Automations: schedule an autonomous agent.

Until now, Morrison did the work when you asked. You opened Chat, ran a Skill or a workflow, read the answer. The agent was good at the work - it just waited for you to start it. Automations remove that last step. Write the instructions once, give them a schedule, and Morrison runs them on its own - then brings you the result.

It’s the same agent you already use in Content Chat, with the same crawl, the same integrations and the same Skills. The only difference is that nobody has to be in the room when it runs. You can find it in the new Automations tab in the sidebar.

What an automation is

An automation is a saved prompt that runs itself. Three things define it:

• Instructions. What you want done, written the same way you’d ask in Chat - “compare this week’s Search Console numbers to last week and flag any page that lost real traffic.” You pick the model and which tools and Skills are in play, exactly like a chat turn.
• A schedule. When it should run - every hour, every day at a set time, or a chosen weekday - in your own timezone. Or no schedule at all: leave it on manual and run it with a button when you want.
• What happens to the output. Keep it in the run history, or have it emailed to whoever needs to read it.

That’s the whole model. If you can ask for it in Chat, you can put it on a schedule.

Set a schedule in seconds

The trigger builder keeps it to the three rhythms that actually matter: every hour, every day at a time you choose, or every week on a chosen day. It reads back in plain language - “Every Monday at 07:00” - and runs against your local timezone, so an 8am report lands at 8am where you are.

New automations start turned off. You build it, run it once by hand to see the output, and only flip it on when you’re happy. Nothing fires by surprise.

It runs the full agent - not a cron script

This is the part that makes it Morrison and not a reminder. When an automation fires, it runs the exact same agent as Content Chat: the same system prompt, the same tools, the same models, the same Skills. It reads your crawled pages and pulls live data from Search Console, GA4, SerpAPI and Ahrefs, chaining as many steps as the job needs to reach a grounded answer.

So an automation isn’t limited to “send me a metric.” It can diagnose a ranking drop, check who overtook you on the SERP, read the page that slipped and tell you the one fix most likely to win it back - the same depth you’d get from asking in Chat, arriving while you sleep.

Get the result where you want it

Every run produces a Markdown report. Two ways to receive it today:

• Just output. The run finishes and the result waits in the run history. Good for things you check on your own cadence, or automations whose job is to keep a log.
• Send email. The output is delivered to one or more recipients, in a clean branded email. A Monday-morning digest can land in your team’s inboxes before standup, or go straight to the stakeholders who asked for it.

Webhook delivery is next on the list. The action model is built to take more channels without changing anything about how you write the automation.

Every run on the record

Scheduling work you don’t watch only works if you can trust it, so every run is fully accounted for. Each one records its status - queued, running, success, error or cancelled - with the time it took and the credits it cost. Open any run and you get the full activity trace: every reasoning step and every tool call the agent made to get to its answer, the same view you see in Chat.

There’s a per-automation history and an organization-wide view across everything you’ve scheduled, so you can see at a glance what ran overnight and what needs a look. A run you started by mistake, or one taking longer than expected, can be cancelled mid-flight.

Built on the Skills you already use

You don’t have to write every automation from a blank prompt. An automation can run Morrison’s pre-defined Skills - the same playbooks you already use in Chat for audits, briefs, decline diagnoses and governance checks. Point an automation at the Skill that fits the job, set the cadence, and it runs that playbook on its own. A few of the things teams put on a schedule:

• A weekly Search Console digest - top movers, winners and losers, week over week.
• A daily ranking-drop early warning - get pinged the moment a money page slips, with who overtook you and the highest-impact fix.
• A weekly striking-distance list - the queries one nudge away from page one, each mapped to a page and a specific change.
• A content-decay and cannibalization check - pages quietly losing traffic, and URLs competing with each other, each with a refresh / merge / redirect call.
• A weekly leadership update - a plain-English read on traffic, wins, losses and the next three actions, written for the reader rather than the dashboard.

Pick the Skill or write the prompt, set the schedule, and you have a working automation in under a minute.

What it costs

An automation run costs credits exactly like the same work in Chat - by the model you chose and the work it did. A heavier model on an hourly schedule costs more than a light model once a week; you set the frequency, so you control the spend. Every run’s credit cost is recorded in its history, so there are no surprises.

Try it

Automations are live now for every workspace on the beta - look for Automations in the sidebar. Point one at a Skill that matches something already on your to-do list - or write your own instructions - run it once to see the output, then give it a schedule and switch it on.

Then tell us what you put on a schedule, and what you wish it could do next, at hello@morrison.app. The more we see how teams use this, the faster the next channels and triggers land.