By Sagar Shankaran, Founder of CallSphere
AI drafts the air waybill, so what does a new export agent actually learn? A forwarder's week-one curriculum, the ramp arithmetic, and what stays certificated.
Key takeaways
Seven-ten on a Monday in September. The export desk at a 22-person forwarder ten minutes from O'Hare is already an hour behind. A shipper in Elk Grove Village sent a letter of instruction listing three skids; four came off the truck. The 1400 lift to Frankfurt has a 1030 cutoff, and the agent who used to hold all of this in her head left in June for a 3PL paying four dollars an hour more. The person in her chair started nine days ago and is counting pieces with a flashlight.
Every forwarder I have walked through this year has a version of this scene. It is not a technology problem. It is the ramp — the six or eight weeks between a new export agent's first day and the day you stop re-checking her air waybills before they go to the carrier. In 2026 that ramp is what changed.
Air export is a job of small, unforgiving accuracies. Chargeable weight against the volumetric divisor. The right consignee on the house air waybill so the master matches when the carrier builds the manifest. The Internal Transaction Number back from AES before the shipment can leave. The screening record that proves your Certified Cargo Screening Facility handled the freight in an unbroken chain of custody. A commercial invoice that agrees with the packing list, because if it does not, the freight sits in the carrier's terminal while your agent stands at a counter at 11:50.
The traditional way you got someone through this was time. You put the new agent next to a senior one, gave her the easy accounts, and let her make two hundred small mistakes under supervision. Six weeks was normal, eight in peak — and peak is exactly when you are hiring, because the September-through-Thanksgiving run is when the volume shows up and the tenured people burn out.
Ask an owner what a new export agent costs and you will hear the salary. The real number includes the senior agent who loses a third of her day for six weeks, the carrier amendments when a house air waybill goes out with the wrong consignee, and the truck that came back because the freight was tendered without the security status.
The workaround everyone pretends is fine: you hire people who already did this job somewhere else. In most US gateways there is no pool left, and the ones who exist are being bid up by the integrators. So you hire from warehousing, from a customs broker's mailroom, from a rental car counter, and you eat the ramp.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for logistics in your browser — 60 seconds, no signup.
Two numbers tell the story. US small-business AI adoption reached 66% this year, up from 55% a year earlier, and roughly 70% of owners say their people need more training to use it well. The bottleneck moved. It is no longer whether you can get a capable assistant into the office — Claude Cowork landed in January and reached mobile and web in July, ChatGPT Work shipped on 9 July, and both take a plain-English goal and hand back finished work. The bottleneck is whether the twenty-four-year-old at the export desk knows what to do when the draft is confidently wrong.
Here is the cleanest way I can put it: in an air export office in 2026, the entry-level job is no longer data entry with judgment attached — it is judgment with the data entry already done. That is a different hire, a different week one, and a different set of things you can stop testing for.
flowchart TD
A["Shipper emails SLI and commercial invoice"] --> B["Assistant drafts HAWB in the forwarding system"]
B --> C{"Any DG line on this shipment?"}
C -->|Yes| D["Stop. DG-certified agent checks the Shipper's Declaration"]
C -->|No| E["New agent verifies chargeable weight and pieces on the dock"]
D --> F["New agent confirms AES filing and screening record"]
E --> F
F --> G["Supervisor spot-checks one in five before tender"]
G --> H["Freight tendered to carrier before the 1030 cutoff"]
The old week one was software training: where the house air waybill screen is in CargoWise or Magaya, how to copy a shipment, which fields the carrier rejects. Most of that is now the assistant's job — it reads the letter of instruction and the commercial invoice, drafts the house air waybill, calculates chargeable weight against the volumetric divisor, and flags where the packing list and the invoice disagree.
So day two is about the things a draft cannot settle. Teach the new agent to walk to the dock and count pieces against the paperwork, because the assistant read what the shipper claimed, not what came off the truck. Teach her the escalation rule: if any line on the invoice smells like a battery, an aerosol, a paint, a magnet or a perfume, she does not touch it — a dangerous goods certified agent does. Teach her what a broken chain of custody looks like on your screening records, and why re-screening at the carrier's terminal costs money and a slot.
And teach her to distrust the draft in one specific way: it will happily fill a field it should have left blank. A commodity description reading "machine parts" when the invoice says "lithium ion battery pack, installed in equipment" is the failure mode that ends careers.
Three things I would no longer screen for in an air export agent. First, typing speed and ten-key accuracy — the drafting is done. Second, memorised rate structures and surcharge mechanics; the assistant pulls the applicable rate, the fuel and security surcharges and the minimum charge and shows its working, and your pricing desk still owns the spot rates. Third, prior experience in the exact software you run, because the person is checking output rather than producing it.
What I would screen harder for: the willingness to physically go look. Comfort saying "I stopped this one." And a clean background, because that has not moved an inch — the Security Threat Assessment under your Indirect Air Carrier Standard Security Program, the fingerprint-based criminal history check and the airport badge with the ten-year employment history still gate the job, and no assistant shortens that clock.
Illustrative, for a forwarder hiring one agent in August. Substitute your own wage rates; the shape holds.
Still reading? Stop comparing — try CallSphere live.
See the logistics AI agent handle a real call — complete, industry-specific, and live in your browser. No signup.
| Line | Old way | 2026 way |
|---|---|---|
| Senior agent coaching hours across the ramp | 48 hrs at $38 | 22 hrs at $38 |
| Weeks before unsupervised tendering | 6 | 3 |
| Unproductive hours during ramp (at $25) | 120 hrs | 60 hrs |
| Carrier amendments and re-tenders, first 90 days | 34 | 11 |
| Added dangerous goods and security classroom time | none | 12 hrs plus $650 course |
Coaching saves 26 hours at $38 = $988. Unproductive time saves 60 hours at $25 = $1,500. Twenty-three fewer amendments at roughly 25 minutes each to chase and correct, at $25 an hour, is about $240. Against that, put $300 of extra classroom hours and a $650 dangerous goods course. Net first-90-days: about $1,778 per hire. Bring four seasonal agents on in August and you are near $7,100 — and more usefully, you get them tendering unsupervised before the peak instead of during it.
Be blunt with yourself here. Dangerous goods acceptance is a certificated human act. The DOT recurrent training clock in 49 CFR 172.704 runs three years; the IATA competency clock your carriers audit you against runs 24 months. An assistant can read a Shipper's Declaration and tell your agent the packing instruction does not match the UN number — but the person who signs the acceptance check sheet has to be trained, current and awake. Do not let a good draft create the impression that the check happened.
The same goes for the badge. Nobody works your dock until the Security Threat Assessment and the criminal history check clear, and that queue does not care how quickly your new hire learned the software. Plan your August hiring around the badge clock, not the training clock.
The Monday version of this: take the last ten house air waybills your newest agent produced, have an assistant re-draft them from the original letters of instruction, and put the two side by side. Where they agree is training time you can stop spending. Where they disagree is your week-one curriculum, written for you.
Less than you did. The parts that took experience to do fast — building the house air waybill, getting chargeable weight right, spotting an invoice-to-packing-list mismatch — are now drafted. The parts that take experience to do safely, like dangerous goods acceptance and knowing which carrier terminal will actually take a late tender, still take experience. Hire one of those people per shift, not five.
Probably with the same number handling more shipments, not with fewer bodies. The dock work, the piece counts and the phone calls do not shrink. What shrinks is the amount of a senior agent's day spent re-checking a junior's paperwork.
That is the 70% number in the flesh. The answer is not a class. Pick one recurring task — drafting house air waybills for your three highest-volume shippers — and have exactly one agent do it that way for two weeks with the supervisor checking every one. When the checks come back clean, the rest of the desk will ask for it themselves.
Your Indirect Air Carrier Standard Security Program governs who handles freight and information, not who drafts a document. Keep screening records and chain-of-custody entries made by a badged, assessed person, and keep the assistant's drafts clearly marked as drafts until a human accepts them.
One last thing that tends to get missed in all of this. While your new agent is learning to count pieces on the dock, the phone at the export desk is ringing with shippers asking whether their freight made the lift. CallSphere builds AI voice and chat agents that answer the business line and the web chat around the clock, take the details, and capture the ones that need a callback — so a nine-day-old hire is not choosing between the counter and the phone at 11:50 on a cutoff morning.

Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
Which jobs on an independent used-car lot change shape when AI drafts the work, what a new title clerk learns in week one, and what leaves the job posting.
A repair station types the tail number six times before invoicing. Here is how 2026's connector between AI and maintenance software cuts eleven days to two.
A new client service rep used to need six weeks on Cornerstone. Here is what to teach in week one now, what stops being a hiring requirement, and the cost math.
Charter fraud meets AI that can act. What a Part 135 operator must keep human, how to scope Avinode and inbox access, and the arithmetic of the approval click.
Concealed-damage claims cost a warehouse manager half a Friday. On-site AI searches your own footage and scan trail without any data leaving the property.
Vendor acknowledgments quietly restate price, date, quantity and freight terms. Why checking all 1,200 a month became economic in 2026, with worked arithmetic.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI