This is an illustrative scenario, not a real client. It shows how we'd approach a situation like this. Figures describe what we'd aim to measure, not results we're claiming.
The situation
A forty-vehicle haulage firm collects daily walkaround checks, defect sheets and delivery notes on paper. The office re-keys the important bits into a spreadsheet, defects sometimes get missed for days, and finding the record for a specific vehicle on a specific date is a hunt through a folder.
Where AI could help
The realistic opportunity here is document extraction: reading the handwritten and printed paperwork and turning it into structured, searchable records, with defects flagged automatically. It's one of the things current AI does genuinely well, provided the inputs are consistent enough.
How we'd approach it
- Intake. We'd look at the actual forms and how they flow back to the office.
- Opportunity scoring. Extraction would be scored against simpler fixes, because sometimes a better form beats any AI.
- Proof of concept. We'd run a batch of the firm's real, anonymised check sheets through a working demo and show the extracted, searchable result, so they can see the accuracy before committing.
- Report. Where it's reliable, where a human still needs to check, and the running cost per document.
- Handoff. A clear brief for a build, or an honest "the paperwork's too inconsistent for this to pay back yet".
What success would look like
In the proof of concept we'd measure extraction accuracy on their real forms, the office time saved on re-keying, and how quickly a defect surfaces. Those numbers, measured on their data, are what the decision should rest on, not a figure from someone else's case study.