Where the hours actually go in inspection report writing
31 July 2026 · 6 min read · Aiman Salem
Ask an inspector how long a report takes and the answer is usually somewhere between two and five hours after a site visit. Ask what happened during those hours and the answer is much less clear, because most of it is not the part anyone thinks of as writing.
It is worth breaking down, because the parts that consume the time are not the parts that carry the professional value, and they are not the same parts.
Sorting and identifying photographs
This is usually the largest single block and the least visible. Forty photographs come off a phone in capture order, and the job is to work out what each one shows, discard the duplicates and the blurred ones, group them by area, and give each survivor a caption that locates it.
None of this is skilled work, and all of it is expensive, because the information needed to do it — where you were standing and what you were looking at — existed on site and was not recorded. Reconstructing it from the image is slow and sometimes impossible.
This is the block that responds best to changing what happens on site rather than what happens afterwards. A caption and a plan location captured at the moment of the photograph costs about ten seconds each and removes the entire task.
Retyping the skeleton
Every report has a substantial body of text that is the same every time: scope, standing limitations, the description of what a visual inspection does and does not involve, terms of reliance. Depending on the report type it can be more than half the document.
Copying it from the last job is the obvious answer and it is how the drift starts: after twenty jobs there are four slightly different versions of the limitations section in circulation, and nobody knows which one is current. The version that ends up in a disputed report is whichever one happened to be in the file you copied from.
This part is worth automating properly — one maintained source for the boilerplate, applied to every report — not because it saves the most time, but because the failure mode of doing it manually is the one with consequences.
Turning notes into prose
This is the part everyone means by "writing the report", and it is genuinely less of the total than it feels. Field notes are fragments; the report needs sentences. The translation is real work but it is fast, provided the notes are good.
When it is slow, the cause is almost always upstream: the notes were thin, so the writing is partly remembering. That is also when errors enter, because a half-remembered observation written up three days later is exactly the kind of statement that turns out to be wrong.
Formatting
Placing photographs so they sit near the text that refers to them, keeping headings consistent, fixing a table of contents whose page numbers no longer match, stopping an image from pushing a heading onto its own page. It is pure overhead, it produces nothing, and it reliably eats the last part of the evening.
Reviewing
The part that should get more time than it does, and usually gets the least, because it comes last and by then the report is late. Checking that every finding has its photograph, that no placeholder survived, that the recommendations follow from the observations, that the address is right in all six places it appears.
What is worth automating
The pattern is fairly clear. Photograph handling, boilerplate, formatting and mechanical checking are all work with a right answer that does not require judgement, and all of them are better done by software — reliably, identically, every time.
The interpretation is not. Deciding what a crack means, whether an item is a defect or fair wear, and what the client should do about it is the reason the report has professional value at all, and it is also where an AI draft is least trustworthy. A model can produce a confident paragraph about a mechanism it has no evidence for, and it will do so in the same tone as everything else in the document.
Which sets the useful division: automate the mechanical work completely, let software draft the prose from evidence you captured, and read every word of the interpretation before it goes out under your name.
Where the software fits
reportai.au is built around that division. Capture on site — photographs with captions, spoken notes, markers on a floor plan, measured floor levels — then a draft assembled from that evidence, with a quality-assurance pass that catches empty sections, leftover placeholder text and non-Australian spelling, and export to Word so the final edit happens in the tool you already use.
What it does not do is check whether a finding is correct. AI drafting can be wrong, and the draft is prepared for you to check and sign off, not to send. The review step above is the one part of the list that does not go away.
Written for practising inspectors and engineers. This is general guidance about documenting an inspection, not engineering advice about any particular building. reportai.au is software for writing reports — it does not carry out inspections, and it does not certify anything. Professional judgement stays with the person who signs the report.
More field notes · What reportai.au does · Pricing
Home · Pricing · FAQ · About · Field notes · Contact · Create an account · Sign in · Terms · Privacy · Refunds
reportai.au is a report documentation tool. It does not carry out inspections, does not assess or certify the condition of any building, and does not provide engineering services. Reports are AI-assisted drafts prepared for the inspecting engineer to check, correct, and sign off. All professional judgement remains yours.
Aiman Salem, trading as ReportAi.au · ABN 95 973 765 172 · admin@reportai.au · Prices in Australian dollars.