There was a time when we counted SMS messages, when mobile calls were priced in thirty second blocks, and when home internet had strict caps. Every experiment carried a cost. That shaped how we behaved and it shaped what services emerged.
AI feels very similar today. To build a simple agentic workflow (an automated chain of tasks) I need:
All this if I want to test or automate my system. Every trial run consumes tokens. Every adjustment has a price tag. While it is rational at this stage of the market, it does restrict the imagination.
The internet only took off when it moved from metered and capped to always-on and effectively unlimited. That shift unlocked behaviours the earlier pricing models could never support.
I suspect AI will follow the same path. At some point we will see consolidation or bundling or a shift to inclusive pricing that will allow far more experimentation. Once the meter stops ticking, agentic systems will move from interesting prototypes to everyday tools.
We may look back at this period the same way we remember dial up: a necessary stage, but one that could never show us what the medium would become.
As an Investor Relations professional Iām trying to use AI to simplify tedious workloads. Does the cost of experimentation shape what you attempt with AI?