Theories
Stay up-to-date on our team's latest theses and research
Stay up-to-date on our team's latest theses and research
Buying something online seems so simple: one click and a package shows up on your doorstep the next day. But behind the scenes is a massively complex, carefully orchestrated dance, with multiple handoffs across the supply chain, from the manufacturer’s warehouse operators, to the retailer’s packing facilities and delivery drivers.
It’s a feat that this system works at all, but it doesn’t always bode well: over one billion B2C and B2B shipments per year are lost, damaged, or delayed in the US each year.
The real world is messy; we’ll never stop boxes from being dropped or winter storms from shutting down an airport. But we can change how businesses respond to these issues, helping them proactively identify, investigate, and solve problems to better serve their customers
Sean McCarthy learned this firsthand at Amazon Shipping, and partnered with Henry Ou, with experience from Apple and ByteDance, to build BackOps: a platform to deploy agents which allow every business to build and operate efficient, resilient supply chains. We’re thrilled to partner with them and lead their $26 million Series A, along with our friends at Construct, Gradient, and Ten VC.
—
When the day starts, the issues start rolling in: A customer’s order never showed up. Their shipment was water damaged. They ordered 10 units but only received 6.
Each one kicks off a painstaking investigation and resolution process: Pulling data out of 20-year old transportation, warehouse management, and procurement systems. Filling out claims forms on carriers’ websites. Calling warehouse workers and truck drivers to figure out where something went wrong. Along the way, manually recording everything in a tracking spreadsheet.
Large retailers, manufacturers, and distributors employ dozens or hundreds of people who spend all day solving these problems. But the core problem isn’t the labor cost; it’s the service. Amazon and Walmart have made logistics excellence a must-have. If a supplier loses your shipment, then takes a week to follow up and make it right, they’re likely losing your business too.
These logistics workflows have hardly changed in decades because until now, advancements weren’t possible. These businesses interact with thousands of different suppliers, customers, and transportation partners, each with their own set of legacy software and processes. If you want to ship your goods with Theory Trucking, you’ll have to use their portal. If you want to store them with Theory Warehousing, they’ll have their own WMS. Good luck integrating any of them.
—
But today that is changing. AI systems can read emails, make phone calls, and click through web portals just like a person. When a customer calls or emails with a problem, BackOps Relay platform has agents that take the lead, automating end-to-end, cross-platform investigations and escalating to a human only when their input is necessary.
The impact of automating even a single workflow is dramatic – it resolves customer issues faster and more efficiently, and frees up operations teams to focus on proactive mitigations and customer service.
Less than a year after launching Relay, BackOps is powering hundreds of thousands of shipments each day for some of the largest retail, manufacturing, and logistics companies (more announcements on these to come!). Their agents spend thousands of hours on the phone, finding data, and filling out forms so human operators don’t have to; 84% of issues are resolved fully-autonomously.
But the real power of BackOps comes not from automating a single workflow or integrating with a specific legacy platform. BackOps is building a central agent repository integrating all of the tools, workflows, and reasoning for how logistics and adjacent teams operate. Employees use a simple screen recorder to demonstrate how work should be done; these recordings are then converted to reusable tools that can be assembled into different agent workflows over time. It’s like Claude skills for real-world businesses operating legacy software – and it’s what will make BackOps more and more valuable over time.
Sean, Henry, and the BackOps team are well on their way to an ambitious goal: bringing AI agents to the messy real-world of supply chains & logistics. They’ll help the largest businesses operate more effectively, and might just help you get your package sooner too. If you’re interested to learn more about what they’re building, reach out to [sean@backops.ai].
8 months of freezing cold each year is a great reason to lock in.
So Theory ventured up north to Waterloo, Ontario, to meet the next generation of incredible founders and technologists.
We were treated to the snowiest year since 1950. Classes were canceled and streets were buried, the kind of weather that gives you an excuse to stay home.
Instead, over 100 students from all years and programs joined us for two panels aimed at early-career engineering and AI jobs.
Waterloo has a way of filtering for people who show up even when it’s uncomfortable.
The coat racks were overflowing while Prashanth (LanceDB), Nick (Neolific), and Anton (Clover Labs) started the night chatting with our Head of AI, Bryan Bischof (Theory Ventures) on how startups hire.
The conversation was candid, practical, and refreshingly casual, much like how early-stage hiring really works.
We discussed:
For students, it offered a clearer picture of how things actually work. For founders and operators in the room, it was a reminder that the best hires are often the ones who show initiative before they’re “ready.”
The second panel shifted from hiring mechanics to career risk.
I led a conversation with Jakob (Voltra), Sam (Upside Robotics), and Jerry (Akatos House) on what it actually means to break into the startup space.
Our conversation was centred around a few deceptively simple questions:
A recurring theme was that early careers are more about exposure than optimization. Spending your time in an environment that forces you to grow matters more than optimizing for title, brand, or timing.
Theory’s welcomed more than 8 Waterloo interns to our engineering team to work on AI and Data applications. Our internship program focuses on real-products that we actually use: like Pipelines to process, extract, and store call recordings, Multi-modal AI Tools for financial reporting, and Evaluating LLMs.
Additionally, our portfolio of investments is constantly expanding, and many of these companies are also turning to Waterloo to search for interns and full-time hires.
Our visit to campus was focused on recruiting directly for our own team, high-trust introductions to our portfolio companies, and building relationships with founders at the very beginning of their journey.
The benefits of being on campus outweighed the inconvenience of the snow.
Huge thanks to everyone who made this event possible: Waterloo Venture Group for their collaboration on the event, Communitech for the event space, and everyone who helped make the night what it was, including teams from LanceDB, Neolific, Akatos, Voltra, Clover Labs, and Upside Robotics.
Most of all, thank you to the students who showed up despite the weather.
We’ll be back.
.png)