Using the AI Copilot: Propose, Confirm, Execute

The AI Copilot is a conversational assistant that lives in a dock on every screen in CampaignSwift. You ask it things in plain language; it reads your live data and gets to work.

What you can ask

Anything about your agency or a client. For example:

• “What’s pending approval for Acme?”
• “Draft next week’s posts for this client.”
• “Which clients are at risk?”

Copilot is always in context: it reads your real analytics, approvals, content, inbox, and client outcomes, so answers are grounded in your actual accounts, not generic advice.

You can scope a conversation to a single client or ask across your whole book of business: one client’s pending approvals, or every at-risk client agency-wide.

Propose, confirm, execute

The core rule: anything that changes something is proposed first, and only runs after you confirm.

• Reads (questions, summaries, lookups) are answered directly.
• Writes (creating posts, scheduling, assigning tasks, inviting teammates) are gated. Copilot shows you what it intends to do, you click confirm, and only then does it execute.

Nothing ships behind your back. If you don’t confirm a proposal, nothing happens.

What Copilot can do

Copilot does the work, not just chat. Among the things it can create and run (each gated behind your confirmation):

• Create and schedule posts
• Build tasks and campaigns
• Plan content onto the calendar
• Generate reports

Copilot does what you ask, on demand. For the recurring work you’d rather not do at all (chasing approvals, triaging the inbox, recovering failed posts), CampaignSwift has named agents that run on your rules. See the agents overview and the per-agent docs in this help center, starting with Approval Shepherd.

Every action is auditable

Every action Copilot takes is logged in an activity ledger, so you can always see what was done, when, and at whose confirmation. The same ledger covers agent actions. See AI Credits and Budget Guardrails for how the audit trail and cost controls work together.

Cost

AI usage in CampaignSwift is shown as simple Credits, with per-agent and per-client breakdowns, a monthly budget cap, and an optional hard cutoff. Copilot usage is part of that same system. See AI Credits and Budget Guardrails for details.

Tips for good results

• Name the client when you want client-specific work (“draft next week’s posts for Acme”), or ask agency-wide questions without one.
• Review proposals before confirming: the propose step exists so you can catch anything off before it runs.
• Start with reads if you’re new: ask about pending approvals or at-risk clients to see how grounded the answers are, then move to letting it draft and schedule.