About Siteglass

Two sides of the same problem.

Construction businesses already have the knowledge required to quote well. It is spread across owner experience, old PDFs, supplier changes, site photos, inboxes and accounting systems.

Siteglass is being built where construction operations and agent infrastructure overlap: software that can assemble that evidence without pretending the owner’s judgement is optional.

The founders

No invented origin story. One founder works on cross-system AI and data infrastructure; the other runs commercial construction work.

Co-founder · Product and AI systems

Miles Cass

Perth, Western Australia

Miles builds CalibrAIte, a containerised intelligence layer for mine-site data. His public work focuses on correlating information from systems that were never designed to speak to one another, classifying it against local company documents and keeping an auditable route back to the source.

That background shapes the Siteglass architecture: connect to what a company already runs, keep tenant context local, expose narrow tools to the model and make every consequential action inspectable.

Miles on LinkedIn

Co-founder · Construction operations

Ryan Piconi

Brisbane, Queensland

Ryan is the founder of Nature Pools and Landscapes, a commercial pool construction company in Brisbane. His public project work includes complex rooftop and high-rise pool delivery where access, structure, waterproofing, sequencing and finish choices have real commercial consequences.

That operating experience keeps Siteglass tied to the actual job: supplier prices move, labour carries a real cost, exclusions matter, evidence is incomplete and a quote must leave enough margin for the business to finish what it promised.

Ryan on LinkedIn

Why construction AI can help

The useful part is not writing faster. It is reducing repeated handling while keeping calculations, evidence and authority clear.

The company is the source of truth

Past quotes, approved materials, labour rates, forms and terms should shape the result before generic model knowledge does.

Calculation and language are different jobs

The agent can explain a quote, but deterministic code must own cents, GST, quantities, margin and the profit floor.

External actions need an accountable human

A connector preflight should name the target, show the proposed change and wait for the configured authority before sending or posting.

The aim is simple: less admin around each job, with more evidence behind every commercial decision.

Siteglass should feel like the company’s operating memory—not a new system everyone has to babysit.

Talk through your workflow