One prepares an account list. Another drafts the outreach. A third reviews campaign results. Each completes its assignment, but they use different definitions of a qualified customer.

The research looks competent. The email looks polished. The report looks precise. Together, they describe three different strategies.

Delegation made the work faster. A shared definition would have made it coherent.

Dots make the coordination problem worth solving

OpenAI introduced dots at DevDay on September 29. Its documentation describes an agent that works between conversations, uses a cloud computer, and draws on relevant context and memory.

A dot can use background agents and delegate tasks to ChatGPT Work or Codex. People can reach it through channels including Slack and Teams. Permitted plugins provide access to connected apps.

OpenAI also documents permissions and action review. The platform already addresses business responsibilities and boundaries.

For a dot coordinating an ongoing responsibility, useful delegation involves work that depends on earlier work. Research shapes the outreach. Outreach produces results. The report helps the team decide what to change.

The company's definitions have to remain consistent across those assignments.

Our companion essay on company work's missing repository explains the broader case for preserving the knowledge those tasks depend on.

Qualified Account needs to mean the same thing in three assignments

Consider an illustrative campaign where the approved target is a US B2B company with at least five outbound sales representatives.

The research assignment uses that definition of Qualified Account. The writing assignment works from an older brief that also includes sole proprietors. The reporting assignment counts every company contacted as qualified.

Each specialist can produce a reasonable deliverable from its instructions. The combined campaign no longer represents the agreed target.

The researcher excludes a company the writer assumes is in scope. The report includes contacts the researcher would never have selected. A change in the measured results may reflect a change in the definition rather than a change in campaign performance.

When dots delegate work to background agents or tasks in ChatGPT Work or Codex, this is the kind of company instruction that should stay consistent across the assignments.

The specialists can have different jobs while relying on the same meaning of Qualified Account.

Meeting Booked can hide a second disagreement

Take another hypothetical example from the campaign.

The outreach summary counts any proposed calendar slot as Meeting Booked. The campaign report counts only invitations the prospect accepted. A rescheduled meeting appears twice in one summary and once in the other.

The team cannot compare the results usefully until it agrees on what the measure means.

For this illustrative campaign, the approved definition could be one prospect meeting with an accepted invitation, counted once even if it is rescheduled. Attendance would be a separate measure.

That instruction changes several tasks at once. The outreach summary needs to distinguish proposed slots from accepted invitations. The report needs to avoid counting a reschedule as another meeting.

Those are concrete conditions the team can agree on before assigning the work.

Ongoing work needs an explicit decision when a definition changes

Dots work between conversations and keep making progress as circumstances change. That makes a mid-campaign decision relevant to work already under way.

Suppose the sales lead changes the minimum team size from five outbound representatives to ten on Wednesday.

Monday's research used five. Thursday's outreach could use ten. Friday's report could accidentally compare the two as though nothing had changed.

The team has a decision to make: finish the current campaign under its original definition, or revise the active list and make the report reflect the change.

The approved choice should reach the assignments it affects. An agent continuing the work can then use the team's decision rather than infer it independently for each task.

HexaHQ keeps the definitions with the work they guide

HexaHQ is where the company keeps its current definitions, corrections, and approved workflows for supported AI clients.

Qualified Account can carry the agreed campaign criteria. Meeting Booked can carry the conditions that make the report comparable. The campaign workflow can preserve how those measures are used together.

The benefit is coherent work: specialists produce different parts of the result from the same approved understanding.

Before delegating your next campaign, settle what qualified accounts and booked meetings mean, and give each assignment the agreed definitions.

Delegated work only adds up when its definitions agree.