Decoding the Supply Chain

Welcome to the first edition of Decoding The Supply Chain where I’ll be sharing my unsolicited opinion on different trends in logistics. As someone who’s spent the last five years investing in supply chain and logistics and now building at BackOps AI, I’ve been on both ends of investor hype and the operational reality.

This article will explore how AI copilots and autonomous agents are shaping the future of work, where they actually add value today, and how to best deploy AI in your organization. As a disclaimer, I am in no way, shape, or form an AI or supply chain expert, but I hope to share a unique perspective on a constantly shifting industry. We’ll establish definitions and baseline knowledge to ensure that everyone starts on the same page. From there, each article will build on the previous one, creating a clearer and more complete understanding over time.

The broader series will deep dive into supply chain trends in an easily digestible manner - while slipping in some Bay Area sports references.

If you have any interest to connect and learn about what we’re building at BackOps AI or if you just want to chat, feel free to reach out to me at: sean.elmurib@backops.ai

AI Copilots and Autonomous Agents

In recent years, artificial intelligence has moved from being a buzzword to a practical tool that people utilize in both a professional and personal capacity. With the rise of platforms like OpenAI, Gemini, and Anthropic, using AI in your everyday workflow has become the norm - whether it's helping me come up with dinner ideas based on the ingredients in my fridge or offering deeper insights at work. It's clear that AI is essential for staying competitive, but navigating the best way to implement it in your organization can still be a challenge. That being said, there are so many AI tools and approaches available, it can be difficult to determine which one is the best fit. To understand the direction AI is heading, it's helpful to examine two key developments: AI copilots and autonomous agents.

We’re going to start with a quick history AI lesson. Following the launch of ChatGPT in late 2022, there was a surge of entrepreneurs trying to apply the generative AI technology to real world problems. AI copilots became the hottest thing since sliced bread and were designed to enhance human productivity by offering real-time suggestions, automating certain tasks, and improving efficiency - all while leaving the final decision making to the end user. A way that helps me conceptualize this, is by thinking of an AI Copilot like adaptive cruise control in a vehicle. They help regulate your speed, staying in the correct lane, and detecting hazards, but at the end of the day, you as the driver remains in control.

AI copilots are especially valuable in industries where human judgment and creativity are essential. For example, lawyers are using tools like Harvey AI to summarize lengthy documents, retrieve relevant information, and even generate preliminary legal analysis. However, despite these efficiencies, the final decision always rests with the human professional. In sectors that require nuanced decision-making or creative input, AI copilots serve as powerful assistants rather than replacements.

As AI technology has advanced, however, the scope has expanded to include autonomous agents - AI systems that can operate largely without human intervention once a goal is set. These systems are capable of managing entire workflows, from start to finish, based on predefined rules. Think of them as the equivalent of a self-driving car. Once you provide a destination, the car takes over, navigating traffic and adjusting speed, without requiring a human at the wheel - only seeking assistance in highly complex or untrained scenarios.

When the time for full adoption comes - logistics, e-commerce, and manufacturing are industries that have well-defined and repetitive processes that are primed for autonomous agents. These agents excel in environments where routine tasks - like order processing, inventory management, and customer support - can be easily automated. At BackOps AI we are already automating back-office functions, learning from real-world data to continuously improve their performance with minimal human oversight. As they handle more tasks, they become more efficient, able to process large volumes of work faster and with fewer errors.

The rise of AI copilots and autonomous agents shows just how far things have come, from tools that help you make decisions to systems that can actually carry them out on their own. But no matter how impressive these technologies seem, they’re still only as good as the data behind them. If the data is messy, outdated, or incomplete, even the smartest AI will give you bad results. That’s why the next big focus isn’t just on building better models, it’s on feeding them better data. We’ll dive deeper into the logistics data problem in the next post - stay tuned.

Is The Logistics Industry Ready For AI?

There’s often a disconnect between venture capitalists, startups, and operators when it comes to deciding which technologies should be prioritized and developed. From my time in VC, I saw firsthand how nearly every investment thesis now includes AI in some shape or form. This sends a strong signal to founders, who often feel pressure to build AI-native products that promise to disrupt entire industries. The problem is that, the pressure can lead entrepreneurs to prioritize investor expectations over the needs of the operators and end-users who actually have to implement and rely on the technology.

I don’t disagree with the broader VC belief that AI is becoming essential for long-term competitiveness. But there’s a clear gap between what investors are excited to fund and what operators are realistically ready to adopt. That misalignment creates friction and sometimes failure when the white board vision doesn’t match the reality on the warehouse floor.

Nowhere is that more apparent than in logistics. On factory floors and in warehouses, the sentiment toward AI typically treads on the line of caution. Some of that hesitation stems from the disruption new technology can cause to established workflows. But there’s also a deeper, more personal fear: that AI could eventually replace jobs altogether. Now that for these solutions to see widespread adoption, operators need to be fully bought in. That means entrepreneurs must work closely with the people on the ground to empower them with AI and identify the real bottlenecks they face instead of coming up with a square peg solution for a workflow that doesn’t need it.

As exciting as autonomous agents are, not every company in the logistics may be ready for them to run mission-critical operations just yet. And that’s okay. Honestly, we need to slow our horses. Most teams don’t want, or need a black-box system making decisions on their behalf. What they’re open to - and in many cases, eager for - are copilot solutions: assistive tools that provide suggestions, automate tedious steps, and push to a human when decisions carry operational or customer risk.

Adoption won’t happen if AI demands a complete rip and replace of the systems operators already know and trust. Instead, AI should feel like a natural extension - an assistive layer that enhances existing processes without upending them.

Earning operator buy-in is just the first step in a successful bottoms-up approach. The next and often harder battle is convincing the decision makers. Entrepreneurs have to show that AI is not only powerful in theory, but reliable in the real world scenarios. Unfortunately, many execs have been burned before, either by the flood of cold outreach from AI-powered BDRs, or by a tech provider’s AI that hallucinated them into hot water with a customer.

Whether you're deploying a co-pilot that loops in a human or an autonomous agent that fully automates a workflow, the key to winning over decision makers is building trust in the AI. Jumping straight into full-scale automation can backfire, as it's often more effective to start small. At BackOps we prefer to focus on a single, repetitive, high-friction task for one customer (e.g., automatically responding to tracking inquiries for Patagonia). Once the AI proves its value there, you can either scale that use case across more customers (like North Face and Arc’teryx), or expand into additional workflows for the same customer (such as handling automated reshipments for Patagonia).

The path forward with AI in logistics is building trust - one task at a time with humans still in the loop.

The Landscape

If it feels like there’s an AI agent for every part of the supply chain these days, it’s because there is. Freight, fulfillment, customer service, claims, procurement, planning all have a startup (or five) promising an AI solution for it. The hard part isn’t finding an agent, rather it’s figuring out which ones are actually useful for your operations and which ones you can scale with

First off, we salute everyone who is taking the jump and building AI solutions. As mentioned, it’s not easy to sell AI into legacy industries, so to see everyone going about it in different ways is very admirable. Secondly, I’m not an AI expert that can tell you the minute differences between all of the providers, so I won’t pretend that I am. What I will do is give a big shout out to all of our peers innovating in the space!

Broadly speaking, the market is falling into a few buckets:

Front-of-house agents are a natural and critical starting point for many logistics teams exploring AI. These tools focus on customer-facing workflows like quoting, order tracking, or support agents that can reliably handle inbound across chat, email, and phone. They are able to offer a visible and immediate ROI, especially for teams interfacing with end customers or retail partners. They also provide a lower-risk entry point into automation, allowing companies to get comfortable with AI adoption without touching the more critical operations. This is the perfect wedge for tech vendors to prove themselves - successfully offloading repetitive front-end tasks, puts you in a strong position to tackle more complex back-end processes next.

Back-of-house agents are the big unlock to freeing up human capital. Manual repetitive like claims, reshipment, scheduling dock appointments, invoice reconciliation plague workers’ time as they have to jump through multiple systems and communicate with different parties in order to resolute their issue. This is exactly where AI can make the biggest dent. These are the workflows that are somehow everybody’s job and nobody’s job at the same time and usually fall at the wayside.

Blended agents are built to help satisfy both the customer experience and internal operations. Fulfillment claims are a perfect example of this. When a customer reports a missing item, instead of kicking off the typical support queue, an AI agent can engage immediately: verifying the order, checking shipment data, identifying the issue, initiating a claim with the 3PL or carrier, and automatically trigger a reshipment.

For the customer, it means faster resolution with less friction. For the back office, it significantly reduces manual workload - no one will be finger pointing at their peers claiming it wasn’t their job, there won’t be any more redundant data entry, no status chasing, and no toggling between systems. The entire process becomes smoother and more consistent and will dramatically change how your team scales.

Then you have vertical-specific agents:

• Freight-focused solutions are tackling key areas like quote automation and capacity sourcing. These tools are especially valuable for freight brokers who manage high volumes across multiple carriers. Automating tasks like check calls, load updates, and appointment scheduling, these solutions help reduce manual work, improve decision-making, and enhance operational agility in a dynamic shipping landscape.

• Fulfillment-focused agents streamline warehouse tasks, handling order exceptions. These tools tend to be more operationally intense and usually need to integrate deeply with their WMS. If the agent is fed the right data, they can really simplify logistics, reduce errors, and make fulfillment workflows much more efficient.

• Manufacturing agents are designed to optimize production planning, inventory management, and quality control. These agents help manufacturers predict demand fluctuations, automate production scheduling, and maintain consistent quality assurance.

• Procurement agents are automating the sourcing process, helping businesses identify and negotiate with suppliers, manage inventory levels, and track purchasing trends. These tools can reduce manual effort in sourcing decisions, streamline supplier communication, and ensure better alignment between procurement needs and inventory requirements.

Many AI vendors will claim they can handle any workflow as long as they’re given the right guidelines. And while that’s not necessarily untrue, the real question is how efficiently they can do it. A general-purpose solution might eventually learn a freight claim process, but it will almost always lag behind an AI tool purpose-built for that exact workflow. Similar to the rise and success of SaaS point solutions, AI point solutions have industry specific context and domain specific logic that can drive a quicker ROI. The onboarding process doesn’t need weeks of prompt tuning, instead it just works, straight out of the box.

As mentioned in the previous section, the best approach when selecting an AI solution is to start with a point solution that solves a real pain point out of the gate and then expand capabilities across similar workflows to build trust. You don’t need an AI that might do everything eventually, you need one that does something extremely well today.

About the Author and BackOps AI

A big thank you for the folks who made it this far into the article. I hope I was able to break down the AI space in a non intimidating and digestible manner.

For those of you who do not know me, my name is Sean Elmurib and I am the Chief of Staff for an AI logistics startup called BackOps AI. I grew up in the Bay Area, played water polo for Loyola Marymount University. As much as I loved LA, I moved back to the Bay to do early stage investing for Plug and Play Ventures where I focused on investing in the supply chain and logistics industry. I was lucky enough to participate in roughly 25 deals spanning raw material procurement, manufacturing, warehousing, all way down to the last mile.

Across my five years doing early stage investments, I met with thousands of founders. There were a select few that had so much conviction in what they were doing that if given the opportunity, I would drop everything to work for them. That was the feeling I had 6 months ago when I met Sean McCarthy and Henry Ou, the Co-Founders at BackOps AI.

Sean and Henry saw firsthand how inefficient warehouse operations were and decided to dive in full time to tackle these problems head on. In doing so, we are building out autonomous agents and copilots to help automate manual repetitive tasks that 3PLs face like claims and reshipments. At a high level, we are able to ingest any customer inquiry or issue through any mode of communication (Phone, messaging, email) and our AI agent is able to fully autonomously resolute those customer inbounds.

If you’re interested to learn more about our capabilities then please drop me a DM or email me at sean.elmurib@backops.ai. We’d love to show you how we can help automate your operations.