What background agents are
A background agent is an automation you configure once and Spiich then runs on its own, on a schedule or when a CRM record changes. Each one has instructions, a trigger, and permissions you control.
Background agents are how Spiich keeps working while you are in meetings. You describe the job once in plain language, decide when it should fire and what it is allowed to touch, and it runs by itself from then on. Every run is recorded, so you can always see what the agent did and which records it changed.
What a background agent is
An agent is three decisions bundled together:
| Decision | What it means |
|---|---|
| Instructions | What the agent should do, written in plain language as Objective, Process and Output. |
| Trigger | When it fires: on a schedule, when a CRM record changes, when an external service calls its webhook, or on demand against records you pick in a table. |
| Permissions | What it may do with the result: email, post to Slack, spend enrichment credits, or write to the CRM directly instead of staging changes for your review. |
Nothing else is required. There is no scripting, no data mapping and no branching logic to draw. If you can brief a new rep on the task, you can brief an agent on it.
The three parts of the instructions
The instruction box is split into three fields on purpose, because agents follow a clear brief far better than a paragraph.
- Objective answers
What do you want done?. One sentence. For example, research this company and find out who we should be talking to. - Process answers
What rules or steps should it follow?. This is where your constraints go: which regions count, which source to try first, what to skip. - Output answers
What should the result look like?. Say exactly what you want to exist at the end: a CRM record, a note on the deal, an email to you, a Slack message.
A legend under the boxes explains the writing syntax: plain text is kept as written, a [Fill-in field] is filled in for you, and text in (Rule) form tells Spiich how to write something rather than what to write. Agents built before this format keep a single freeform instruction box instead, and they still run normally.
Where agents live
Open Background Agents from the left sidebar. Settings > Background Agents lands on the same page.
The list gives you one row per agent with a plain-English sentence of when it fires, such as "When a Deal is updated and the Stage field changes, run the agent" or "Every weekday at 5:00 AM, run the agent". Each row also shows a live or paused dot, its Cost per run, its Created date and its Next run. Below the list, Usage History and Analytics itemise and chart every credit charge agents have made.
The sidebar has its own collapsible Background Agents section. Agents with a run in flight show a spinner, hovering one opens its five most recent runs, and a badge counts any CRM writes waiting for your review.
Who can create one
Creating, editing and deleting agents is a workspace decision. Workspace admins can do it by default, and an admin can grant the same right to an individual member or to a whole team.
| Action | Who can do it |
|---|---|
| View agents and read run history | Everyone |
| Run an agent on demand, where the agent allows manual runs | Everyone |
| Approve or dismiss CRM changes an agent staged | Everyone |
| Create, edit, pause or delete an agent | Admins, plus members and teams an admin has granted access |
| Point an agent at another member so it runs as them | Admins only |
Every member can open the page, read an agent, review its run history and approve staged changes. Without the grant, the controls that create, edit, pause or delete an agent are not offered. Ask an admin to switch it on from Settings > User Management, or for a whole team from Settings > Team Management. See Manage your team.
What an agent can do during a run
Inside a run an agent has the same reach a rep has in Chat. It can read and filter your CRM records, search your emails, calendar events and meeting summaries, read meeting transcripts, search the web, look for prospects and companies, create and update CRM records, write notes, manage tasks and fill Spiich tables.
Sending email, posting to Slack and enriching contacts happen only when you have switched that permission on for that specific agent. See What an agent is allowed to do.
Agents Spiich suggests
Spiich occasionally proposes an agent based on patterns in your workspace. It appears at the top of the list marked Proposed with a Review link. The pitch page says why it was suggested, summarises what one run would do, states the cost, and lists its guardrails, for example "Does not send email". Nothing runs until you choose Activate, Adjust or Not for us.
What agents do not do
- Agents do not appear for post-meeting work. Meeting notes and CRM updates after a call are handled by the meeting pipeline and are deliberately kept out of this list. See Meeting notes and CRM updates.
- An agent does not run without credits. If the workspace runs out mid-run, the agent is paused automatically and workspace admins are emailed.
- Deleting an agent is permanent and cannot be undone.
If something goes wrong
Every agent stopped at once and admins got an email. The workspace ran out of credits, so the agents were paused rather than left firing. Top up in Settings > Billing, then resume each agent with the play button. Set a Low credit warning at the bottom of the Background Agents page so admins hear about it earlier next time.
The Create new button is missing. The workspace has no agents yet, so the template library has taken over the whole page. Pick a template card directly to create your first one.
An old `Settings > background-agents` link. It redirects to the top-level Background Agents area, so old bookmarks still work.