Perplexity Computer and Automations for marketing work

#01What can Computer do for a marketing team?

At 100 credits per $1, a weekly research report costs a few dollars a month. A rebuilt dashboard every week costs much more. Lesson 2 covers allowances by plan.

Computer can also run Research and Model Council as skills. You open them from the + menu or by typing "/".

#02What are Automations, and how do they work?

Perplexity's 29 September 2026 post introduced Automations for Computer:

• Schedule triggers: daily, weekly, monthly or a custom cadence.
• Event triggers: events in Slack, Gmail, Outlook, Linear or GitHub. You can narrow them with conditions, such as "emails from a specific sender".
• Memory across runs: each run knows what earlier runs found, so it can report changes instead of repeating everything.
• Project context: an Automation can use a Project's files, tools and instructions.
• Management: an Automations dashboard to edit, pause or delete.
• Credits: Perplexity says watching for a trigger doesn't use credits, but each run does.

The older Perplexity Tasks (scheduled searches managed in notification settings) are separate. Perplexity says they can be migrated into Automations.

#03Which marketing Automations are worth building first?

Our view: if an Automation's output isn't read by a named person every time it runs, delete it. Automations that nobody reads just burn credits.

#04How do you set up an Automation step by step?

1. Write the job as a brief. Include the inputs, the steps, the output format and where it goes.

Every Monday at 07:00 CET:
1. Visit the pricing and changelog pages for Competitors A–D (list in project file).
2. Compare with last week's run.
3. Post to #competitive-intel: only changes, each with the URL and a one-line "why it matters for Ledgerline".
4. If nothing changed, post "No changes".
Do not post anything else. Do not contact anyone.

1. Run it once by hand as a normal Computer task. Check the output and the credit cost.
2. Create the Automation by describing the task and cadence, then confirm the schedule Perplexity proposes.
3. Attach it to the right Project so it uses your instructions and competitor file.
4. Watch the first three runs and compare them with a manual check.
5. Set a review date. Is anyone using it? Is it accurate? Is the credit cost fair?

#05What guardrails should you set?

• Human in the loop for external actions. Emails to customers, social posts, ad changes and CRM updates should only ever be drafted for a human to send.
• Least-privilege connectors. Connect only the apps and channels the Automation needs.
• Credit budgets. Note the expected cost per run in the Automation description and review it monthly.
• Named owner. Every Automation has one person responsible.
• Prompt-injection awareness. Pages Computer reads can contain instructions aimed at agents. Keep Automations read-only where you can, and treat odd outputs as a signal to check.

#06Computer vs the API vs n8n: which should you use?

Many teams use both: Computer for ad hoc work and prototypes, then n8n or the API once a workflow is proven and needs to be cheaper or more controlled.

#07Worked example: Ledgerline

Ledgerline is a fictional accounts-payable software company. Details are illustrative.

Ledgerline's marketing ops lead built three Automations in the "Competitive intel" and "AI visibility" Projects.

Walkthrough

1. Wrote each Automation as a brief with explicit "do not" rules.
2. Ran each one manually and recorded the credit cost.
3. Created the Automations and attached them to Projects.
4. Checked the first three runs of the competitor diff against manual checks. One false positive came from a cookie banner change, so the brief now says "ignore layout or cookie text changes".
5. After a month, cut the dashboard Automation and moved that job to n8n.

#08Common mistakes

• Scheduling before testing. Run it once by hand first.
• Vague outputs. "Summarise what's new" produces essays. Specify a format.
• No "nothing changed" rule. Without it, agents tend to make up something to report.
• Agents with send rights. Draft, then approve.

#09Try it yourself

Write the Monday competitor diff brief for your top two competitors. Run it once as a Computer task, note the credits used, and decide whether to schedule it. If you're on Free, run the same brief as a Pro Search to test the instructions.

#10Quick quiz

1. Which apps can trigger Automations, according to Perplexity's launch post?

Answer: Slack, Gmail, Outlook, Linear and GitHub, plus schedule triggers.

2. Why is memory across runs useful for marketers?

Answer: It lets an Automation report changes since the last run instead of repeating everything.

3. What should an Automation do when nothing changed?

Answer: Say so explicitly, for example by posting "No changes".

#11What should you check before the next lesson?

• Each Automation has an owner, a cost estimate and a review date.
• No Automation sends external messages without approval.
• The first three runs were checked by hand.

Previous: Projects for teams. Next: The Comet browser for marketers. Or return to the Perplexity for Marketers track. Related: Plans and pricing, Automate Perplexity research with n8n, n8n for Marketers.

#12Common questions

What is Perplexity Computer? Computer is Perplexity's agent for multi-step work. It can research, use connected apps, run code in a sandbox and produce files, reports and dashboards, paid for with credits.

What are Automations? Automations are recurring Computer tasks. They run on a schedule or when an event happens in a connected app such as Slack, Gmail, Outlook, Linear or GitHub, and each run remembers earlier runs.

Can Computer send emails or post for me? With connectors, it can act in connected apps. Keep a human approval step for anything that leaves your company, such as emails, posts or customer messages.