From Site Boots to Software: Why I'm Building Practical AI for Construction
A personal note after leaving Operance and getting back to building
Last month I completed my exit from Operance, where I served as CEO. Stepping away gave me something I’d been missing: time to build again. And not “AI” in the abstract—tools I wish I’d had on site and in the office. I’m focused on practical AI that respects how construction really works: complex, time-pressured, and unforgiving of sloppy and ill informed decisions.
Why chatbots aren’t enough (and where they are brilliant)
Everyone can see the potential of AI—just look at how quickly chatbots like ChatGPT have become part of daily work. Chatbots are fantastic for unique questions and exploration. But many construction tasks don’t need “unique.” They need consistency. Consistency is the prerequisite for automation, and automation is how you get speed, reliability, and the ability to build more on top.
If a task can be written down as clear, repeatable steps, a human probably shouldn’t be doing it. Our value is in judgement, context, and creativity. That doesn’t make domain experts less important—quite the opposite. We need experts to define the business logic and the boundaries so machines can handle the grind and we can focus on the decisions that matter.
Construction is full of these repeatable workflows. They’re often unique to our industry and, historically, they’ve been done by hand.
I’ve done the grind
I’ve spent time on site hand‑marking drawings—circling changes, annotating, and assessing the implications. Done properly, it’s slow work. Done quickly, it’s risky. Miss a critical change in a revision and the fallout can be costly, cause delays, or, worst of all, be dangerous. That tension—too little time, too much at stake—has been with the industry for decades.
Where AI actually helps: the messy edges of a workflow
Generative AI lets us add intelligence to workflows that don’t fit neatly into rigid process maps. Take drawing revisions as a simple example. A good review isn’t just “spot the difference.” It’s:
- Is this definitely the same drawing that’s been revised?
- Is it the next revision, or are there intermediate versions we’ve missed?
- What has changed specifically? What’s the impact and significance?
- How do changes affect interfaces and knock‑on trades?
- Are the changes permitted within the contract?
- Are the changes compliant with the latest standards?
- Do they save money or cost money? Do they shorten or extend programme?
This is complicated, important work. Not every step can be automated—but many can be assisted, triaged, and accelerated. Anything that helps teams get to the important changes faster is a win for safety, cost, and schedule.
Decision latency is the silent killer
I still remember how we used to process information on one project—before email was everywhere. Each day’s post and faxes went into a manila folder with the site management team listed on the front. You’d tick your name after reviewing, then pass it to the next manager. On a complex project it took days for everyone to see the same information, then more time to file and act. We’re better today, but decision delays and unimportant noise still creep in. The longer it takes to surface the signal, the more expensive everything becomes.
What I’m building now: tools I wish I’d had
This is why I’m building MyDrawIt. It’s my first post‑Operance venture and it starts with something painfully familiar: extracting reliable data from drawings and safety documentation.
MyDrawIt turns unstructured technical drawings into structured JSON you can actually use. Drag and drop a file, get clean data or a downloadable Excel. No training courses. No six‑month rollout. Immediate value. And when you’re ready to go further, there are documented APIs so teams can automate parts of their own workflow and pull richer information from the drawings.
RAMS: from compliance burden to something genuinely useful
We’re also expanding beyond drawings with a RAMS (Risk Assessment & Method Statement) service. RAMS are critical but time‑consuming to review. Our new processor doesn’t just extract fields—it assesses document quality against Method Statement Assessment standards. It scores across 11 core criteria (from risk validation to emergency procedures), provides evidence for each determination, flags critical failures, and suggests improvements.
This isn’t about replacing professional judgement. It’s about giving construction professionals back time and focus. RAMS data extraction is live in the app. The assessment is accessible via API today and will be available in the drag‑and‑drop app in the coming weeks.
What I learned at Operance: obsession with workflow
At Operance I was obsessed with automation. We were running construction‑specific processes over and over across customers and projects. Every improvement to our workflow made things better for clients, reduced costs, and improved outcomes. That same mindset is baked into what I’m building now: remove friction, standardise outputs, and let experts spend more time on judgement.
An API‑first foundation that compounds
Construction’s unique processes mean we can’t transform everything at once. We need foundations: consistent outputs, interoperable data, and APIs that let us connect steps into flows. Link a few trusted processes, automate the predictable parts, and suddenly you can make small improvements that compound. That’s how we climb out of productivity stagnation—incremental wins that add up to transformational change.
Trust before hype
There’s plenty of hype around AI. I’m not chasing it. I’m focused on “trust‑building AI”: clear wins you can measure—time saved, errors reduced, compliance improved—without breaking what already works. Once teams trust the outputs, we can layer more intelligence where it helps most.
Join me
If any of this resonates—if you’ve been the person with a pen and a stack of drawings, or the manager waiting days for a folder to land on your desk—I’m building for you. Try MyDrawIt. You’ll get 10 free credits to start. Use the APIs if you want to go deeper. And follow along as I keep sharing what I’m learning about practical AI in construction.
Ian Yeo is the former CEO of Operance and founder of MyDrawIt. He specialises in helping construction organisations navigate digital transformation through practical AI implementation at Yeo Innovation. Learn more at ianyeo.com or try MyDrawIt at mydrawit.app.