About Punch List AI

Why it exists

Every app in this market is form-first. Pick a project, pick a location, pick a trade, type a title, choose a severity: and only then attach a photograph. That is four taps of administration before you have recorded anything at all.

So in practice nobody uses them on the walk. People photograph forty things, scribble abbreviated notes, and write the report that evening. By then a third of those photographs no longer have a clear memory attached, and the item that reaches the subcontractor is vaguer than the one that was seen.

That gap (not finding defects, which experienced inspectors do perfectly well) is the actual bottleneck. Punch List AI exists to close it: photograph the defect, get the write-up, correct it, move on. The item is finished before you walk away from it.

What it is honestly good at

• Turning a clear photograph of a visible defect into specific, trade-appropriate written English
• Consistent categorisation and severity grading, the same defect graded the same way on a Friday afternoon as on a Monday morning
• Drafting the remedial action for standard, recurring defects
• Reading patterns across a body of work that no individual could hold in their head
• Producing a PDF report that a client, GC or subcontractor will actually work from

What it deliberately does not do

We would rather you know the boundary than be disappointed by it.

• It does not find defects for you. It describes what you photographed. Walk the building the way you always have.
• It cannot see what is not in the frame, behind the plasterboard, inside the wall, what the drawing specified.
• It does not assess structural adequacy. Ever. That is engineering, not image description.
• It does not judge code or specification compliance, because it has not read your project's documents.
• It is not accountable. Your name goes on the report. Read every generated field before you save it, that review is what makes the output yours.

We wrote this out at length in How AI is actually changing site inspections, including where it fails and how to use it without putting your professional judgement at risk.

Who builds it

Punch List AI is made by Nice Express. It is a small, independent product, not a module inside a construction platform, and it is not venture-funded, which is why it costs $59.99 a year, or $9.99 month to month rather than a per-seat enterprise licence, and why there is no sales call.

The roadmap is shaped largely by what inspectors ask for. Feature requests sent to punchlistai@niceexpress.org are read by a person.

How we handle your data

Projects, issues, photographs, annotations and generated reports are stored on your device, in a private database. If you are signed in to iCloud they also sync, since version 1.1, through the private iCloud database of your own Apple ID, so an iPhone and an iPad stay on the same inspections and a new phone does not lose them. We cannot read that database. We do not run a server that holds your content, and there is no account to create.

When you tap Analyse, the photograph, with the project name, site address and a few recent issue titles as context: travels through a small relay we operate to Anthropic's Claude API, and comes back as a draft you review before it is saved. The relay exists so that no API credential ships inside the app; it stores nothing. Per Anthropic's published policy, API content is not used to train models. Subscriptions are handled entirely by Apple; we receive no payment details.

The full detail is in the privacy policy.

Platform

iPhone and iPad, iOS 17 or later. The iPad build uses a split-view layout and shares its projects with the iPhone through iCloud. An Android version is on the roadmap but has no date we would be comfortable promising , email us and we will tell you when it is in testing.