At 12:01 a.m. Eastern on July 24, a 10% surcharge that touched nearly every import into the United States quietly expired. At the exact same instant, a new tariff took effect covering 60 trading partners and 99.4% of US imports. If you blinked, you missed the handover. Your customs broker did not.
That was the third time in six months that the ground rules under US freight moved. And this is the point where you would expect a logistics software company to explain the tariff to you.
We are not going to do that. We build software for freight operations. We are not customs lawyers, and we are not going to pretend to be. What we watch every day is something else: what these changes do to the people running operations desks. The providers that come through rule changes cleanly are not the ones that predicted them. They are the ones agile enough to absorb them.
So this article is about where that agility comes from, and about where we stand on the question behind most freight technology conversations right now: what should AI actually do in an operation, and what should it never touch?
The tariff part, kept short
The compressed version of the last six months. In February, the Supreme Court struck down the emergency-powers tariffs that had covered most imports since 2025. The replacement, a 10% surcharge under Section 122, carried a hard 150-day limit written into the statute. Courts fought over it all spring, but in the end the clock decided, not a ruling. One of the few deadlines in this story that held. It expired at midnight on July 24.
The successor was already waiting: new Section 301 duties covering 60 trading partners at 10% to 12.5% depending on the country. USMCA-qualifying goods and a handful of sectors stay exempt. And one detail most coverage skipped: the in-transit grace period covered ocean freight only. Truck, air, and rail got no window. Somewhere out there is a driver who cleared at 11:58 p.m. and saved a customer real money without ever knowing it.
That is all the tariff explaining we are going to do. The rest is about what churn like this does to the work.
What tariff churn actually costs a logistics desk
Every rule change arrives at the desk as busy work, and the busy work is compounding faster than the headlines.
The schedule under your quotes will not sit still. The US issued 32 revisions to the Harmonized Tariff Schedule in 2025 and 10 more by early June. That is roughly one edit every two weeks, for eighteen months, to the table your quotes depend on. A shipment coded correctly six weeks ago can carry the wrong duty today, even though the code still looks right.
Every low-value parcel now needs a customs entry. The $800 de minimis exemption is fully gone, and freight that used to move informally now needs declared values, HTS codes, and duty calculations, in volumes that never used to touch a customs desk. The customs desk has noticed.
Enforcement is looking at intermediaries, not just importers. A June executive order directs Customs to tighten the rules governing forwarders and brokers themselves. More documentation requests, slower clearances, even on lanes where no rate changed.
And the volume swings are policy-made. Importers front-ran the July 24 deadline exactly the way you would expect: ocean spot rates rose roughly 120% from mid-May to mid-July, then retreated before the deadline even arrived as bookings went soft. A ten-week boom and bust manufactured entirely by a policy date.
Two things we would actually bet on. The churn is permanent: three legal vehicles in six months means the question is never whether the rules will change, only which statute is next. And the paperwork keeps growing: every change above adds entries, checks, and records requests to each shipment, so the desk workload rises even while freight demand cools. The work compounds at the operations desk, not on the truck.
Here is what that looks like on the ground. Reclassifying lanes after every schedule edit. Re-quoting because the duty math moved mid-negotiation. Chasing origin paperwork one document at a time. Answering the same "what does this mean for my shipment" email forty times. None of this is why anyone got into logistics. All of it stands between a rule change and a steady operation.
Absorbing tariff churn is the skill, and busy work is what limits it
The past six months rewarded one capability above everything else: absorption, the amount of rule change an operation can take without breaking stride. Nobody in February could have told you how the courts would rule, which statute would come next, or which modes would get transition relief. Absorption, not prediction, is what separated the steady operators from the scrambling ones.
And absorption has a bottleneck. Watch a good operations desk during a rule-change week and you will see that the judgment calls are usually quick. An experienced operator knows within minutes whether a shipment is exposed, whether a customer needs a call, whether a quote needs to be pulled. What eats the week is everything around those calls: finding the affected shipments, rechecking the codes, redrafting the confirmations, keeping the audit trail straight. The scarce resource is not judgment. It is the hours lost to the busy work wrapped around judgment.
That is the observation our whole approach to AI hangs on.
Where we stand on autonomy in logistics
Here is our position, stated plainly, because the industry conversation about AI in freight tends to be either breathless or terrified and we do not find either useful.
Traditionally, all of this work is handled by people. We think some of it can now be handled by AI: reading the emails and documents as they arrive, working out what each request actually is, pulling the shipment references together, drafting the routine response, flagging the case that needs a human before anyone had time to look for it.
AI should take the busy work, and never the judgment.
And we hold two lines while doing it.
We would never take work that belongs with humans away from humans. Judgment calls, exceptions, customer relationships, the "why is this one different" cases: those stay with people. The job of the software is to make those cases easier to see and faster to act on, not to bury them under confident automation. Not fewer people. More meaningful work.
Autonomy has to know its own limits. The last wave of freight software promised visibility: dashboards that told you, in great detail, what had already gone wrong. The tools that win the next wave will do two things differently. They will know when they are not sure, and route those cases to a person instead of guessing. And they will get ahead of exceptions instead of reporting them after the fact.
The question we keep asking ourselves is the one July 24 made concrete: if the routine work handled itself, how much more rule change could the same team absorb? That is what autonomy is for. Not replacing the operator. Injecting agility exactly where the churn is landing.
What measured autonomy looks like in a logistics operation
The teams that came through July 24 cleanly were already running this way, whatever tools they used. Their habits share a shape:
They verify classification at quote time, not at clearance, because the schedule underneath the code keeps moving.
They quote with duty scenarios and state their assumptions on the rate confirmation, instead of quoting one number and hoping.
They track effective dates by mode, because ocean, truck, and air are no longer treated the same.
They keep origin and USMCA qualification paperwork current, because exempt and non-exempt now sit one document apart.
They assume clearances will be slower and build that into transit promises, because enforcement scrutiny is rising industry-wide.
Look at that list again. Every habit is repetitive, rule-following, deadline-driven work. Exactly the kind of work that breaks under churn when it depends on someone remembering to do it. Exactly the kind of work autonomy should carry, so the rule change gets handled as routine instead of as an emergency. Less chasing. Fewer surprises at the border. Clearer answers for customers.
The last six months made one thing plain: in this market, the advantage that lasts is not predicting the next tariff. It is an operation that keeps answering its customers while the rules change underneath it.
We build Ibis for the people doing exactly that work. Routine work handled. Human judgment elevated.