Precon in the Age of AI

In this episode of Construction Conversations, I sat down with Patrick Murphy (Founder of Togal.ai) and Jeremy Crumley (Pre-Construction Director at Balfour Beatty) to talk about what’s actually happening in pre-con right now, where AI fits, and why the gap between what’s possible and what’s being adopted keeps growing.

Two very different seats at the same table. One building the tools, the other deciding whether to use them.

Construction Briefs is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Here’s what stood out.

1. Owners are finally paying for pre-con. That changes the equation.

Jeremy made the point that owners are putting real value on pre-construction services. Not lip service. Dollars. Balfour Beatty actually lost a project because their pre-con fee was too low, and the owner didn’t believe they’d get quality work at that price. Other owners are restructuring RFP scoring to avoid penalizing firms that invest more heavily in pre-con.

That shift matters. When pre-con stops being a loss leader and starts being a paid service, the incentive to tool up changes entirely. The tech investments that felt like a stretch when pre-con was “free” start to pencil out when owners are writing checks for it.

2. AI takeoffs work. The harder problem is earlier in design.

Patrick’s origin story is a familiar one for construction tech founders. He grew up in the family business, came back after a stint at Deloitte, looked at the books, and realized estimators were spending 50-60% of their time coloring and counting. That’s the lightbulb moment behind Togal.

For hard bid work with defined quantities, automated takeoffs are already delivering real value. But Jeremy pointed out something important. On CM-at-risk projects where you’re working from a couple of floor plans, one section cut, and some elevations, the task requires more imagination than automation. The design just isn’t far enough along for the software to do its thing.

That tension, between what the tech can do and where the design actually is, might be the most honest framing of where pre-con AI sits today. The tool works. The inputs aren’t always ready for it.

3. The prompt-to-CDs vision is closer than you think. And further.

Patrick laid out a bold claim. In three to five years, a basic building, apartment, hotel, house, could go from a prompt to a full set of construction documents. His reasoning: Togal is already processing hundreds of thousands of plans daily, and the number of ways to run plumbing in a four-bedroom house isn’t infinite. Once you have a database of a couple million houses, the patterns converge.

He’s not wrong about the logic. But Jeremy’s reaction, measured and honest, was a good check on the timeline. Most GCs are watching, not buying. Balfour Beatty’s current stack is Destini Estimator, Bluebeam, some Autodesk products, and a whole lot of Excel. The design community hasn’t been brought along for this ride yet. And technology always takes longer than the builders of it think. Patrick acknowledged that too.

The interesting question isn’t whether prompt-to-CDs will happen. It’s what the industry looks like when it does. Patrick invoked Jevons’ paradox. As the barriers to development drop, more gets built. More projects need builders. The pie gets bigger. That’s the optimistic case, and it’s not unreasonable.

4. The data goldmine nobody’s harvesting.

Jeremy pointed to something that should bother every pre-con director listening. The industry spends hundreds of thousands documenting projects. RFIs, change orders, scope gaps, architect tendencies. All of it captured, filed, and then basically forgotten.

Right now, most of that documentation is designed for litigation, not learning. Jeremy wants to flip that. Which architects need support in certain areas? Which subs come in low and then change-order you to death? Which RFI questions keep showing up in construction that should have been resolved in pre-con? The data to answer all of this already exists. It’s sitting in project files across every office.

Patrick’s take was pragmatic. Structured data is better, cleaner inputs produce more reliable outputs. But unstructured data is better than nothing. And the technology to make sense of messy project files exists today. The bottleneck isn’t the AI. It’s the organizational will to actually invest in a data strategy.

This is the unsexy work that separates firms who are “using AI” from firms who are actually getting value from it.

5. Entry-level jobs are the blind spot nobody wants to talk about.

Both guests circled around a tension that doesn’t get enough airtime. AI is displacing junior-level work first, and that’s exactly where the next generation of construction professionals is supposed to learn the business.

Patrick drew from his own experience. His first job after Coastal was at Deloitte. That entry-level role eventually got outsourced, then automated. The same pattern is emerging in construction back offices. Companies are hesitant to hire fresh grads when AI can handle what used to be a two-year training ground.

Nobody has a clean answer for this yet. Patrick’s optimistic take is that education systems will catch up and prepare graduates to manage AI tools rather than compete with them. But “eventually” is doing a lot of heavy lifting in that sentence.

Jeremy offered a different angle. A 40-year career means every year is 2.5%. Spend three years doing something you’re merely okay with and you just burned 10% going nowhere. His advice to younger professionals: find what you’re passionate about, invest in your contacts early, and stop treating your network like a LinkedIn vanity metric. Construction isn’t that big of an industry. The people you’re working alongside on a two-year project are the same ones you’ll be across the table from a decade later.

Patrick went bigger. The price of knowledge is going to zero. AI will be superintelligence in your pocket. The humans who work with their hands, the mechanics, the plumbers, the electricians, are positioned to become some of the highest-paid professionals in the economy. The paradigm is shifting in favor of the trades, and construction careers are going to look very different on the other side of this.

---

Conclusion

This conversation sits at the intersection of two realities. The first is that AI tools for pre-con are real, they’re improving fast, and the firms that lean in early will have an edge. The second is that most firms aren’t leaning in yet, and for good reason. The integration isn’t there, the design workflows haven’t caught up, and the data foundations are a mess.

The companies that pull ahead won’t be the ones chasing every new tool. They’ll be the ones who invest in getting their data right, give their teams room to experiment, and stop treating pre-con like a cost center.

Pre-con is quietly becoming the place where projects are won or lost. The teams that figure out how to pair human judgment with AI speed are going to have a very different next decade.