Human decisions, supported by AI
Different models have different jobs. We scope the task, evidence and human review before discussing a test.
AI close to the work
Put your own information to use where decisions are made. We help you define a local AI test around a clear task, the right source material and results that people can review.
Large and small language models (LLMs and SLMs) are different-sized models for working with text: searching, interpreting and summarising documents and operating rules. A smaller model may suit a bounded local task; the right choice depends on the material, review process and test criteria.
Computer vision uses specialist models to analyse camera and drone images or video. Language models have a separate text role: they do not themselves validate an image assessment. People check the original evidence and make the decision.
In a planned test, a discreet local gateway could run a language model close to the work and summarise information about a plastic tool. The technician reviews it and decides what happens next:
Scope the test around where queues form, which signals golf staff need and how an intervention will be evaluated. The goal is clear information for the team’s next decision.
Plan an inspection round with defined checkpoints and a trained operator in control. Image analysis and summarisation have separate tasks; the person makes the decision.
Planned development with NVIDIA Jetson, not a ready-to-use service or verified customer installation. Models, data and hardware are checked per test. Smart glasses and onboard AI remain development tracks, not verified functionality.
Scope your first AI testPractical reading
Välj rätt modell för rätt arbetsuppgift
Skilj textsökning och sammanfattning från specialistmodeller för bildanalys. Med ett teknikerexempel vid ett plastverktyg och ett arbetsblad för källor, granskning och testkriterier.