The Coordination Tax: The $1.6 Trillion Cost of Running Supply Chain on Human Hands

A Memo to Investors, Customers, and Anyone Who Wants to Understand What We're Building

Before we built Logistify AI, my co-founders and I ran a third-party logistics business. At our peak, we were operating ten warehouses across 70,000 square feet, managing over $20 million in inventory for some of the largest B2B e-commerce companies in our market. Every day, we received thousands of orders — by email, by WhatsApp, by phone, by paper. We hired 50 college interns whose sole job was to read those orders, figure out what they actually meant, and type them into our inventory management system by hand.

We were not a small or unsophisticated operation. We had technology. We had processes. We had experienced people who knew our customers' quirks, our suppliers' habits, and the institutional rules that had accumulated over years of doing business. And yet, every single morning, the first couple of hours of work was spent copying information from one place into another. Not because anyone chose to make it that way. Because that was the only way it could work.

Our accounts receivable was its own version of the same problem. We extended 30-day payment terms to our B2B e-commerce clients — standard practice, necessary to win the business. But following up on outstanding invoices meant emails, WhatsApp messages, and sometimes driving to a customer's office in person. There was no system that tracked who owed what, sent a reminder at the right moment, or flagged when an account was quietly becoming a risk. Collections was a manual pursuit, entirely dependent on whoever had bandwidth to chase. By the time we wound down the business, we had written off over $100,000 in bad debt — money that customers owed us and simply never paid. That is not a rounding error. That is the measurable cost of coordination failure in accounts receivable.

We had the same problem on the warehouse floor with driver coordination. Across all ten warehouses, we were loading and offloading roughly 700 vehicles per day — delivery trucks of every size, plus hundreds of motorcycle delivery riders. Every single one of them arrived with no prior appointment, first come, first served. Drivers would sign in on a paper sheet at reception, then wait outside beside their delivery trucks until someone came to call them in. Our receiving team had no visibility into what was coming or when. Labor planning was guesswork. At 700 vehicles a day, even a conservative 45 minutes of avoidable dwell time per vehicle amounts to over 500 person-hours of collective waiting — every single day, across our network. That is the equivalent of more than 60 eight-hour working days, lost daily, to a problem that had a straightforward fix: appointment scheduling. We just did not have the system to do it.

That experience — living inside the coordination gap across order entry, collections, and the warehouse floor — is what this company is built on. What follows is my honest account of what that gap looks like across the industry, why it has persisted for so long, and why we believe AI agents are finally the right tool to close it.

The Software Did Its Job. Coordination Is What Remained.

The great ERP rollouts of the last three decades solved a real and important problem: they gave companies a single source of truth for what had already happened. An invoice paid, a shipment received, a purchase order confirmed — all of it captured, stored, auditable. For financial reporting, compliance, and historical analysis, ERPs are genuinely transformative. We used them. Our customers used them. SAP, Oracle, NetSuite, Microsoft Dynamics, QuickBooks, Unleashed, Odoo, Zoho — every one of them is a remarkable engineering achievement. The systems of record did exactly what they were designed to do.

What they were not designed to do is coordinate. Coordination is something different. It is the act of taking a WhatsApp message that says "send me the usual 50 crates" and understanding that it means SKU-4471, not SKU-4472, for this particular customer who has a special pricing agreement expiring next week — and getting that into the system without a human in the middle. It is reading a supplier's confirmation email and catching the quiet update buried in paragraph three before it becomes a crisis. It is matching a lump-sum payment from a customer who routinely pays three invoices at once. It is noticing that the paper delivery note doesn't match what the driver actually offloaded.

No system of record was ever designed to do that work. So the work fell to people. And it has stayed with people — because until very recently, there was no technology that could take it from them.

"The Coordination Tax (n.): the cumulative cost — in labor, errors, delays, and write-offs — that a business pays every day because its systems of record cannot bridge the gap between each other on their own. It is not a line item on the P&L. It is the sum of every order typed twice, every invoice chased by hand, every shipment confirmed over WhatsApp, and every truck that waited an hour in a yard because nobody knew it was coming."

"A $26 trillion global supply chain, coordinated through Outlook, Excel, phone calls, and WhatsApp."— Accenture, Resiliency in the Making (2023)

The Coordination Tax, Witnessed Across Ten Warehouses

I want to be specific, because abstractions about "supply chain inefficiency" are easy to dismiss. Let me tell you what the coordination tax actually looks like — through three companies we have worked with directly.

The New Jersey Beverage Distributor

A beverage distributor in New Jersey processes 3,000 to 4,000 orders per year across email and EDI. Their system is QuickBooks Desktop — a tool they know well and trust deeply. The problem is the orders that arrive by email: PDFs formatted in each customer's own template, with product descriptions that do not match the distributor's catalogue names, delivery addresses sometimes embedded in footnotes, and pricing occasionally reflecting contracts that were updated six months ago and never communicated to the buyer. Each of those orders required a human to read it, interpret it, and translate it into QuickBooks. Not occasionally. Every single time. If you want to understand why this is so hard to fix, our definitive guide to sales order entry explains the anatomy of the problem in detail.

The London Craft Brewery

A craft brewery in London receives orders from its trade customers by email, WhatsApp, and voicemail. Their ERP is BREWW — a system purpose-built for breweries, well-designed and well-suited to their business. The coordination problem was in the gap between the channel and the system. Customers do not order using BREWW's catalogue names. They say "the usual pale ale" or "two more of the ones we had last month" or name a product in a way that matches two or three SKUs depending on format — cask, keg, or can. Every order required a human to read it, interpret it, match the description to the correct SKU in BREWW, and enter it manually. Not just for the ambiguous ones. For every single order that arrived outside of a structured channel, which was most of them. The risk of a mis-matched SKU going unnoticed — cask dispatched instead of keg, wrong volume, wrong account — was not theoretical. It was a daily operational exposure.

The Nairobi Bakery

A bakery in Nairobi was receiving 1,500 orders per day via WhatsApp. Fifteen hundred. Every day. A team of data entry clerks would wake up before dawn to start working through the messages — reading them, interpreting what each customer meant, cross-referencing against the product catalogue, and typing each order into the system before the first delivery runs left the building. On a good day, they caught up. On a bad day — a public holiday, a sick team member, a customer who sent an ambiguous message at 5 a.m. — they did not. Wrong quantities shipped. Customers waiting. Drivers dispatched with incomplete manifests. (We wrote about how AI handles this channel specifically in our guide to AI sales order entry for Brightpearl.)

This was not a failure of the people involved. They were working as hard as any team I have seen. It was a coordination failure. There was no system whose job was to coordinate between the channel where customers placed orders and the system where those orders had to live. Every coordination act was a human act. And human coordination, at 1,500 orders per day, is inherently fallible.

These are not outliers. These are representative. The pattern we saw across every warehouse we operated, every customer we served, and every supply chain operation we observed was the same: excellent systems of record at the perimeter, and a coordination tax in the middle — paid by people, every single day.

The Anatomy of a Taxed Workflow

Across manufacturers and distributors, there is a class of workflow that consistently generates the most cost, the most errors, and the most organizational pain. These workflows are not random or isolated — they share a precise anatomy. Recognizing it is essential to understanding why they are so resistant to conventional automation.

What makes this anatomy so consequential is not that any one marker is unusual. It is that all seven appear together, in the same workflows, repeatedly. When a workflow checks all seven boxes, conventional software cannot automate it — because conventional software was built to work inside systems, not between them, and to follow explicit rules, not interpret tacit knowledge.

The Full Arc of Exposure

These taxed workflows do not cluster at one end of the supply chain. They run across the entire order-to-cash cycle. At every handoff — from the moment a demand signal arrives to the moment cash is collected — there is a manual coordination step where a knowledge worker steps in to interpret, transcribe, reconcile, or escalate.

Every stage in that arc is a place where information that belongs in a system is currently living in a person. In the New Jersey beverage distributor, it was order capture and invoicing. In the London craft brewery, it was order capture and SKU matching. In the Nairobi bakery, it was the demand signal stage. In every case, the person standing in the gap was not adding strategic value — they were performing a translation. Converting information from the format a human sent it in to the format a system could accept.

The Coordination Tax Doesn't Show Up on the P&L

What makes this coordination tax so persistently under-addressed is that its true cost is nearly invisible to leadership. The labor cost of a five-person data entry team shows up as a line item. The cost of the coordination failures that team absorbs — return orders, stock discrepancies, incorrect invoices, delayed payments, disputed shipments — is distributed across departments, absorbed as "normal" operational friction, and rarely attributed to its actual source.

But the numbers, when surfaced, are clarifying. A 1% pricing discrepancy on a €6 million invoice batch is €60,000 — found in an inbox that no one was systematically monitoring. A data entry error rate of even 2% on 1,500 daily orders means 30 wrong orders every day, each one a potential return, a stockout, an unhappy customer, a chased reconciliation. A backorder notice buried in a supplier's follow-up email, unread for three days, is an engineer's deadline missed and a production schedule thrown.

The cost is not marginal. It is structural. And it compounds with every order, every shipment, every invoice, every day.

"Supply chains lose $1.6 trillion every year to disruptions that go uncaught — not because the signals weren't there, but because no system was watching for them."— Accenture, Resiliency in the Making (2023)

Why This Has Not Been Solved Before

The honest answer is that it could not be. Automating a workflow that requires genuine language understanding, contextual domain knowledge, and judgment across multiple unstructured inputs was technically out of reach until very recently. Rules-based automation — RPA, EDI integrations, workflow tools — could handle the parts of supply chain operations that were already structured and systematic. They could not handle the messy, language-rich, knowledge-dependent reality of how supply chains actually communicate.

That has changed. An AI agent (LLM + tools + context) that can read a WhatsApp message and understand it as an order, cross-reference it against customer history and product catalogue, and enter it accurately into an ERP — without a human in the loop — now exists and works in production environments. AI Agents that monitor supplier inboxes for buried backorder notices and surface them before they become crises. AI Agents that reconcile payment documents against contracts and flag discrepancies before the invoice is approved. AI Agents that encode the business rules that previously lived only in the heads of experienced workers, making them persistent, auditable, and scalable.

This is precisely what we have been building since we closed our 3PL business and turned the problem we had lived inside into a company. We were accepted into Microsoft's for-Startups program, the 1871 Chicago tech incubator, and Accenture's supply chain accelerator — not because we pitched a clever idea, but because the people running those programs had seen this problem at scale and recognized that the technology to address it had finally arrived. We presented our work at DLD Munich 2025 to an audience of investors and operators who confirmed, again, that this problem is universal and the solution is overdue.

"The goal is not to replace the knowledge worker. It is to finally give their knowledge somewhere to live — somewhere that doesn't retire, doesn't miss emails, and doesn't make transcription errors at 4pm on a Friday."

What We Are Building — and Why

Logistify AI builds the coordination layer that supply chain has never had. Our AI agents handle the work that currently falls between systems — reading the WhatsApp order, processing the paper GRN, matching the payment, surfacing the buried backorder notice — without requiring a human to perform the coordination act. Each AI agent automates a specific coordination task in the order-to-cash arc, from demand signal to cash application.

We are not building dashboards or analytics. We are not building another ERP. We are building the intelligence layer that finally makes supply chain coordination automatic — the layer that should have existed all along, and that AI has finally made possible to build.

"We are building the intelligence layer for supply chain coordination."

The coordination workflows that have resisted automation for decades are not inherently unautomatable. They were waiting for technology that could meet them where they actually live — in language, in context, in the space between systems. That technology is here. The question is which companies will move first to eliminate their coordination tax, and which will still be paying it when their competitors are not.

If you are running a manufacturing or distribution business, and any part of this memo sounds like a Tuesday morning in your building, I would like to talk. And if you are an investor who sees what we see — that coordination is the defining unsolved infrastructure problem of modern supply chain — I would like to talk to you too.

This is why we built Logistify AI. We didn't stumble into it as a market opportunity. It is the coordination problem we lived inside for years, and the mission we chose because we believed, and still believe, that supply chain deserves better than a $26 trillion industry coordinated through Outlook, Excel, phone calls, and WhatsApp. We are building the intelligence layer that finally makes supply chain coordination automatic.

— Daniel Emaasit, CEO, Logistify AI