributor sales representative is part seller, part service desk, part troubleshooter and part local intelligence network. Hours spent retyping orders are hours not spent inside a customer’s restaurant, noticing what is selling, discussing a substitution or asking for another line of business.
Pepper’s customer examples translate software into returned attention. Flanagan Foodservice moved online ordering adoption from 29 percent to 62 percent after launch. Palmer Food Services reported saving 20 to 40 accounting hours each week. The numbers matter because they describe behavior, not screen views: customers choosing a new channel and employees getting part of the week back.
The road after the round
Pepper raised $16 million in 2021, $30 million in 2024 and another $50 million in February 2026. The Series C, led by Lead Edge Capital, brought total capital raised to $100 million. By then Pepper said its customer base and revenue had more than tripled since the prior round, with more than 500 distributors and 100,000 active restaurant and retail operators representing over $30 billion in annual volume.
The capital is being directed toward new agentic products, deeper operating functions, more reliable integrations and larger customer-facing teams. Acquisitions fill in the same map. Pepper bought Kimelo in 2025, adding another vertical distribution toolset, and Y Combinator-backed Alima in March 2026. Alima’s founders joined Pepper to work on product-content matching, data infrastructure and AI-assisted implementation. In July, a partnership with Essendant added enriched content for more than 32,000 items in foodservice and JanSan catalogs.
The story has expanded far beyond a prettier way to order groceries for a restaurant. Pepper is trying to become the system a distributor uses across the day: receiving the strange order, spotting the sales opportunity, showing the right product, collecting the invoice and passing clean information into the ERP. Cheung’s aspiration is an end-to-end platform for critical operations. The discipline is to keep each new piece tied to a job someone already has to do.
“The more we dig into distributor businesses and workflows, the more opportunities we find.”Bowie Cheung, 2026
The useful place to look
There is a recognizable founder move in Cheung’s path: step out of a visible success and inspect the machinery it taught you to notice. Uber Eats made the consumer side of restaurant commerce legible on a phone. Pepper is making the upstream side legible to the people who already run it. The interface may begin with a chef tapping through a catalog, but the harder value sits behind that moment, where prices, pack sizes, inventory, credit and fulfillment have to agree.
Cheung’s public writing carries an operator’s optimism about independent distributors. He argues that technology can give smaller companies the reach of large incumbents while preserving their local knowledge and service. That possibility depends on restraint. Software should remove clerical drag without pretending relationships are inefficiencies. AI should take on a messy order without inventing certainty. Integrations should meet the system already in the building.
The back room rarely gets the glossy product launch. It gets a phone blinking with another voicemail and a stack of checks that still need attention. Cheung has decided that this is exactly where ambitious software belongs. The bet is simple enough to test tomorrow morning: fewer lines retyped, fewer payments hunted, and more people free to do the part of distribution that was human all along.