Most firms that moved off paper moved onto a form. Forty fields, a photo slot against each, tick as you go, export at the end. It solved real problems: consistency across a team, structure without retyping, nothing skipped.
It also quietly changed what a site visit produces, and not many people have looked squarely at the trade.
What a form is good at
Genuinely a lot, and worth saying before the criticism.
A form makes coverage auditable. If the schedule says every riser door gets checked, the form proves each one was. It makes a junior's output legible to a senior. It stops the thing everyone has done at least once, which is get back to the office and realise nobody photographed the meter room. And it produces data that lines up across jobs, so a portfolio of surveys can actually be compared.
None of that is nothing. If your work is genuinely repetitive and the failure mode you fear most is omission, a form is the correct tool and this article is not for you.
What a form cannot hold
A form can only capture what someone anticipated at design time.
That sentence sounds mild. In practice it means the observation that decides the job is routinely the one with no box. Some examples of the shape:
- The defect is ordinary, but its position is the problem. Nothing on the form asks where relative to what.
- Three separate items are one cause. The form records three items.
- The thing is fine now but will not be after one wet season, and the form has “condition: satisfactory”.
- The occupier told you something on the way through that changes the interpretation entirely.
- It is a judgement call between two readings, and you want the record to show that you considered both.
Every one of those is expertise. None of them fits in a dropdown. So they go in the free-text box at the end, if there is one, or they go nowhere.
The workaround is the tell
Watch what experienced people actually do with form-based tools and you see the same adaptation everywhere: the fields get filled in quickly and carelessly, and the real content goes in “additional notes” as a wall of text typed in the car afterwards.
That is not misuse. It is a professional routing around a container that does not fit the work. But it means the structured data is the part nobody trusts, and the trustworthy part is unstructured, written from memory, and hours late.
You now have the costs of both approaches and the benefits of neither.
The second cost: where the work happens
The other thing a form does is move labour onto the walk.
Tapping is slow. It is slower with gloves, slower on a ladder, slower in glare, and slowest exactly when the site is most demanding, which is when the findings are most complex. Every minute spent completing fields on site is a minute not spent looking at the building, and it is the most expensive minute in the whole job to spend that way.
The write-up has to happen somewhere. Putting it on site does not remove it. It relocates it to the point where your attention is worth the most.
What we think the shape should be
Capture should be as fast and as unstructured as the moment allows, and the structure should be imposed afterwards, by something that read everything.
Concretely: talk while you walk. Say where you are, what you are looking at, what you think it is, and what needs to happen. That takes no fields, no taps, and no anticipation of what you might find. Then let the structure get derived from the capture, and check it.
That is the bet Sully is built on. The walkthrough and the narration go in together, and what comes back is a structured summary with action items, each line traceable to the frame and the words behind it. You review it, correct it and sign it. The form, if you need one, gets filled from a record that already contains everything, rather than being the thing that decides what gets recorded.
The honest caveat
Unstructured capture has a real failure mode, and it is the mirror image of the form's: nothing forces coverage. A form will not let you skip the meter room. A voice memo absolutely will.
If your work genuinely depends on completeness against a fixed schedule, keep a checklist for coverage. Just stop asking it to hold your findings. Those two jobs got merged into one tool for historical reasons, not good ones.