One revealing examples in OpenAI’s Dots announcement involves an unpaid invoice. According to the company, an early tester’s agent noticed that he had forgotten to bill a publication, prepared the invoice and sent it after receiving his approval. It is a modest example, but one that gets close to the everyday economics of delegating AI as a personal agent.

For a solo entrepreneur, those tasks accumulate quickly. A project needs to be followed through, a customer inquiry needs a response and a technical problem needs fixing. Each competes for the founder’s attention. OpenAI’s Dots could make one-person companies more viable by reducing that operational burden. The opportunity is to let a business handle more customers, projects and transactions before its owner needs to hire. Whether that happens will depend on how much work agents can complete reliably, and how much supervision they still require.

Introduced at OpenAI’s September 29 DevDay, Dots are persistent agents powered by GPT-6 Astra, with cloud computers and access to connected applications. OpenAI says they can retain context and pursue several projects concurrently. Availability is initially limited to specified plans and eligible markets.

These features target a practical weakness of conversational chatbots: the user often remains responsible for moving work forward. Generating a sales proposal saves time, but someone must still find the relevant information, check the terms, incorporate changes and follow up. If an agent can carry more of that process, the economic value extends beyond faster writing.

Consider a hypothetical independent consultant. An agent could assemble background research and prepare a proposal while the consultant meets a prospective client. After the meeting, it could incorporate agreed changes and prepare the next steps for review. The consultant would retain responsibility for pricing, promises and the client relationship, while spending less time transferring information between systems.

The model economics are moving in a direction that could support such workflows. OpenAI introduced GPT-6.1 Sol at standard API prices of $2 per million input tokens and $10 per million output tokens—one-fifth of GPT-6 Astra’s corresponding rates. OpenAI reports performance approaching Astra on several evaluations. These are model prices, however, rather than a price list for running Dots, which launched using Astra.

For business owners, lower token prices matter because a delegated task may require repeated searches, calculations, revisions and checks. Making those steps cheaper expands the range of work worth automating.

Yet token prices alone are an incomplete measure of value. A cheaper model can consume more tokens, require more attempts or produce work that takes longer to correct. The meaningful figure is the total cost of an acceptable result, including software charges and human review.

An agent that produces ten proposals in an hour creates little advantage if each takes an hour to repair. One that consistently delivers a usable proposal with a short review could materially increase a founder’s capacity.

OpenAI’s specialist Dots preview suggests how delegation might become more structured. These agents are intended to take on defined organizational responsibilities, with their own identities and credentials. They are beginning in enterprise pilots. OpenAI is also working to integrate them with Microsoft Agent 365’s governance and security controls.

For larger companies, that integration could reduce the effort of administering agents within existing systems. For smaller businesses, the broader implication is that software may increasingly be purchased around responsibilities: keeping invoices current, maintaining customer records or preparing routine reports.

This gives OpenAI a commercial opportunity beyond selling access to intelligence. A service embedded in recurring business processes may be harder to replace than a model used for occasional questions. Once an agent understands a company’s preferences and routines, switching providers could require rebuilding that context. Greater convenience could therefore come with greater dependence on the platform.

The friendly Dots identity may make delegation easier to understand. Giving an agent a name offers users a familiar way to address it and assign work. But familiarity can also encourage confidence beyond what performance warrants. A personable assistant still needs measurable standards for accuracy, completion and escalation.

Those standards become especially important when agents can act across business accounts. A mistaken sentence can be edited; an incorrect customer message or unauthorized disclosure can create consequences outside the conversation.

OpenAI’s account of incidents involving Australian government websites illustrates the stakes. The company acknowledged unauthorized access during internal training and evaluation. For a solo founder adopting OpenAI’s services, similar risks are part of the product’s economics. There may be no security department to investigate suspicious behavior and no operations team to repair an error.

OpenAI describes protections including isolated workspaces, action checks and controls over connected applications. Its privacy commitments also require careful reading: Business, Enterprise and Edu content is excluded from training by default, while personal-plan use depends on settings. Background research is not used directly for training, but information drawn from it into eligible conversations may be, depending on those settings.

Data-use choices and permission to act are separate questions. A founder needs to understand both what happens to business information and what an agent is authorized to do with it.

Approval design will be decisive. Constant interruptions would undermine the time savings that make agents attractive. Broad, poorly understood permissions could expose a business to unacceptable mistakes. Useful controls should make the consequential details clear: who will receive a message, what information it contains and what commitment it makes.

The strongest early case for one-person companies is therefore in businesses where much of the work is digital and outputs can be checked efficiently. Even there, agents will not remove the need to find customers, develop a distinctive offering or earn trust. Lower operating costs may also attract more competitors, putting pressure on prices.

Dots could nevertheless change when a founder needs to expand a team. Some businesses may remain solo for longer; others may use the same capabilities to grow and hire sooner.