The Task Creation Assistant

The Task Creation Assistant drafts a task from an audit finding: what the problem is, why it matters, the steps to fix it, and a definition of done. You can then refine it in conversation before creating it.

It spends credits, and it says what it will cost before it runs. There are two charges: one for the initial draft, and one per refinement turn.

Why it exists

A finding is not a task. "Twelve pages are missing a meta description" is a fact; what someone needs is which pages, what to write, in what order, and how to know when it is done.

RankX AI can flatten a finding into a task without the assistant, and that is free and often enough. The assistant is for the findings where it is not: where the fix depends on your site's shape, where the steps are not obvious, or where the task is going to a person who was not there when it was found.

When to use it

Use it when the task is going to somebody else. A task written for a developer who has never seen the audit needs context that a one-line finding does not carry.

Use it for a finding whose fix is genuinely non-obvious: a canonical disagreement, a mixed-content problem across templates, an indexing question with several possible causes.

Do not use it for a finding whose fix is one line. "Add a meta description to these four pages" does not need drafting, and creating the task directly is free.

Do not use it in bulk. The value is in the specific piece of thinking, and drafting fifty tasks produces fifty pieces of prose nobody reads.

What it produces

The same structured shape every task has, filled in properly:

• The problem, stated in terms of your site rather than in terms of the check.
• Why it matters, which is what makes the task survive being deprioritised and picked up a month later.
• Numbered steps.
• A definition of done, so completion is checkable.
• The affected pages, where the finding recorded them.

Refining it

You can refine the draft in conversation, and each refinement turn is charged. That is worth knowing before a long back and forth: two good turns are usually better value than six vague ones.

The most useful refinements are the ones that add context the assistant cannot have: which template the pages share, who owns that part of the site, what was tried before.

What it does not do

It does not fix anything. It writes the task. Applying a fix to a connected WordPress site is a separate action with its own dry run, and it currently covers the SEO title and meta description. See Tasks.

It does not invent findings. It works from an audit issue that already exists.

It does not know your team. Steps are written for a competent person with access to your site, not for a named role.

Cost

Two charged actions: the initial draft, and each refinement turn. The current figures are on the credit cost reference, which is generated from the product. Prices are not repeated in prose here, because rates in RankX AI are versioned and republishable.

Creating a task without the assistant is free, on the board or over MCP.

Where to go next

• Tasks, for the board itself.
• Running an audit, for the findings it drafts from.
• Credits and metering, for the spend model.