AI vs. Software: B&B Fight Night - Kp Reddy Vs Dustin Devan

Feb 4, 2026

KP Reddy, serial entrepreneur and construction-tech investor, and Dustin DeVan, seasoned construction-tech CEO and former general contractor, clash over whether AI can clean messy project data or firms must first structure it. They debate systems of record, ontologies, LLM roles, Excel’s staying power, data longevity, and what builders really need from tech. Quick, opinionated, and provocative.

Ask episode

AI Snips

Chapters

Books

Transcript

Episode notes

ADVICE

Verify LLM Outputs For Critical Decisions

Avoid treating LLM outputs as authoritative without governance and verification.
Dustin Devan advises layering deterministic systems for auditable business-critical decisions.

ANECDOTE

CityCenter Lien Experience

Dustin recalls being part of the largest lien in American history at CityCenter, Las Vegas.
He uses that experience to warn about messy records during litigation.

ADVICE

Standardize Before Scaling AI

Standardization helps AI make better inferences because garbage in yields garbage out.
Dustin Devan advises companies to build agreed schemas and SOPs before heavy AI adoption.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Should construction companies sort out their data before using AI? Or can AI sort it out for them?

We put two heavyweights in the ring to fight it out.

In one corner: Dustin DeVan, one of the most decorated CEOs in construction tech. Exited Building Connected for $275M. Now running Ediphi. He's seen $400M liens and knows what happens when your data falls apart in litigation.

In the other corner: KP Reddy, serial entrepreneur, founder, investor, and the man with 48 career experiences on LinkedIn (yes, we counted). He's backed by the biggest names in AEC and has been building stuff since he was a kid.

Dustin says get your house in order first. He's been through a $400M lien. Messy records in court? Not fun.

KP disagrees. Projects end, teams move on, data ends up everywhere. That's just how construction works. AI can make sense of the mess.

One thing they agreed on — Excel isn't going anywhere.

No punches pulled. No fluff.

Who won? You decide.

🎧 What's your take — do we need structured systems before AI, or can AI help us build structure from the chaos?

Our Sponsors:

Provision helps pre-construction teams find risks and scope gaps early so they can bid with confidence: https://www.provision.com/

Presented by Track3D - The first reality capture intelligence platform for smarter construction: https://track3d.ai/

HVAKR - Develop mechanical designs under budget and ahead of schedule: https://www.hvakr.com/

ALICE Technologies - Automates 'what-if' scenario exploration with AI — enabling you to rapidly test options, mitigate risk, and optimize schedules: https://www.alicetechnologies.com/home

Datagrid - Agentic AI that reads your docs, connects your tools, and gets the job done: https://datagrid.com/

00:00 Intro

01:00 Introduction to the Debate

03:25 The System of Record: Centralized vs. Decentralized

06:17 The Role of Language Models in Construction

09:18 Understanding Systems of Record

12:14 The Importance of Data Structure

15:07 Ontology and Context in Construction

17:58 The Expiry Date of Data

21:10 Business Building and Venture Capital in AEC

24:05 Excel vs. Cloud Tools

26:54 The Value of Historical Data

30:08 Defensibility in Business Models

33:10 Conclusion and Final Thoughts