OpenAI Dots: 24/7 agents and the hidden cost in Codex

OpenAI has introduced Dots, AI agents that run continuously and, during the launch period, do not consume the user's normal quota. However, this free access has nuances that should be understood before integrating them into our workflows. Thibault Sottiaux, head of core products at OpenAI, explained on X that each user's main Dot is included in the plan and available 24/7 without affecting the usage allocation. But if that Dot delegates a task to Codex, consumption is deducted as usual. That is, the agent can be on all day at no cost, but as soon as it starts creating tasks in Codex, the bill can go up.

This distinction is key for any IT team evaluating how to incorporate autonomous agents into their daily operations. The promise of always-on agents is attractive, but the underlying billing model can turn an apparently free tool into a source of unexpected expenses if task routing is not controlled.

The devil is in the consumption details

Sottiaux stated that the basic functionality “will always be included in your plan,” but the terms published by OpenAI are more restrictive: the use of Dots will not count toward eligible allocations during the first month after launch, and afterwards the conditions per plan will be published. This leaves in the air what will happen with the included quota once that window closes. For developers planning workloads based on Dots, the uncertainty is notable.

Furthermore, OpenAI halved the usage included in its $200 Pro plan on the same day Dots launched, and added a higher $500 tier. This move suggests that the company is adjusting its business model to monetize the enormous cost of keeping millions of agents online. Sottiaux expects “several million dots online in a matter of days,” which represents considerable expense even if they do not call other products.

Implications for SMEs and IT teams

For an SME, adopting always-on agents can be a double-edged sword. On the one hand, they allow automating repetitive tasks and maintaining a constant digital presence without human intervention. On the other hand, the fact that the agent can decide on its own where to execute the work (whether it does it itself, uses an included tool, or delegates to Codex) introduces a factor of unpredictability in spending. Without proper supervision, we could find that the agent has consumed the entire monthly quota on coding tasks that could have been resolved otherwise.

This scenario recalls the need to establish clear usage policies and cost control mechanisms in any AI initiative. Just as in software development we measure token consumption or choose models based on budget, now we will have to add a layer of agent management that decides which tasks are delegated to metered services and which are not. The lack of warnings before a Dot goes from included work to metered consumption is a risk that teams must mitigate with proactive monitoring.

At ForgeNEX we recommend starting with scoped pilot projects, defining spending limits, and periodically reviewing usage patterns. It is not about giving up automation, but about integrating it with the same financial discipline we apply to any cloud service. Experience with other AI tools has taught us that efficiency without control can backfire.

Relationship with other trends

This launch is part of a broader trend toward autonomous agents that execute tasks continuously. We have already seen how AI accelerates exploits and how a defensive mindset is no longer enough; in that context, managing always-on agents adds a new dimension of operational and financial risk. It is also worth remembering that there are alternatives on the market, such as AWS's open source agent which, according to the company, costs 45% less than Claude Code or Codex. The choice of platform and cost model becomes a strategic decision.

In short, OpenAI's Dots offer interesting potential to automate tasks at no initial cost, but their integration into enterprise environments requires careful analysis of workflows and consumption limits. The key is not to lose sight of the fact that what is free today may cease to be so as soon as the agent crosses certain boundaries.

Source: The New Stack. Analysis and adaptation: ForgeNEX.