Blog

Field notes on agentic AI for the natural and built environment

How agent teams change the work of owners, developers, planners, consultants, contractors and the authorities who approve what they build.

Everyone14 July 2026 · 6 min read

What is an agentic AI operating system for the project lifecycle?

Chatbots answer questions. Agent teams carry workflows. The difference decides whether AI changes how projects run, or just how they are searched.

Read the post →

Contractors

Forecast Final Cost: the one number every contractor should see daily

Most projects discover their final cost at the end, one variation at a time. The teams that stay in control watch one live number instead.

30 June 2026 · 5 minRead the post →

Approving authorities

Review the model, not the PDF

Submissions arrive as static documents describing a live reality. Reviewers deserve the reality. Read-only access to living models changes the review, and the queue.

16 June 2026 · 5 minRead the post →

Engineering consultants

The re-keying tax: five packages of work, one environment

A ground investigation crosses five software packages on its way to a report. Every hand-off re-types data, and every re-type is a place where versions drift.

2 June 2026 · 6 minRead the post →

Government agencies

Agentic AI the public sector can audit

Public agencies cannot adopt AI they cannot explain. The requirements are stricter, and they are exactly the right requirements.

19 May 2026 · 6 minRead the post →

Developers & asset owners

Ground risk: what to ask before you commit to a site

The ground is the one part of a development you cannot redesign. The cheapest risk decisions happen before purchase, with evidence most buyers never assemble.

5 May 2026 · 5 minRead the post →

Environmental consultants

One data spine for environmental workflows

Baseline studies, impact assessments and monitoring commitments describe one environment — and usually live in disconnected documents. They do not have to.

21 April 2026 · 5 minRead the post →

Engineering consultants

Why AI in engineering must cite its methods

An engineer cannot sign what cannot be checked. The rule that governs human calculations should govern AI outputs, word for word.

7 April 2026 · 5 minRead the post →

Asset owners

The ground model should outlive the project

Owners pay for subsurface knowledge on every project, then receive it in a form that dies in an archive. The model, not the report, is the asset.

24 March 2026 · 5 minRead the post →

Contractors

Where margins go to die: the reconciliation gap

Between contract, claim, certificate and payment there are three hand-offs. Unmatched amounts hide in all of them, and workbooks find them too late.

10 March 2026 · 5 minRead the post →

Planners

Planning runs on documents. It could run on decisions.

Planners synthesise more documents per decision than almost any profession. Agentic AI will not make the judgment — it can finally carry the synthesis.

24 February 2026 · 5 minRead the post →

Everyone

Screening-level analytics: what they are, and what they are not

The most useful AI outputs in engineering are also the most misunderstood. A precise definition of 'screening-level' protects both the work and the trust.

10 February 2026 · 5 minRead the post →