Activity log
Every automation run is recorded with a full audit trail.View recent runs
- Open Automations for your site
- Click an automation to open its detail sheet
- Click Activity (history icon) to open the activity panel
- Status — completed, failed, or running
- Summary — natural-language description of what the automation did
- Action count — how many real actions were taken (articles created, links inserted, etc.)
- Timestamps — start and finish time
Tool-call drilldown
Click a run to expand every tool call:- Tool name and input parameters
- Output (truncated for large responses)
- Latency per call
Notifications
Configure notifications when creating or editing an automation.- Toggle Email notifications on
- Select team members as recipients
- Add optional context — extra instructions appended to the notification (e.g. “Flag anything about competitor X”)
Slack
- Connect Slack for your site at Integrations → Slack (
app.trysight.ai/integrations/slack) - Invite the Sight AI bot to the channel you want notifications in
- In the automation’s notification settings, enable Slack and pick a channel
@mention the Sight AI bot in Slack threads to ask questions about your site (read-oriented; mutations still require confirmation in chat).
Slack must be connected per site. Switch sites before connecting if you manage multiple properties.
How scheduling works
- Schedules are stored as cron expressions in UTC
- A platform cron evaluates due automations every 15 minutes
- Each automation fires at most once per 15-minute evaluation window
- Schedules whose minute doesn’t land exactly on a tick boundary (e.g.
5 * * * *) still fire once per window
schedule_cron empty) never run on the cron — trigger them with Run now or via Agent chat.
End dates
Set an optional end date to auto-disable an automation after a campaign. Past end dates are skipped and the automation is turned off automatically.Common issues
Automation never runs
Automation never runs
Check these in order:
- Enabled — toggle is on
- Schedule — not set to Manual only (unless you’re triggering manually)
- End date — not in the past
- Plan — site is on Pro or Advanced with AI credits remaining
- Site license — site is licensed on your team plan
- Activity log — look for failed runs with error messages
Run stuck in 'running' status
Run stuck in 'running' status
Sight AI enforces one in-flight run per automation. If a run crashes without finishing, a 30-minute janitor automatically marks stale
running rows as failed and releases the lock. Wait up to 30 minutes, then trigger a new run or check the activity log for partial tool calls.Agent created no articles / took no actions
Agent created no articles / took no actions
Usually one of:
- Empty queue — no eligible opportunities (connect GSC, wait for opportunity sync)
- Score threshold — opportunities below the agent’s threshold were skipped (check activity log for skip reasons)
- Dedup — articles already exist for those keywords
- Action budget exhausted — previous rows in the same run used the full budget
Articles created but not on my site
Articles created but not on my site
Automations create articles in Sight AI first. They sync to your CMS when:
- A CMS integration is connected and active
- The agent calls
sync_article(template agents do this for refreshes; new articles may need Planner scheduling depending on your settings)
Outreach Agent not sending
Outreach Agent not sending
Confirm:
- Advanced plan active
- Outbound email domain verified in Outreach settings
- Outreach Opportunities have eligible rows
- Activity log shows email lookup and send tool calls (not skips)
Slack notifications not arriving
Slack notifications not arriving
- Slack connected at
/integrations/slackfor this site - Bot invited to the selected channel
- Slack notifications enabled on the automation with a channel selected
- Run completed successfully (failed runs still notify if configured)
Custom automation hit tool step limit
Custom automation hit tool step limit
Custom automations cap at 15 tool steps per run. Split work across multiple automations, reduce scope in instructions (e.g. “process at most 3 items”), or switch to a template agent with a dedicated engine.
Credit usage
Every automation run consumes AI credits from your team’s pool — the same credits used for article generation, AI visibility checks, and agent chat. Credit cost varies by:- Number of LLM calls (custom automations)
- Number of actions taken (template agents — article generation is the largest cost)
- Tools invoked (research, image generation, outreach drafts)
Best practices for monitoring
- Review activity logs weekly when first enabling an automation — drop to monthly once you trust it
- Start with low action budgets and increase after validating output quality
- Enable email notifications for the first two weeks, then switch to Slack for ongoing visibility
- Pair Search + Boost agents and watch both logs — Boost should target existing URLs, Search should create new ones
Next steps
- Automations overview
- Getting started
- Agent templates
- FAQs — general automation questions