DATA & ANALYTICS
What is forcing the correction now is cost. Token costs were an abstraction when AI was experimental. Now that organisations are operationalising agents across functions, those costs are material, and there are legitimate questions about what you are getting for them.
The ones that come out ahead will treat governance not as a brake but as the infrastructure that makes progress repeatable: knowing what data agents are working with, who owns it and whether the context behind it is still accurate.
Q. DO YOU HAVE A BOLD AI PREDICTION FOR NEXT YEAR?
ยป The cost of frontier AI models is going to force a fundamental rethink of where enterprise AI actually runs.
AI governance will continue to become more critical. The next year will see enforcing separation between AI used for personal productivity and AI deployed for specific enterprise workloads. For the latter, smaller models, open-source models and industry-specific models running closer to the data will take on a much larger share of the work. Not because they are better at everything, but because they are good enough for well-defined tasks and economical in a way frontier models are not.
The organisations that get ahead of this will abstract that complexity away from the user entirely. The technology decides where work should go, which model, at what cost, for what task. The user just gets an answer.
aimagazine. com 133