We are sceptical of AI projects that begin with the technology. Ours begin with a queue that is too long, a forecast that is too wrong, or a review process that takes three people a week.
Where AI earns its keep
- Documents. Application forms, invoices, land records, scanned registers. If people are retyping information from a PDF, that is a candidate.
- Forecasting. Demand, footfall, collections, staffing. Anywhere a planner is currently guessing from last year’s spreadsheet.
- Detection. Duplicate beneficiaries, irregular transactions, quality defects — the needles humans miss at volume.
- Retrieval. Search across circulars, contracts and manuals that returns the clause, not a list of 400 files.
Our position on the hype
Not every problem needs a model. We will tell you when a rules engine, a better form, or a fixed database index would do the job for a tenth of the cost. That recommendation has cost us work before. We still make it.
