Does speeding up work with AI just move the mistake further down the line?

Chris Condon runs an AI quoting platform for air freight. He says speed without judgment only moves the error, and the fix is checking what you commit to first.

It does, if speed is the only thing you automate. Chris Condon, founder of Aircon, says the costly errors in freight are the details nobody caught before the quote went out: the hazmat, the missing dimensions, the shipper who is closed on Fridays. His answer is to let the system catch those before you commit, and hand the judgment calls to a person with the context in front of them.

Chris started in 1995 driving a forklift in a container yard in Kent, Washington. As I said when I introduced him, he went on to spend sixteen years at Expeditors, turn around a freight company and build and sell another one. Today Aircon checks routes, pricing, feasibility and margin for air freight forwarders before a quote is sent. I wanted him on the show because most people in logistics will tell you the problem is speed, and Chris thinks that's the wrong diagnosis.

Why does speed matter so much in the first place?

Because in air freight, the first answer usually wins. When Chris took over the export department in El Paso, the first thing he built was a quote desk. He told the woman running it, Angela, to answer every request within thirty minutes. According to Chris, it's well understood in the industry that the first forwarder to respond has a 40 to 60% better chance of winning the business, regardless of the rate, and he says his customers have documented it for years.

So the pressure to move faster is real. Every quote desk has a dashboard counting quotes per hour. The trouble, in Chris's view, is what that pressure does to the person reading the request.

Where does the mistake go when you move faster?

It goes to the end of the file, where it's too late to fix. Chris told me about a panel he sat on in Morocco a couple of months ago. One forwarder said he had to read through twenty emails to figure out the dimensions of a shipment. Now put that person under a quotes-per-hour target. Things get missed. Was it hazmat? Was it batteries? Did anyone notice the shipper never opens on a Friday, and the pickup is booked for Friday?

None of that shows up when the quote goes out. It shows up when the file closes and lands on the P&L. By then the forwarder has already committed, and Chris says most of them just eat the cost rather than go back to a customer who told them the details up front.

Our job is to know what you're committing to at the time of quote before you're exposed.

Chris Condon, 30:34

That line stuck with me. A fast wrong quote is worse than a slow right one, because the fast one looks like a win until the invoice arrives.

Why do experienced people miss the routine details?

Because routine stops people thinking. Chris was careful to say this isn't a character flaw. It's human nature. You write the procedure once, learn it, and then you work from memory.

We write all these beautiful SOPs and standard operating procedures, and this customer needs this. But once you start doing it, you never refer back to it, right?

Chris Condon, 28:57

I've watched this in every business I've worked with. The procedure lives in a binder and the real process lives in someone's head, and the two drift apart a little every month. Chris's point is that a machine reads the checklist every single time. In his words, the agents don't miss hazmat, and they don't miss a request that came in without dimensions. If you're writing procedures for an agent to follow, the same rule applies: it will only check what you wrote down.

Which decisions should stay with a person?

The ones that need judgment. Chris draws the line early in our conversation, and he comes back to it again and again.

Where is human judgment required? Where is human input required? And if it's routine, let the routine be routine and let the system handle that.

Chris Condon, 04:46

His example is Dallas to London. Plenty of flights, stable rates, lots of options. Quote it automatically off the rate sources you already trust. Now make it hazmat to London. The system can still find the hazmat rate, but someone has to ask whether you hold the license to move it. Oversized cargo, or a shipment headed to an embargoed country, goes the same way. Chris wants those in front of a person, with the whole request laid out, so the operator's time goes to the calls that need an operator.

That's the split I'd use in any business. Routine work with clear rules goes to the agent. Anything where a wrong answer creates liability, or where the rules run out, goes to a human who can see why it was flagged.

What happens when automation skips the human entirely?

Chris has a story about this, and it's the part of the episode I'd want every founder to hear. About six to eight months ago, someone pitched him on automating his inbound leads. He takes almost none of those calls, but he took this one. The vendor pulled up a dashboard and announced they had already reached out to a thousand of his customers.

You reached out on my behalf without my authority, without my approval, and without my voice.

Chris Condon, 11:45

He knew it was true because one of the names on that dashboard was one of his customers. He called it an ethical problem and ended the conversation. What struck me is that the vendor thought the thousand messages were the proof it worked. Speed was the whole pitch. Nobody had asked whose name was going out, or whether the owner had agreed to it.

How do you find the mistakes you can't see?

You look at what never made it into the system. Chris made a point about dashboards that I keep thinking about: a dashboard only shows what's in the system. It says nothing about what's missing. By his account, most of Aircon's customers are missing 30 to 40% of their inputs. One customer told him they handle fourteen thousand quotes a month, and the CEO believes he's missing thirty to forty percent of them too.

Chris also turns the question around. If a customer asked for a hundred quotes this month and shipped nothing, why? Was it the rate, the routing, the transit time, the pickup? A good forwarder answers that by instinct. With the data, an owner can have the direct conversation: are you just benchmarking me, or am I thirty percent too high on this routing, and would a different airline win it back?

I don't need visibility. I need to know what happens next.

Chris Condon, 19:44

If you're deciding where AI should go first in your business, Chris's test is a good one. Don't start with the step that's slow. Start with the step where a missed detail costs you money later, and make sure a person sees the exceptions before you're committed.

Cut from AI Is Making Logistics Faster. But Are We Solving the Right Problem?.