The Software Paradox: Why More Technology Isn’t Creating Less Work

Stand behind your best carrier rep for 10 minutes tomorrow and count the tabs. Don’t help. Just count. 

Then pay attention to what happens between them. I’ll bet the rep pulls a load from one system, hunts for capacity in another, checks a carrier somewhere else, opens email, grabs the phone, and eventually returns to the TMS to document the whole adventure.

I did this on my own floor once, and it ruined my week. We had spent years buying technology, yet every product still depended on the same integration: the person at the desk. Robert Solow noticed the same software paradox in 1987 when he saw computers everywhere except in the productivity statistics. He would’ve felt right at home in freight.

I run Envoy, where we build Ellie, so I’ll spare you the fake neutrality. I’m selling something. I also bought nearly every category of freight software first, then paid people to carry the work from screen to screen.

I judge technology by loads per rep, operating cost per load, and whether the brokerage can take on volume without opening 20 new seats. Once those figures move, the operating model has changed. 

Until then, the old work is still running the company through newer screens.

The Stack Knows Plenty. Your Rep Still Has to Move the Load.

A TMS gives the load, customer requirements, and history a reliable home. Load boards put capacity within reach. Compliance platforms bring authority, insurance, safety, and fraud signals into the decision. Ellie uses the record and intelligence these systems provide.

Now give the stack a live request: “I need a reefer out of McAllen by 2, and I can’t pay over $2,300.”

The rep takes over from there, finding the right carriers, reaching the dispatcher who’s ignored 37 emails, working the rate, confirming the equipment, and following up until the truck is covered.

The execution layer carries out that work. Ellie uses the tools already on the floor to source, communicate, negotiate, and verify within your rules. The rep’s capacity no longer lives entirely inside one person’s ability to work the tabs.

The Integration Layer Is a Person You Pay

Follow one ordinary load across the carrier desk. The rep copies the details from the TMS, checks a load board, pastes the MC into a vetting portal, sends the email, calls the dispatcher, negotiates the rate, confirms the truck, and circles back to update the TMS.

We’ve given that process respectable names, but the integration layer is still a person with a salary, commission plan, health insurance, and 401(k). Their brain holds the context while their hands move the information. Human middleware also gets tired and occasionally takes a vacation during produce season.

The portals will remain part of freight, and a beautiful API won’t persuade a carrier to come down $200. The calls, counteroffers, confirmations, and exceptions keep landing back on the rep.

Meanwhile, the dispatcher who would’ve taken your Joliet load for $1,850 gives it to the next broker who called. Your systems worked exactly as designed, which is cold comfort when the load is still sitting there.

A Beautiful Draft Still Leaves the Load Uncovered

I’ve watched plenty of AI copilot demos where the machine produces a lovely carrier email, summarizes the response, and waits for the rep to finish the job. I’ll take the saved minutes. I lose interest when the vendor presents them as a new operating model.

The easiest way to separate the products is to follow the work: The rep still sets the direction, controls the money, approves where required, and takes the edge cases. They can own the outcome without personally typing every email and chasing every reply.

Your Hiring Plan Knows Whether the Technology Worked

Put 2019 loads per carrier rep beside last month’s figure, then add operating cost per completed load. That review will tell you what another hour of slides about AI strategy probably won’t.

If the company wins 20% more freight and immediately needs roughly 20% more people, labor is still attached to volume at the same old ratio. The TMS may be cleaner, the compliance decision may be safer, and the rep may work faster, but the brokerage hasn’t gained much operating leverage.

Count the software bill and the people pushing work through it. Watch service, margin, and exceptions too, since moving a task elsewhere only improves productivity on the department’s PowerPoint.

I’ve said before that adding applications isn’t scaling. The next renewal should explain which operating figure moved. “The team likes the interface” may be true, but I wouldn’t take it to the board as evidence that the business can scale.

Robinson Changed the Ratio

C.H. Robinson makes this conversation uncomfortable because it has results. Its Q2 2026 report puts productivity gains above 60% since the end of 2022 in North American Surface Transportation and Global Forwarding. NAST volume grew 1.5% year over year while average headcount fell 11.6%.

At Citi’s 2026 conference, its leadership described the strategy in language any CEO can understand: decouple headcount growth from volume growth. Robinson paired its agents with a lean operating model, redesigned the work around them, and changed the economics of the floor.

Robinson has hundreds of engineers, years of investment, and enough scale to make that undertaking rational, though. A $200M brokerage is a different story. You could spend two years trying to copy that operation, only to discover you accidentally started a very expensive software company.

That doesn’t do you any favors. Robinson will still show up at the same RFP with a more productive operation, and your shipper won’t hand out points for building the technology yourself. That’s where we come in. We built Ellie so you can hire the execution layer and put it to work across the systems you already own, and compete with the big dogs.

What Ellie Is Taking Off the Floor Today

Roughly 75% of freight in our live accounts is booked through Ellie. Reps use her on real freight, where she handles work such as:

• Finding and Reaching Carriers: Ellie reads the load, ranks carriers using the brokerage’s data, and works email, text, and voice in parallel. The rep can review the response instead of personally introducing the same load to every dispatcher with a pulse.

• Working the Rate: Ellie handles the back-and-forth inside the customer’s buy caps and approval rules. Decisions outside her authority reach the rep with the carrier, rate, equipment, and history assembled.

• Checking the Carrier Who Called: Through our Highway integration, Ellie captures the caller’s DOT number and checks it against Highway’s compliance data and the brokerage’s rules. Highway supplies the intelligence, and Ellie puts that intelligence to work inside the transaction.

The experienced person makes the calls that deserve experience and gets to them before 4:30 because Ellie has handled the routine execution.

Quit Shopping for Software and Hire the Work Done

Solow’s paradox faded when businesses redesigned work around computers instead of treating them like expensive typewriters. Freight has reached that point. The execution layer changes how the load gets covered and how much volume the floor can absorb.

Digital Operations defines what Ellie owns, how much authority she has, and who receives an exception. TOAS shows her work as it happens, and each booked load adds to the Carrier Context Graph the brokerage owns.

Give Ellie the work, keep your people on the judgment, and measure the result in loads per rep, operating cost per load, service, and margin. If those figures move, you can grow without asking the COO to celebrate a new customer by calling a recruiter.

Book 30 minutes. Bring the load that rolled last week and the portal it failed in, and we’ll cover it live. Or email me, robby@tryenvoy.ai, with your loads per rep for 2019 and last month. If those two numbers match, I already know what your software’s been doing. So do you.