Oberman Logistics owner Wes Oberman created a custom transportation management system using ChatGPT and Claude Code for just $55 per month, automating back-office processes and saving the company 20 to 25 hours each week.
- Oberman built the TMS using ChatGPT and Claude Code, acting as a mediator between two AI tools, with no coding background
- The custom system costs $55 per month versus $1,500+ quoted by existing TMS providers
- The TMS saves the company 20 to 25 hours per week
- The system automates rate confirmation extraction, load posting, invoice generation and delivery documentation
- Oberman uses GitHub, Supabase, Vercel and Resend as the technical foundation for his custom platform
Wes Oberman sold his big rig last month. As the owner of Oberman Logistics, an owner-operator trucking company out of Tennessee, he had to make the tough call to come off the road. Though only a 14-truck operation, his company that started just seven years ago has grown significantly enough that he had to park himself behind a computer in his home office in a shift from his office on wheels.
It’s there where Oberman spent a handful of weeks behind three computer monitors creating his own transportation management system (TMS) with zero coding experience. He said it was worth it.
“It took a lot of time, but it’s saving us a lot of time and money too,” he said.
Growing challenges
Oberman Logistics is in the process of onboarding an additional three owner operators. Oberman and his wife and business partner Laura were already taking on water prior to that, so they also onboarded a new dispatcher — a former driver who had retired with their company. Oberman said that created another set of challenges: Laura and the dispatcher were constantly calling each other for details the other needed, significantly slowing down operations.
It was time to find a TMS that could automate manual back-office processes, Oberman said, but that proved fruitless after multiple demos and trials with existing solutions on the market. He said most options seemed to be more tailored to broker operations instead of trucking and even less so for spot freight operations.
“None of it worked,” he said. “My problem with TMS is they're all so bloated (having more features than needed). They all look like they were made from the '70s. They're not very intuitive, and man, they cost a lot.”
He said the cheapest quote he received was somewhere around $1,500 a month. The cost of his self-made TMS is just $55 per month — the total of the subscription costs associated with each tool used to maintain the system.
Inspiration strikes
Oberman’s wife, who manages accounting, handles settlements on Thursdays. He said it used to be a 12-hour day for her as she manually entered and checked every number.
“She was drowning,” Oberman said.
Oberman threw her a buoy. Using ChatGPT and Claude Code, he created a settlement payroll program that uses AI to pull in all the numbers and does all the math for her. Now, she can wrap up that task within an hour from her spot on the couch.
He is currently working to integrate this standalone tool into the new TMS, which he said is saving even more time — 20 to 25 hours per week.
The idea to create his own TMS was also inspired, in part, by his wife. Oberman said he was already using ChatGPT for little things, like troubleshooting. He then got his wife hooked on it, and a comment on a Facebook post she made about it led Oberman to take on the challenge.
“One of her friends said I should try Claude and that he owns a business — I don't know this person at all — and that he used Claude to make some sort of point-of-sale thing,” Oberman said. “That gave me the idea of using Claude Code and ChatGPT in conjunction to create a program that would essentially work as a TMS.”
Assembly required
With ChatGPT and Claude on a widescreen monitor and the live website and test website on two separate, vertically mounted side monitors, Oberman began to assemble the pieces to his TMS, which functions as a subdomain on the Oberman Logistics website.
“I used Claude Code to build it, and I used Chat GPT to talk to Claude Code,” Oberman said. “I told Chat GPT what I wanted, and then it would give me a prompt to tell Claude Code. Then I would basically be the mediator between the two, talking and reading and adjusting what they were telling each other so that it was tailored to the way I wanted.”
He said ChatGPT told him everything he needed to do, from the coding prompts to which platforms to use, including GitHub, which stores, manages, tracks and shares software code, Supabase for building the web application, Vercel to host and deploy the website and Resend, which automates emails.
“Many times, it gave me way too many steps, and I’d have to ‘okay, slow down. I don't know what you're talking about. Give me one step at a time,” Oberman said. “It actually gave me slower, stupid steps for a layman that has no idea what he's doing.”
He later added an AI extractor (Claude Sonnet 4.6, backed up by Claude Opus 4.8) when the dispatcher shared how long it took to upload rate confirmations.
“Whenever Bill (dispatcher) or Laura upload a rate con, they have to verify it. It’s not foolproof 100% automation,” Oberman said. “But instead of sitting there reading over a rate con and then inputting everything, it just fills the boxes in for them.”
Before and after
Prior to implementing the TMS, an owner operator would book their own loads using the DAT load board for dry van and Truckstop’s load board for open deck. The rate con would be delivered to one of the Oberman Logistics emails, where it was then signed, sent back to the broker and sent to the driver via email. Once the bill of lading (BOL) was received, Laura would manually create an invoice in QuickBooks.
Oberman offered a full breakdown of the process with the TMS, which has been in operation for three to four months:
Open the load board and click “new load” on the left side of the screen. The system prompts the user to upload the broker’s rate con PDF. The user then clicks “extract with AI,” and the system automatically inputs the details into all the fields within 15 to 20 seconds, including a section at the bottom that notes special instructions such as broker stipulations like monetary deductions if the carrier doesn’t provide tracking or is late for their appointment.
Once the user clicks “save load,” the load is posted to the load board and assigned to an owner operator, who receives an email with all the details, including miles, origin and destination, reference number, special notes, etc.
The system is then updated each step of the way, from dispatch to pickup to delivery.
Once delivery is complete, the system changes the status to Proof of Deliver (POD)/BOL needed, and the owner operator emails an image of those documents to the dispatcher, who inspects it and uploads it into the TMS.
Accounting then receives a notification that the POD/BOL has been uploaded and approves or denies it. If denied, it goes back to the owner operator and dispatcher with notes. Once approved, the system automatically emails the broker a copy.
The status then moves to invoice. Once accounting clicks “invoice load,” the system creates an invoice and combines the invoice with the rate con and POD/BOL into a single PDF to be sent to the broker.
Oberman said next steps would include the addition of a driver app and integrating the ELD so the system would automatically email the broker when a driver is a certain distance from pickup or delivery.
He noted that updates to the broker consume a significant amount of the back-office staff’s time.
“They contact us constantly about it,” he said, adding that he’s not quite ready to take those steps just yet. “I don't want to automate too much because that's where failures can happen.”
