Our latest confidence-scoring model runs on a third of the compute with no loss in accuracy — now rolling out across existing AI-powered software deployments.

The smaller footprint means lower latency for real-time decisions and lower running cost for partners operating at scale, without retraining or migrating existing integrations.

The rollout is staged: current deployments switch over automatically over the coming weeks, with monitoring in place to confirm confidence scores stay consistent with the previous model.