Table Of Content
A debate about systems of record became a preview of the SaaSpocalypse
On February 4, 2026, Bloomberg reported a $285 billion wipeout in software stocks. That same day, we hosted two of construction tech’s most experienced voices to debate whether the system of record is dead.
They didn’t know they were providing live commentary on a market crash. But listening back, the timing is almost eerie.
Dustin Devan, the CEO of Ediphi who previously exited BuildingConnected for $275 million, squared off against KP Reddy, a serial entrepreneur and investor through Shadow Ventures. The question: in a world of AI agents and LLMs, do we still need centralised software systems to manage construction data?
KP’s opening salvo didn’t mince words: “My view of software is it’s generally dead.”
Dustin’s response was immediate: “Generally does not work when we’re talking about construction and contracts.”
What followed was an hour-long debate that anticipated almost every argument now dominating tech media.
Construction Has Different Requirements
This is where Dustin pushed back hardest. Construction operates on contracts. On litigation. On documentation that must be immutable, auditable, and defensible in court.
“I was part of the largest lien in American history, $400 million in City Center in Las Vegas,” Dustin said. “You better have a system of record if you want to get through that and get paid.”
His argument centres on the industry’s fundamental structure: temporary projects, fragmented teams, contractual liability. These demand something more durable than an LLM wrapper.
“Language models are probabilistic,” Dustin explained. “And in construction, we write contracts. We need systems of record that are auditable, traceable, and we need to minimise the likelihood of errors.”
This echoes what SaaStr’s Jason Lemkin wrote this week: “Nobody is building a homegrown CRM in Replit to replace their Salesforce instance. Shipping a v1 is maybe 2% of the work.”
The LLM Wrapper Problem
KP’s most provocative claim focused on the AI startups themselves.
“All these LLM wrapper companies, that’s why they’re doomed in many ways,” he said. “I’m saying LLMs are like 5% of what I do. LLMs are tiny, tiny what we do. There’s so much more around the construct of AI.”
This cuts against the prevailing narrative. The fear driving the SaaSpocalypse is that AI agents will replace software. KP’s argument runs differently: the AI startups that think they can win by wrapping an LLM in a pretty interface are just as vulnerable as the incumbents they’re trying to disrupt.
“The construct of LLM being AI is one of the biggest misses that people are not working on,” he said.
Recent analysis supports this view. The “Year of Technical Debt” narrative suggests that 2026’s surge of AI-generated code will require extensive cleanup in 2027, as developers hunt down bugs introduced by “vibe coding.”
Where They Agree
Despite the sparring, both agreed on several points that matter for anyone building in this space.
Excel will survive. “I think Excel is the greatest software ever built,” Dustin said. “It’s the undisputed king.” KP agreed: “They keep trying to tell me to use Google Sheets around here and I won’t do it. Like I’m not a child. I need Excel.”
Interfaces will change, even if systems of record persist. Dustin acknowledged that “the interface of how we interact with our systems of record will change. That’s KP’s point. And I think it will be much more agentic.”
Data alone won’t protect you. When KP raised the “conspiracy theory” that Autodesk and Procore might use their data to compete with their own customers, Dustin’s response was dismissive: “For all the GCs or subcontractors or designers worried Autodesk or Procore are gonna come into your space, please don’t worry. That would be hysterical.”
KP’s punchline: “I would love to see them try. It would be an amazing Netflix special watching them try.”
The Real Question
IDC predicts that by 2028, pure seat-based pricing will be obsolete, with 70% of software vendors refactoring their pricing strategies around new value metrics such as consumption, outcomes, or organisational capability.
The debate between KP and Dustin centred on what replaces the current model and who captures the value.
Dustin believes the winners will be companies that “make you more efficient in your system of record” rather than those trying to build new systems of record with AI wrappers.
KP believes the entire category is being repriced: “I don’t think 10 years is long enough in our industry to do anything of impact.”
Both perspectives may prove correct. The SaaSpocalypse is killing the assumptions that made software feel inherently safe. What emerges will require both the structured, auditable foundations Dustin champions and the semantic, adaptive intelligence KP has spent 15 years building through ontologies.
Construction has always been an industry where theory meets reality at the job site. This debate was no different.
As KP put it, reflecting on what makes founders successful in this space: “If you want to solve problems, not all these problems are happening behind a Zoom screen.”
The full debate is available on the Bricks & Bytes podcast. Clips and highlights are available on our YouTube channel.
About Bricks & Bytes
Bricks & Bytes is the leading media platform for construction technology, covering the intersection of AEC, venture capital, and emerging technology. Subscribe for weekly insights from the people building the future of the built environment.
