Skip to content

How tasks work

Tasks are the single currency for AI work across grids, workflows and Nova — here's what each one costs.

On this page
Tasks drawn down by a scheduled run.

One currency for AI work

Every plan includes a monthly pool of tasks. Grid cells, workflow steps and Nova messages all draw from the same pool, so you don't have to think in tokens, credits or per-feature limits. The pool is per organization and resets monthly: 100 on Brands Starter, 500 on Brands Scale, 2,500 on Brands Enterprise; 200 on Agencies Launch, 2,500 on Agencies Growth, 5,000 on Agencies Enterprise.

Running prompts doesn't draw from the pool — a scan, a run or a re-run is not a task.

What a grid column costs

TasksColumn
1Formula, Fetch URL, Publish, Google Doc, Human review, YouTube transcript
3Scrape page, Google search, GSC metrics
5AI prompt, Grounded research
10AI citation check, Competitor overlap, Citation intel

The grid editor prints a live estimate on the run button, so you see the cost before a run rather than after. Columns set to run only manually are left out of that estimate.

What a workflow step costs

Control flow is free — triggers, branches, filters, code, notify, loops, human review. A single outbound request (HTTP, connector, CMS publish, Google Doc, staging rows in a grid) is 1. Fetch- and search-shaped steps (scrape, web search, Reddit search, screenshot, PDF to text, GSC metrics) are 3. LLM-shaped steps are 5, and the citation fan-outs are 10.

Model tier changes the price of an LLM step: economy 2 · standard 5 · premium 10. Running a step on your own API key is a flat 2.

Going over your pool

The pool soft-caps rather than stopping work dead: usage past the allowance bills at $0.015 a task, up to twice the plan's pool. Usage sits in Settings → Billing next to prompts, workspaces and seats.