A 20-minute call
Tell us what is not working. No deck, no discovery phase, no invoice.
Book the callAI in data work
Research and classification that took a person a week, done in an afternoon — with the accuracy measured, not assumed.
- Timeline
- 3 to 4 weeks
- From
- from $6,000
- Commitment
- Fixed price
Classifying, enriching and researching at volume is exactly what these models are good at, and exactly where most people deploy them without ever checking the error rate.
What it covers
- Classification, extraction, enrichment and summarisation at volume
- A human-labelled sample first, so accuracy is a measurement rather than a claim
- Confidence scoring, with low-confidence rows routed to a person
- The pipeline handed over as something your team can re-run without us
What you get
- The working pipeline plus documentation
- A measured accuracy report against the human-labelled sample
- A cost-per-thousand-records figure so you can decide whether to scale it
Not included
- Buying the source data
- Model API costs at volume, billed at cost
- Any use that would breach a source's terms of service
Start small
Three doors. The first one is free.
You should be able to find out whether we are any good without signing anything.
One first engagement
A published band, a fixed scope and a date.
See the price bandsWe run it for you
Outbound, a care plan, or an ongoing build team.
Get a quote