Kambi says AI now trades 85% of bets across its network. In its 4 November 2026 results update, the supplier also reported zero-second live delays for 59% of live bets traded by AI.
A closer look at live-bet acceptance
The company’s Q3 report says its system sets delays for individual bets using the match state and expected value. It also says NBA betting moved to AI trading after quarter-end, with work continuing on American football and table tennis.
These are supplier-reported figures. The 59% applies specifically to AI-traded live bets; it is not a percentage of all network bets. The 85% describes current coverage rather than a stated quarterly average.
What operators should measure alongside speed
Our assessment is that the next operational question concerns the full acceptance journey. A bettor experiences the interval between submitting a request and receiving a clear answer. Operators evaluating a trading service should therefore record that interval alongside the outcome, so that an improved delay setting can be assessed against the experience delivered.
A useful comparison would track acceptance rates, rejection reasons and changes to offered prices across equivalent events. The event mix matters: aggregate figures can conceal differences between sports or between quiet periods and decisive moments. Preserving that detail would help a trading team identify where a faster process improves availability and where further investigation is needed.
Our earlier coverage of BTA’s planned AI analysis for prediction markets examined explanations presented to users. Kambi’s update concerns trading and bet acceptance. For product teams, that distinction helps keep measures of an assistant’s usefulness separate from the reliability of the system processing a request.
Commercial review should connect these records with risk outcomes over the same period. More accepted requests alone would leave the quality of those decisions unresolved. Operators need enough evidence to review the trade-off between availability, pricing and exposure before drawing conclusions about a system’s performance.