AI for Engineering Firms: Why I Started Obnovit

It was January 2025, a cold one, and I was in my home office on a virtual coaching call with a handful of other entrepreneurs. Our coach told the room that every one of us should be spending at least five hours a week learning AI and working out how to put it into the business.

I had been a practicing engineer since June 2012. Chemical plant operations first, then plant engineering, then EPC work, then specialized consulting in the industrial energy and oil and gas sectors. Our firm was called Obnovit, Czech for “to renew,” a name we picked years earlier because we wanted to do renewable energy projects inside the engineering practice.

Sitting on that call, I had no idea that inside of a year I would put a decade of engineering work, and a client relationship we were being paid on at that moment, onto the table in one sentence I had not planned to say.

What I did not say on that call

The rooms I was in that winter were software founders, nine figure companies, and advisors who work with billion dollar companies, and they were all saying the same thing, that SaaS was going away and that every company would need to run AI first.

I felt overwhelmed and excited at the same time, and what I remember most is looking at a blue ocean with no idea where to start and no idea who could even help me. I did not say any of that to the room. I said it to my co-founder after the call, and we agreed we needed to work it out right away.

The first honest conclusion I reached is that it was too much for one person to digest alone, so I went looking for people who were further ahead.

A few weeks later a referral came through for a 12-week program built on the thing I already believed, which is that you start with foundations, build the system from scratch, and work up from there. We did not study AI in that program. We built working AI systems inside our own business, which is where the rule came from that I have never broken since. Everything gets tested in our own company before it goes out to a customer.

The five seconds that redirected the work

Two things were happening at once that spring. The hours were coming back in my own week, on proposals and reports and the unglamorous work that eats an engineer’s time. And when I asked the companies around me what they were doing, engineering firms, fabricators, construction, design build AEC, the answer came back the same almost every time. Some ChatGPT or Copilot here and there, a few people using it and a few not, and nothing behind it.

Then I was on a call inside that 12-week program, and the trainer said that anybody with a specialization of skill and knowledge should be building their own AI systems to help others in their industry, because those people are going to need a lot of help and you now have the expertise.

I went into that call thinking about how to put AI into our engineering business, and I came out of it knowing the new mission was helping other engineering companies do it.

That was the turn, and it took about five seconds.

The sentence I had not planned to say

Not long after, I was on a Zoom call with the engineering manager and head of engineering at a company we were doing engineering projects for. An active client, paying us at that time for something else entirely. We were talking about that project work, and I asked whether they had looked at AI at all.

They were aware of it. Scattered usage, a few people using it and a few not, no formal program, no strategy, and no clear sense of what it could do for the business. Underneath that they had a fair question about quality assurance, whether the output was valid, and how anyone would go about validating what came back.

I was listening to my own January story.

So I asked if they would want help with it, framed as competitive advantage and better operations. I want to be plain about what was in play in that sentence, because I did not think it through before it came out. I had no curriculum, no name for the offering and no price, and I had never run a training session for anyone in my life. Behind me was a decade of work as a professional engineer and a firm built on engineering services, and in front of me was a working relationship with a client who was paying us right then.

They said, sure.

The first session

We ran it on Zoom with over a dozen engineers and support personnel. I built it the way I build anything new, crawl before walk, walk before run, and nobody goes to sprint because they are excited about the technology.

I knew it landed from the questions they asked in the first hour. The intake-to-exit survey showed gains in efficiency and skill across the group, and we wrote up case studies out of their own wins. More of their team came through the sessions after that, and then they engaged us to build their AI strategic plan.

That answered the only question that mattered, which was whether there was an appetite for this inside an engineering company. There was. So we went all in and moved the business over to AI education and enablement for engineering powered businesses.

The name turned out to be about something else

We chose Obnovit for renewable energy work, and it fits better this way.

Some of us ran calculations on paper in college, and then CAD arrived, and Aspen and ANSYS and PV Elite, and work that used to take a week by hand could be modeled in a couple of hours or a couple of days. Nobody today mourns the drafting board. The computer did not replace the designer at the drafting board, it replaced the drafting board, and the engineers and designers became more valuable because the hours that went into drawing went into judgment instead.

That is what AI does inside a firm that uses it well, and it is also the answer to that engineering manager’s question about validation. You work with the model the way you work with a PhD-level intern. It will get you 80 or 90 percent of the way there, you iterate it to 95, and an engineer picks up the last five percent and owns the output. The peer checks and the approvals that protect the public do not go anywhere. We are engineers first, not tech experts, and everything we teach gets tested before it reaches production.

If you are sitting in your own January 2025 right now, overwhelmed and not sure where to start with AI, the thing that worked for me was not reading more news. It was getting around people further ahead, and then building one working system inside my own business before I said a word about it to anybody else.

Tell me which process in your firm eats 10 hours a week and produces nothing an engineer would consider true engineering, and I will tell you whether it is the one to map first.