How a PE Portfolio Company CIO Built a Production AI Application Without a Development Team

Can small-company IT teams be order shapers, not just order takers? Thomas Sweet, CIO of private equity-owned Industrial Refrigeration Pros, shows how rapidly advancing AI technology helps make it possible.

On December 27, 2025, I removed the "deny all" rule from the firewall, and RefrigAgent went live — a free tool that generates up to 90% of a regulatory compliance plan for refrigeration facility managers, and in turn delivers qualified leads to our sales team at Industrial Refrigeration Pros (IR Pros). Four months earlier, this product existed only as a proof-of-concept unit test. We had no development team, no software vendors, and no seven-figure budget. What we had was a three-person IT department, four local computer science students looking for a capstone project, a set of rapidly maturing AI coding tools, and a CIO – me – who wanted to show it could be done.

If you are a technology leader at a small or mid-sized company, especially a private equity (PE) portfolio company, and you have been wondering whether AI can move the needle for your business, it can. And you do not need the resources of a Fortune 500 to make it happen.

The business opportunity

IR Pros is a Sawmill Capital portfolio company serving over 2,350 customers nationwide – comprising eight subsidiaries assembled (“rolled up,” in PE language) into a national platform for the industrial cooling marketplace.

We saw opportunity on two fronts. First, our traditional customer base is in agriculture and protein processing, but the rapidly growing data-center cooling market presents significant expansion potential. The challenge: traditional customer acquisition through site visits, consultations, and trade shows is expensive and difficult to scale into a new market. And second, as a PE rollup, we faced a perception challenge that directly affects the company’s valuation: how do we show potential acquirers we are not just a collection of legacy businesses that “turn wrenches,” but an integrated, technology-driven organization?

Our CEO at IR Pros challenged me to build an application that would address both problems. He wanted it in 2025, and we committed to our PE firm that it would be done. The formal industry launch is scheduled for the International Institute of All-Natural Refrigeration conference in March 2026, but we shipped in late 2025 as promised.

What RefrigAgent does, for prospects and for us

RefrigAgent accepts piping and instrumentation diagrams–the detailed blueprints that document industrial refrigeration systems. It then uses machine learning and large language models to generate inspection, testing, and maintenance plans compliant with IIAR Bulletin 6, the industry safety standard. The output achieves up to 90 percent accuracy.

For facilities that need a compliance plan, this is transformative. Previously, getting that plan required engaging for a professional on-site assessment, and the average price of an in-person IIAR 6 assessment is $35,000. RefrigAgent provides refrigeration pros with substantial value immediately, with no meeting or consultation required. It is not intended to completely replace the work of a licensed Professional Engineer with onsite inspection, but still provides up to 90% of what would be identified through engineering review of plans. A RefrigAgent report allows users to immediately begin planning for needed maintenance, for example. In return, we ask only for contact information, which helps to collect qualified leads for our sales team.

However, the value for IR Pros extends beyond lead generation. Every diagram uploaded teaches us about the equipment installed at that facility. We can identify replacement part opportunities, develop custom service quotes, and build a richer understanding of our market — all before a salesperson ever picks up the phone.

How small IT teams can deliver on big ambitions

Here is the part that matters most for readers at smaller companies: this project would not have been possible twelve months earlier. The enabling factor was the rapid maturation of agentic coding tools, with some releasing updates monthly.

We used GitHub Copilot, Visual Studio Code, and Claude Code. I drove the architecture and stayed hands-on with my team – the three of us also responsible for running all IT operations across eight subsidiaries. The AI did not replace our judgment; it amplified what we could accomplish. Tasks that once required specialized developers have become achievable for a small team with demonstrated willingness to learn.

Let me be specific about the economics. We partnered with the University of Texas at Dallas on a student capstone class, a semester-long project where students apply classroom learning to real business problems. Total cost, a $5,000 donation plus tooling expenses. Our cloud-native infrastructure runs approximately $2,000 per month, including enterprise security tools. Development tools cost $700 per month. All other labor was absorbed within existing salaries. No new hires, no contractors, no managed service providers. These are results that would have required a full development team just a year ago, and months of budget negotiation.

One thing I underestimated: the learning curve for students working in our secure environment. They learn to code in school but rarely encounter an enterprise with layer upon layer of endpoint and cloud controls. Our environment was foreign territory. Though not all their code made it into the final release, we valued the partnership and the fresh perspectives they brought to the project.

Why PE sponsors should care

Private equity firms evaluating portfolio company technology capabilities often focus on the ERP, infrastructure stability and cybersecurity posture. Those fundamentals matter. But the more interesting question is whether the technology function can contribute to enterprise value beyond keeping systems running.

RefrigAgent demonstrates that a lean IT organization can ship revenue-impacting, customer-facing products. For portfolio companies preparing for exit, this changes the narrative. IR Pros is not a legacy rollup with bolted-on IT; we are an AI-focused company with a team that has demonstrated it can build. That matters to strategic acquirers evaluating integration complexity and cultural fit.

Some industry reports suggest that tech-savvy mid-market companies can command one to two times higher valuation multiples. Whether that specific premium materializes, the strategic value of demonstrating AI capability is real. We are leading our competitors, not following them.

I recently presented our approach to the CFOs of all Sawmill Capital portfolio companies. While our specific application is not directly applicable to roofing or flooring businesses, the methodology is. The playbook is to treat AI as a strategic partner, not a tactical convenience, and ask a simple question: What have you always wanted to build, but didn’t think you could?

What this means for CIO careers

For technology leaders at smaller companies, this kind of work changes your standing in the room. When you can walk into a board meeting and demo a production application your team built in four months for $15K, IT stops being a cost center. You are no longer the person who keeps the lights on and defends the IT budget. You are a builder who delivers business value — “the order shaper.”

PE sponsors notice. Boards notice. And frankly, the talent market notices. The CIO who can use AI tools to ship products is a different profile than the CIO who manages vendors and infrastructure. As these tools continue to mature, that distinction will only sharpen.

Our CEO asked IT to create competitive advantage, not just support it; we committed and shipped. But this project also reflected something personal. I have always wanted to build and ship a customer-facing web application. For years, the gap between vision and execution was too wide for a small IT team to cross. The tools finally caught up with the ambition. For CIOs at smaller companies, the opportunity has never been clearer: the tools exist, the economics work, and you do not need a massive budget or a team of developers. You need curiosity, commitment, and the willingness to trust a new way of working.