Gopichand Kotana

It feels like we’ve hit a strange equilibrium in the AI space.

The big players have remained the same over the past year. The benchmark wars still happen, but most users don’t really care anymore. In practice, people are choosing tools based on workflow fit and integration, not on marginal differences in reasoning scores.

At the same time, access is unusually generous.

In many companies, employees have near-unlimited use of powerful models. I have seen people using the highest reasoning settings to rewrite one-line emails. Advanced AI systems are being asked to summarize meeting notes or tweak phrasing. And the marginal cost feels close to zero.

But my hunch is that this is temporary.

These tools are heavily subsidized by AI companies, driven by the need to capture market share. But the companies building these tools and models are no longer research-heavy outsiders. They’re becoming the next generation of large software firms. As they grow, maintaining revenue discipline will become crucial for them. I wouldn’t be surprised if they are already considering ways to charge more or offer fewer tokens or less power for the same price.

In large corporations, access to tools is likely to be more structured. There will probably be usage limits, tiered permissions, and the need for internal approvals. This is similar to the existing infrastructure. None of this means the tools will get worse. In fact, they’ll probably get better. But the environment around them may become tighter.

If you have access to powerful models, you can freely test workflows, experiment with ambitious integrations, or even push their limits a bit just to understand their boundaries. That freedom may not seem special, but it probably is.

We may look back at this period as the phase when AI was both powerful and unusually frictionless.