TOAS: Defining the Transportation Observability Action System

I ran a top 50 Freight Brokerage. I always believed that the real work lived in more places than the TMS but we  never once asked how the work would get executed off the data sets in the future.  It is starting to become clear. Very. If you’d asked most , they would have said the TMS, same as you. It took me too long, and too much money, and multiple freight cycles to see with clarity. 

Sit next to one of your reps for an afternoon. Not a bad one. Your best one, the one who knows which carriers ghost and which ones come through when a lane goes tight. 

Watch her cover a reefer load that has to roll Monday. She’s bouncing from the TMS to a load board to a carrier portal to her inbox to her phone, then back to the TMS because the rate moved. 

Nobody built her day on purpose. It piled up one tool at a time until my sharpest salesperson spent her afternoons as the glue between six systems that don’t talk, covering at 6 p.m. a load that should’ve gone out by noon.

What took me too long to see: nobody ever built her an operating system, so she became one, spending her day as the coordination layer freight never got around to building.

So at Envoy, we built the real one. A brokerage can finally run on an operating system instead of six tools and a tired rep holding them together. TOAS, the Transportation Observability Action System, is the backbone that powers it, and our execution layer, Ellie, is the first thing running on it.

Freight Never Had a Real Operating System

Think about the laptop you’re reading this on. A drive, a screen, a hundred programs, and underneath all of it an operating system that makes the pieces behave like one machine. You never think about it. That’s the whole point. It runs the machine so you can do your work.

Freight has the pieces. The TMS, the boards, the portals, the inbox, that’s the hardware. Sourcing, negotiating, covering, that’s the work. What we never had was the operating system in the middle that makes the pieces move as one. So we handed that job to people. Your reps became it, and they’ve paid for it in clicks and late nights ever since.

I’m not judging anyone who built a floor this way. It’s how the whole business ran, and nobody built the real operating system because it took an unlikely pair in one building: people who’ve sat on a carrier desk their whole careers, and engineers who build agents for a living. Put them together, and you get a backbone that can run the machine instead of a person doing it by hand. 

We call that backbone TOAS.

What TOAS Is, and Why It’s the Backbone

Let me lay out how it fits together, because this category loves fog. 

The operating system is the new way your brokerage runs, one coordinated system instead of six that ignore each other. 

TOAS is the backbone that powers it, the Transportation Observability Action System, the layer that sees everything moving across your tools and acts on it, live. Ellie’s the first application running on that backbone, the part your reps work with day to day.

On your laptop, you never see the operating system doing its job. You see the apps. Same here. Your reps see Ellie. Underneath her, TOAS is doing the sensing and the running, on top of the TMS and the boards and the portals, below your people, under the hood.

The TMS Was Storage, Not an Operating System

A TMS is a system of record, and I don’t say that as an insult. Recording is the job. It writes down the rate, the carrier, the pickup, the POD, so there’s a clean file when a shipper or a lawyer comes asking. It’s the disk. A disk earns its keep. I just never covered a single load by staring at a database of loads I’d already covered.

The bill comes due during the workday. The record’s supposed to hold it all, and my reps still lost close to a third of their day hunting for information the system was supposed to have. Our own floor burned nearly 40% of a rep’s time switching between screens. 

I was paying the most expensive people in the building to dig rates and pickup windows out of a filing cabinet instead of selling freight.

It was never going to fix itself, because the work never lived in one place to begin with. It’s smeared across a dozen tools, and the TMS only ever caught a thin slice of it, always late. 

An operating system needs two things a filing cabinet will never have: it has to see the whole floor as it happens, and it has to act on what it sees. That’s the backbone’s whole job, and it’s the two words hiding inside the name TOAS.

Observability Is the Operating System’s Senses

Start with observability, because everybody hears the word and pictures a dashboard, when it’s really the opposite of one. 

Engineers have run observability as a real discipline for a decade, and I’ll take their definition as is: figuring out what’s really going on inside a system from the signals it’s already giving off, not from the few numbers you happened to put on a chart.

A dashboard shows you a metric you already decided to watch. Observability tells you what’s happening and why, before anyone builds a report for it. The TMS is the dashboard. It shows the fields it was set up to show, but can’t tell you a carrier just went dark on a lane you’re exposed on, or that a rep’s one click from firing off a rate that went stale an hour ago.

Now stand on a carrier floor. Every rate check, portal login, email, text, and check call is a signal about what’s really happening. Read them together the way a good ops manager reads a room, and you’ve got the real state of your floor, not the tidy version the TMS logged after the fact.

That reading is what gives the operating system its eyes, and TOAS turns it into memory the whole desk can use.

Action Is What Makes It an Operating System, Not a Dashboard

The word many vendors tiptoe around is action. Seeing is only half a system. Plenty of tools will paint you a gorgeous picture of a problem, then hand it back for you to fix by hand. A real operating system doesn’t just report the state of the machine. It runs it. It does the work.

The way it works, TOAS is the backbone and Ellie’s the tool running on it. She reads the load, sources carriers across your systems, runs outreach over email, text, and voice at once, negotiates inside the max pay you set, checks MC/DOT before anything books, then hands your rep a decision to sign off on. 

Read that back and notice who didn’t lift a finger. Nobody logged into a portal. Nobody retyped a rate three times. The rep owns the call. Ellie did the grunt work under it.

I can’t believe I used to pay grown adults to send those emails. I mean that. The rest of the business world’s already going this way too, with task-specific agents jumping from under 5% of enterprise software to something like 40% by the end of 2026. Freight’s just late, which is the only reason the door’s still open. 

If your software doesn’t change what happens on the floor, it isn’t real yet.

Where TOAS Runs: On Top of Your Stack, Under Your People

Your engineers are already wondering what this costs them to run. Good. That’s my favorite part. You don’t install this operating system by ripping out the one you’ve got. 

TOAS runs on top of whatever stack you’re already on, and under the person who owns the outcome, as a Chrome extension. No migration. No integration project. No 12-week onboarding before anyone gets value.

Ellie works the same portals your reps do, on the same pages, without waiting on an API somebody might never build. Engineering teams tend to relax when they hear that, because a new platform is one more thing to secure and staff and babysit, and this is none of that. It rides on the tools that already cleared your review.

Most of this category wants you to move your whole operation onto their platform first and find the value later, if it shows up. I’m asking for nothing but the tools you already run. It’s the difference between a rep going live the afternoon she meets Ellie, and a six-month rollout that dies the first time it meets your real workflows.

A Rolled Load Is an Operating-System Failure

Go back to that rep, 6 p.m., half her board still open. 

When one of those loads finally rolls, it’s almost never because she didn’t know the freight was sitting there. She knew. What she didn’t have was anything watching the whole floor closely enough to catch the coverage slipping in time to save it. 

Call it coordination. That’s the entire job of an operating system, and she was doing it by hand.

The market’s making it bite harder too. Route guide depth on long hauls hit 1.71 early in 2026, up from 1.28 a year earlier, which is a polite way of saying shippers are digging deep into their backup carriers before a truck ever gets covered. Every step down that list is time, and time, not lane knowledge, is where loads fail now.

A rep with six tabs open might spot that load slipping at 4 in the afternoon. Ellie’s running on TOAS, which is watching the whole board at once, so she’s on it by 10 in the morning, while there’s still a truck to book. Same rep. Same lanes. The load moves.

The 2026 Market Is Forcing the Upgrade

This stops being optional this year, and you can feel it coming. Last time volume came roaring back, we all reached for the same lever and hired. Then the market rolled over, and the shops that staffed up spent the next year handing out severance. 

I don’t know one owner who wants to sit through that a second time, and every one of them can feel the sequel loading.

Especially with the setup here. 

Spot pushed above contract for the first time since 2021, tender rejections are running at multiyear highs, and the people who watch this for a living are calling these levels a new floor, not a spike that fades. Volume’s coming for your desk again, sooner than you’d like.

The shops that come out of this ahead won’t be the ones that hired the fastest. They’ll be the ones who quit tying volume to headcount at all. A brokerage running on this new operating system swallows the surge without five new reps on payroll, because TOAS is doing the work those hires would have done. That window’s open right now, and it slams shut the day a competitor in your lane installs their operating system before you install yours.

The Operating System That Learns: The Carrier Context Graph

This is the part I’d bet the company on. A normal operating system forgets everything the second you close it. This one remembers, and what it remembers is yours.

Every rate Ellie negotiates, every load she covers, every carrier she works with turns into structured memory inside a Carrier Context Graph. That’s the long memory of your floor, and TOAS is where it lives.

It’s also why Ellie negotiates smarter the longer she runs, leaning on what’s booked on your lanes instead of guessing. The carrier knowledge that used to sit in one 20-year veteran’s head, and walk out the door the week he retired, ends up somewhere your whole desk can reach.

More importantly, it’s the quiet reason the second AI tool a brokerage buys rarely measures up to the first. With none of your context, it’s starting from scratch, relearning years of how your floor covers freight. Shared state is what beats siloed automation, and the backbone is where yours lives and stays yours.

What TOAS Is Not, and How to Spot a Fake

Let me be straight about what TOAS isn’t, because the honest version has always outsold the hype for me. It isn’t full autonomy, it isn’t a chatbot in a freight costume, and it isn’t one more tab your reps have to babysit. The rep still reviews, still approves, still owns the rate and the carrier and the relationship. 

Autonomy only grows as fast as trust does, and that’s the only honest way to build it.

I say the “isn’t” list out loud because this category has a lying problem. Gartner figures more than 40% of agentic AI projects will be killed by the end of 2027, and there’s a name for the con now: agent washing, old automation with a fresh coat of paint sold as an agent.

So when a vendor sells you a backbone, make it prove the action half. Can it act, or does it only watch? Does its memory survive the session, or start over every morning? Will it run on your tools, or does it need a migration first? The one that ends most of my demos: what percent of your freight would it book today? 

For context, ours books about 75% in live accounts, on SOC 2 Type II. That’s a product. The rest is a pitch.

The Operating System Is Here, and Ellie Runs on It

I called the TMS my system for 20 years. But it never really was. It was the filing cabinet, and my reps were the operating system, holding six windows together by hand until 6 at night. That job was always temporary. It took someone building the real one to prove it.

Now a brokerage can run on a real operating system, and TOAS is the backbone that powers it, the senses to see the whole floor and the reach to run it as one instead of six. Ellie’s the first thing running on it, and everything she does feeds a Carrier Context Graph that’s yours to keep, and nobody else can rebuild. Your best rep stops being the glue and goes back to closing.

With spot above contract and capacity still leaking out of the market, the shops that put this operating system under their floor this year scale into the upturn without scaling payroll. I’m not hoping that’s true. In our live accounts, about 75% of freight already books through Ellie.

The fastest way to understand any of this is to watch it happen. Book a demo, and I’ll put Ellie on a few of your real loads while you watch her in action.