← Customer stories

CUSTOMER STORY

What if you try to rebuild a moveOS analysis with a general-purpose AI assistant?

A colleague asked HBK engineer Mitchell Marks why he couldn't just use a general AI assistant. He ran the test himself: the same fluid dynamics problem, both ways, on the same morning. The difference was the setup each path needed before the engineering could start.

“Why couldn't you just do this in a general AI assistant?”

Mitchell Marks is Director, Automotive Strategy & Business Development at HBK. In two weeks on moveOS he built an app that troubleshoots electric-drive test data, with expert checks running as workflows. Read that story

Looking back on that build, he said:

I don't think I would have been able to hard-code this. The agentic aspect makes it possible.
Mitchell Marks, Business Development Director - Automotive, HBK

When he showed the work inside HBK, a colleague asked the obvious question: why not just do this in a general-purpose AI assistant? Mitchell decided to test it properly. He took a fluid dynamics problem a friend had posed a few weeks earlier and worked it both ways on the same morning. He knew he could get there on the do-it-yourself path; the question was what it would cost.

The same problem, the same morning, two paths. The steps on the first path are setup anyone would face.

Where the first four hours went

On the do-it-yourself path, the assistant could talk through the physics. Running a simulation was another matter. It needs a solver, and the free CFD tools need drivers, a Linux environment inside Windows, and a Python script to tie it together. None of that is engineering, and anyone on that path faces the same list before the first simulation can run.

Mitchell kept track of where the time went:

I have spent four hours installing drivers and decrypting how to use freeware CFD tools. Not even do the CFD.
Mitchell Marks, Business Development Director - Automotive, HBK

On moveOS, the solver and the tools were already installed and connected, so his first messages were about the problem itself. About 15 messages in, he was working on the engineering. Within just over an hour, he had a deployed interactive app with tunable parameters.

It's the pipes

Mitchell doesn't put the difference down to the model. He puts it down to everything around it: the tools are already installed and connected, the data is already reachable, and the agent works in a clear structure, keeping what it does in files it can refer back to. When he showed his work to colleagues that morning, his summary was three words: it's the pipes.

Explaining it internally, he put it this way:

I don't have to think about connections, data paths or installing tools. I just hit go and do engineering.
Mitchell Marks, Business Development Director - Automotive, HBK

From a white paper to an app

The same applies to methods engineers already have on paper. One of HBK’s technologists had written a white paper on measurement uncertainty. Using its methods and the analysis he had already developed on the platform, Mitchell created an initial app interface in six minutes. About an hour after requesting the interface, it was displaying real measurement-uncertainty results.

After the build: sharing it

Setup is only the first cost. Suppose the do-it-yourself path gets there: after the four hours of setup, a couple more produce a working tool. That tool lives on one computer, in an environment that took a morning to assemble. There is no simple, secure way to share it across the organization. Each colleague would repeat the setup, and someone has to look after access, data permissions and updates.

On moveOS, what an engineer builds is an app or a workflow inside the platform. Sharing it with colleagues is a matter of choosing who gets access, with the organization's security and permissions already in place. That is the part moveOS carries, so the person who knows the method can spend their time on the method.

What engineering teams can take from it

Try the do-it-yourself path, and time it

A general assistant is a good way to learn what AI can do with your problem. Count how many hours went to the problem and how many went to setup.

Ask how you would share it

A tool on one computer helps one engineer. Once colleagues want it, the work becomes access, security, updates and support.

Put engineering hours on engineering

The person who knows the method should spend their time on the method, not on drivers and data paths.

Bring one of your own engineering problems

See how far your team gets on moveOS when the tools and data are already connected, and what you build can be shared the same day.

More customer stories