Guides
Morning brief to Slack
Somebody reads the overnight noise every morning and tells everyone else what matters. It takes forty minutes and it's the same forty minutes on Monday as on Thursday.
This guide moves that job to an agent and puts the result in Slack at 7am, so the first person in reads the brief instead of writing it.
It's the natural second thing to build. The quickstart gets Scout writing a decent brief by hand. This adds Slack, and then a routine so nobody has to ask for it.
What you'll build
- Scout writing a brief you'd actually read, on request.
- One Slack tool so Scout can post it.
- A routine that repeats the request at 7:00 on weekdays.
Twenty minutes, assuming Scout already exists.
Before you start
- Slack connected under Library > Integrations, with the bot able to post
to your channel. Ours is
#service-desk. - An agent named Scout. The quickstart creates one.
- Permission to run Slack actions in this organization.
1. Get the brief right first
Before you touch Slack or the schedule, ask Scout for the brief by hand and read what comes back:
Write this morning's brief for the Northwind IT service desk. Cover what came
in overnight, anything worth watching today, and the next three things somebody
should pick up. Under 200 words.This is worth doing properly, because a brief nobody reads is worse than no brief. Ask yourself, honestly, whether you'd read this at 7am. Usually the first version is too long and too hedged. Tell Scout to be shorter and more definite, and ask again in the same conversation.
Only once you'd actually read it should you put it in a channel.
2. Let Scout post to Slack
- Open Build > Agents and open Scout.
- In the Tools section, add
slack.message.post. - Save.
Back in the conversation, ask for the same brief and add: post it to #service-desk. Scout writes the brief and posts it in one turn.
Check the channel. If nothing arrived, it's almost always Slack rather than Mobius: the integration isn't connected, or the bot hasn't been invited to the channel. Scout's reply will quote what Slack actually said, which is faster than guessing.
3. Ask for it every weekday
Still in the same conversation:
Do this every weekday at 7am: write the overnight brief and post it to
#service-desk. Don't reply to me separately, the channel is enough.Scout answers with a proposal card, not a schedule. It states the instructions it will repeat, when it fires and in which timezone, who can see each occurrence, which connections it may act through, and the spending ceilings. It also shows the most this could cost over the next 30 days.
Read the instructions line on the card carefully. It is what actually runs, and it is the agent's summary of what you asked for, not your words verbatim. If it lost something, say so in the conversation and Scout proposes again.
Click Approve routine. Nothing is scheduled until you do, and only a person can approve one: an agent cannot put its own work on a schedule.
4. Read it the next morning
Each occurrence is an ordinary turn in the same conversation. Open the session and you'll see the 7am instruction, Scout's Slack call, and its reply, sitting directly under the requests you made by hand.
That is the point of a routine returning to the conversation. When Monday's brief is wrong, you ask about it in the same place, and Scout can see the brief you're asking about.
Variations
Keep it out of Slack at first. Approve the routine without the Slack tool and read the brief in the conversation for a week. Add the posting once the wording is right, so your team's first impression isn't a bad draft.
Let Slack do the scheduling. Give Scout slack.message.schedule if you'd
rather it write the brief earlier and have Slack post it at 7am.
Add real data. Grant Scout the tools that reach whatever the brief should cover: alerts, ticket counts, a spreadsheet of backup results. An agent writes better when it's summarizing something specific rather than speaking generally.
Get it approved first. For a brief that goes to a client rather than your own team, tell Scout to get the draft approved with an interaction before it posts.
One per client. Start a conversation per client and approve a routine in each, pointing at that client's channel.
When it stops arriving
- Was there an occurrence? Open the conversation. A
skippedoccurrence means the previous one was still running when this one was due; afailedone leaves its error in the transcript. - Is the routine paused? Pausing keeps the schedule, so resuming does not
mean rebuilding it. Routines are managed through the API and CLI today,
under
mobius routines. - Did it hit a ceiling? The daily ceiling stops further occurrences once the rolling 24-hour spend reaches the limit.
- Did the owner leave? Mobius pauses a person's routines when they leave the organization rather than letting them keep acting on that person's behalf.
Next
- Understand what you approved: routines.
- Set proper limits: guardrails.
- Handle work arriving from outside with triage inbound tickets.
- Add a human check with interactions.