Zero RFIs, $13.8M, and an Inference Inflection Point

Tech veteran KP Reddy has launched Zero RFI, an AI-native platform with $13.8M in seed funding led by General Catalyst. The platform is positioned as infrastructure software for core construction processes - not a bolt-on productivity tool, but something built from the ground up to replace the workflows that generate the most waste on site.

The name makes the ambition explicit. RFIs - Requests for Information - are one of the construction industry's most persistent inefficiencies. They're expensive to process, slow to resolve, and a reliable indicator of where coordination has broken down. A platform that actually reduces them, rather than just digitising the existing mess, would be genuinely valuable.

But it's worth being clear about what "AI-native" means in practice. The platform's real test is whether it reduces friction in the field or whether it adds another layer to manage. Early-stage construction tech has a history of promising one and delivering the other.

Semble AI has been profiled for its work on AI agents for building-system design - specifically reducing the time spent on technical coordination for commercial and residential projects. The approach automates specialised design workflows that currently require significant manual effort from engineers and design managers.

This is a narrower problem than Zero RFI is targeting, but arguably a more clearly defined one. Building-system design coordination - MEP clashes, design iterations, specification checks - is a real bottleneck in the pre-construction phase. If AI can accelerate that reliably, it frees up the design team to focus on the decisions that actually require expertise.

Tech companies - Google most visibly - are running voluntary buyout programmes to reduce headcounts in roles that no longer align with AI-first priorities. It's a softer approach than traditional layoffs: severance, accelerated stock vesting, extended health coverage. But the underlying message is the same. Work is being restructured around what AI can now do.

This isn't a construction story yet. But it will be. The physical site roles are protected - as Geoffrey Hinton noted last week (16 March), dexterity and real-world problem-solving are still the hardest things to automate. But the coordination, scheduling, documentation, and compliance work that surrounds those physical roles? That layer is exactly what platforms like Zero RFI are targeting. The combination of falling AI capability costs and purpose-built construction tools means the timeline for that restructuring is shortening.

Jensen Huang's two-hour GTC keynote this week had one central message: the AI industry has moved from a training-compute race to an inference deployment race. In Huang's framing, the "inference inflection point" means the focus is now on how fast, how cheaply, and at what scale AI can be run - not just how powerful the underlying models are.

For construction, this matters more than it might seem. The expensive, inaccessible AI tools of two years ago are becoming cheaper and faster with every hardware and software generation. The Blackwell and Rubin chips are selling well. NVIDIA's speculative decoding improvements are reported to deliver up to 1.69x speedup over previous methods. Google launched Gemini Embedding 2 - a single multimodal embedding space across text, image, video, and audio. The infrastructure layer for running AI is maturing rapidly.

This doesn't mean construction firms should rush to buy GPU clusters. But it does mean that the cost-benefit calculation for AI tools is shifting. Enterprise-grade capability is moving within reach of mid-market businesses, not just the global contractors with large technology budgets.

JFB Construction has announced a stock split ahead of its merger with XTEND, which will create XTEND AI Robotics. It's a signal of continued investor appetite for construction-and-robotics plays - the bet that the future of construction involves autonomous or semi-autonomous machines doing work that's currently done by people.

This is a longer-term trend than the software story, but it's part of the same picture. As AI capability improves and hardware costs fall, the economics of construction robotics get more interesting. JFB/XTEND isn't a UK company, but the trend it represents is global.