Hemut Wants to Run the Trucking Back Office With AI

Every trucking company has a person, or a room of people, who lives on the phone. They confirm pickups, chase drivers for check calls, argue over detention pay, and retype the same load details into three different systems. It is unglamorous work, and it is where a lot of a carrier's money quietly leaks out. Hemut, a company from Y Combinator's Spring 2025 batch, is betting that most of that work should be done by software - specifically, by AI agents that never sleep and never lose a rate confirmation.

Hemut calls itself an AI-native operating system for trucking companies. In plain terms, it is trying to be the single place a carrier or freight broker runs its back office: dispatch, load booking, quoting, real-time tracking, accounting, and the endless inbound and outbound phone calls that hold the whole thing together. The pitch the company puts on its own site is blunt - "We don't sell software. We drive outcomes."

The industry that runs the country, on software from 2004

American trucking is enormous and deeply unsexy. Freight moves the majority of domestic goods, and the industry is measured in the hundreds of billions of dollars. But the software running the back office of a typical mid-sized fleet has barely changed in twenty years. Dispatchers stitch together legacy transportation management systems, spreadsheets, email, and a phone. Data gets entered by hand, then entered again. The big mega-carriers can afford custom tooling and armies of staff. The thousands of small and mid-sized fleets cannot.

That gap is the opening. Hemut is not trying to sell a single clever feature into an existing stack. It is trying to replace the stack. The wager is that mid-sized operators are finally tired enough of duct-taping ten tools together that they will move to one platform - if that platform is obviously better, not just newer.

It helps to picture the size of the problem. A mid-sized carrier running dozens of trucks might touch a hundred loads a week. Each load carries a small tail of coordination: a quote, a booking, an appointment, a pickup confirmation, a series of check calls while the freight is in motion, a delivery confirmation, and then the paperwork that turns all of it into an invoice. Multiply those handoffs across a fleet and a week, and the back office is not a support function - it is the business. Most of those handoffs are still done by a human reading one screen and typing into another.

A founder who filed the paperwork himself

The reason to take Hemut seriously is the person behind it. Loki Cheema, the founder and CEO, did not study the trucking problem from a distance. By his own account he was homeless at 16 and later worked 100-hour weeks at a motel, where he met a truck driver and started helping him with paperwork. That turned into a dispatching business, and eventually a 16-truck fleet he built from scratch. Six years of hands-on dispatch taught him exactly where the hours and dollars disappear.

He started Hemut in 2024, reportedly out of a UCLA dorm room, and took it into Y Combinator's Spring 2025 batch, where the primary partner was Pete Koomen. The founding story matters here for a practical reason: in an industry that is skeptical of Silicon Valley and allergic to change, a founder who has answered the phones himself is a different kind of salesperson. He can sit across from a fleet owner and describe the exact shape of a bad day - the load that fell through because nobody returned a call, the detention hours that never got billed - because he has had those days.

That credibility shows up in the product decisions too. Hemut did not start with a flashy consumer-style feature and work backward. It started with the parts of dispatch that are pure friction, the tasks a person does only because software could not, and pointed automation at those first. It is a less glamorous roadmap than a sweeping "AI for logistics" pitch, and it is probably a more durable one.

What you actually get

Hemut's product line is organized around jobs rather than jargon. The core system is a full transportation management system with AI built in from the start. On top of it sit a set of agents and modules aimed at the specific tasks that eat a dispatcher's day.

Underneath those sit the task-specific agents Hemut names for exactly what they do: a Quoting agent, a Scheduling agent, a Tracking agent, and an Asset agent. It is a deliberately literal naming scheme, and it tells you how the company thinks - each agent is pointed at one repetitive, money-adjacent task, and the value is in stacking them.

The voice piece is the one that tends to make trucking people lean in. A large share of a dispatcher's day is spent on the phone, and much of it is routine: confirming an appointment, asking where a driver is, reading a rate off a confirmation. Those are precisely the calls that an AI agent can make or take without a human waiting on hold. The promise is not that the phone goes quiet - it is that the humans get pulled off the routine calls and put on the ones that actually need judgment, like a load going sideways or a customer relationship worth protecting.

Who is buying, and how the money works

Hemut sells to trucking carriers, freight brokers, and fleet operators - the mid-sized operations running regional or national fleets that have real volume but not a big software budget. The business is B2B SaaS, but the company frames its pitch around outcomes it can point to: more revenue per load booked, dispatcher hours saved each week, and lower overall software and IT spend. That framing is a sales strategy as much as a product one. It moves the conversation away from "another subscription" and toward "here is what changed on your P&L."

It is a smart place to plant a flag. Software sold on features gets compared feature by feature, and there is always a competitor with a longer list. Software sold on outcomes gets compared to the status quo, and the status quo in trucking is a lot of manual hours. For a fleet owner weighing whether to switch, the question Hemut wants in their head is not "does this have every button the old system had," but "does this get more freight booked with fewer people on the phone." Those are very different conversations.

The competition, and the hard part

Hemut is not walking into an empty field. Established players like McLeod Software and Truckstop own the legacy TMS market, and telematics companies such as Samsara and Motive have deep hooks into fleets. A wave of other AI-freight startups is chasing the same opening. Hemut's differentiator is a specific one: it was built AI-native and it aims to unify the whole back office, rather than bolting an AI feature onto software designed before large language models existed.

The hard part is the same thing that makes the opportunity real. Trucking operators are conservative, and ripping out the system that runs your business is risky. Winning here is less about a demo and more about proving, load by load, that the new way loses fewer confirmations and saves more hours than the old one. So far the company reports approaching $1 million in booked revenue and a team that grew from four people to about twenty within roughly its first year - early, but pointed in the right direction. Hemut has also been selected for Fuel Accelerator's AI/ML cohort focused on trucking tech.

There is also a timing argument in Hemut's favor. The kind of automation it is building - agents that can read a messy scanned document, hold a natural phone conversation, and make a decision inside a workflow - was not practical a few years ago. The capability arrived roughly at the same moment a founder with deep trucking scars decided to point it at freight. Incumbents can add an AI feature, but retrofitting an AI-native design onto twenty-year-old software is a different and slower job. That is the window Hemut is trying to move through before it closes.

None of this guarantees anything. Booked revenue is not the same as durable, renewing revenue, and a twenty-person team is still a small crew against entrenched competitors with thousands of installed customers. The company will have to prove it can support fleets at scale, keep its agents reliable on the calls that matter, and earn the trust of operators who have been burned by software promises before. Those are real risks, and Hemut is early enough that they are all still open questions.

There is a lesson buried in Hemut's approach that other founders can borrow directly: pick an industry everyone finds boring, work inside it long enough to feel the real pain, then build the tool you personally needed. Boring is often where the money hides, and lived experience is a moat that is genuinely hard to copy. Whether Hemut becomes the operating system for mid-sized trucking is still unwritten. But the company is aimed at a real, expensive problem, and it is being run by someone who has felt that problem in his own hands.