How do I brief an AI marketing agent with a quarterly goal instead of individual tasks?

Brief an AI marketing agent with five things: one measurable outcome with a person who owns it and a sourced baseline, the deadline with checkpoints, the constraints that are fixed, the decisions a person must approve, and how you will review progress. Leave the tasks out. The agent should turn the quarter into a weekly plan where every line has an owner and a finish line, and you judge the plan, not the to-do list.
Why does a task list fail as a brief?
A task list is a plan you already made, handed to someone else to type up. Three things go wrong.
It hides the goal. "Publish eight blog posts" is done when eight posts exist. Whether they moved pipeline is a question nobody asked.
It locks in your assumptions. If the channel mix was wrong, the agent executes the wrong mix on time.
It removes the agent's best skill. A capable agent can re-plan when a channel underperforms, flag a blocker early and route work to the right specialist. A checklist gives it nothing to re-plan against.
A quarterly goal fixes all three, but only if the brief is specific enough to plan against. The template below works with any marketing agent; it is also how MIRA, Nagent's AI Marketing Chief of Staff, expects to be briefed.
What does a good quarterly brief contain?
Use these seven fields. The first four are required; the last three make the quarter reviewable.
Outcome. One number, stated as an end state. "120 sales-accepted demo requests from mid-market SaaS in Q1" is an outcome. "Grow awareness" is not.
Owner and baseline. The person accountable for the number, and where today's figure comes from (CRM report, ad account, analytics). An agent cannot plan a lift against a baseline it was never shown.
Deadline and checkpoints. The end date plus monthly checkpoints with an interim number each.
Approval gates. Every decision a person must make: campaign concept, any live budget change, anything published, any email to a customer list, legal review of claims.
Fixed constraints. Budget ceiling, channels in and out of scope, markets, languages, brand rules, claims that need proof.
What is flexible. Say it explicitly: channel mix, content formats, sequencing, test design. This is the room the agent plans in.
Review rhythm and stop rules. When you read the plan (weekly), what triggers a re-plan (a checkpoint missed by a set margin), and what pauses work (a spend cap hit, a claim rejected).
What does a worked example look like?
Here is a brief a two-person marketing team could give an AI marketing agent for a quarter.
Outcome: 120 sales-accepted demo requests from mid-market SaaS companies (200 to 1,000 staff) in the US and UK by 31 March. Owner and baseline: Head of Marketing. Baseline: 31 last quarter, from the CRM "SAL by source" report. Checkpoints: 30 by 31 January, 70 by 28 February. Approval gates: Head of Marketing approves the campaign concept before any creative is made. Every live ad budget change waits for approval. Every published page and every email to a list waits for approval. Product claims are checked against the proof library before launch. Fixed constraints: Paid budget capped at the agreed quarterly figure. No cold outbound this quarter. English only. Brand guidelines in the knowledge base apply. Flexible: Channel mix between search, paid search and answer engines; content formats; test order. Review: Monday plan review. If a checkpoint is missed by more than 20%, re-plan before spending more.
Notice what is missing: no list of posts, no ad count, no keyword list. Those are the agent's job.
What should the agent send back?
A good agent turns the brief into a plan you can audit in five minutes. Look for:
Lines with owners. Each item says who does it, what done looks like and what it is waiting on.
Numbers with sources. Any baseline or forecast names where it came from. If it cannot, it should say the number is missing rather than guess.
Gates in the right places. Every item that spends, publishes or emails a customer shows the approval it needs.
A first week that is small. Audits, read access and drafts before anything live.
Open questions. A good plan asks for what it lacks, such as access to an account or a missing proof point.
If the plan comes back as a confident list of tasks with percentage lifts and no sources, reject it and ask for the sources. That habit, set in week one, shapes the whole quarter.
How do current tools take a goal instead of tasks?
Most marketing agent products now accept a goal as input. What happens next differs. Checked on 24 October 2026:
HubSpot Campaign Agent (Beta) "turns a goal into a fully connected, multi-channel campaign plan" from your audience, brand settings and past performance (HubSpot). It plans one campaign inside HubSpot.
Salesforce Campaign Agent has you define goals and limits and then creates content and adjusts sends (Salesforce).
Relevance AI has you map a playbook of agents, triggers and handoffs on a canvas (Relevance AI). The goal lives in how you design the flow.
Lindy sets up routines from a sentence, such as a recurring report (Lindy). That suits recurring tasks more than a quarter-long outcome.
The brief template above works with all of them. The difference is whether one agent holds the quarter and routes work to specialists, or whether you hold it and brief each tool separately.
What are the common briefing mistakes?
A goal without a baseline. The agent invents one or plans in vague terms. Fix: attach the report the baseline comes from.
Approval gates left implicit. The agent either stalls or acts without asking. Fix: list every gate by name.
Constraints scattered across chats. Brand rules in one thread, budget in another. Fix: put standing rules in the agent's knowledge base and campaign rules in the brief.
Changing the goal mid-quarter without saying so. The plan drifts and the record stops making sense. Fix: re-brief explicitly and note the date.
Reviewing outputs, not the plan. Approving twenty drafts a week while never reading the plan. Fix: Monday plan review first, cards second.
How does goal-based briefing work on Nagent?
We make Nagent, so read this as our account of our own product. Nagent is multiplayer AI for growth: people and AI agents in one shared context, delivering the number.
On Nagent, you give the quarterly brief to MIRA, the AI Marketing Chief of Staff, in the Marketing team room. Her working rules match the template:
She starts from the target a person set. If none has been set, she says so and plans in counts, never in percentages against a number nobody chose.
Every plan line has an owner, what done looks like and what it is waiting on.
She states a number only when a teammate's report, a connected account or the team record gave it, and says where it came from.
She routes rather than does. Search and answer engine visibility go to DRIS, paid campaigns to NIA, creative strategy to HOOK (Beta) and reputation to Echo (Beta). Writing and visuals go to CREA's Content team, and the result comes back to her thread.
Standing constraints, such as brand voice, products and customer profile, belong in the Knowledge Hub, which every agent reads before it works. Campaign constraints belong in the brief.
Approval gates map onto the Earned Autonomy Ladder. Every agent arrives at L2 Assist, where it drafts on its own and changes to records other people rely on wait. Below L5 Lead, every public post and reply waits for a named person. NIA's live ad changes always queue for a person, and campaigns arrive paused.
What it does not do: MIRA does not set your target, write the copy herself or post to social media (MOXA is coming soon). She also cannot report your overnight activity from her Home panel today; the Actions page and the team room show what ran.
How should you review the quarter week by week?
Monday: read the plan. Check owners, sources and gates. Ask for anything missing.
During the week: decide cards as they come up, in the room or in the Inbox.
Friday: compare what moved with the checkpoint, using the source named in the brief.
Month end: if a checkpoint is missed by your stop-rule margin, re-brief before spending more.
Quarter end: write down what worked and what did not in the team record, so the next quarter starts from it.
For more on why outcomes beat tasks, read where AI productivity tools stop and marketing agents begin. To see who MIRA routes work to, open MIRA's page.
Frequently Asked Questions
What is the difference between briefing an AI agent with a goal and with a task list?
A goal brief says what success looks like, by when, within which limits and with which approvals. A task list says what to do, so the planning decisions are already made and the agent only executes them, right or wrong.
How specific should the goal be?
Specific enough to check on a known date from a known report. A number, a segment, a deadline and the report the number comes from.
Can an AI agent plan a full marketing campaign and assign the work to other AI agents?
Yes. On Nagent, MIRA plans the work and hands each piece to the lead who owns it, and each handoff runs under the receiving teammate's own level and budget, so approvals still apply.
Should I put budgets in the brief?
Yes, as a ceiling. Then keep live budget changes as an approval gate. A planner that can move money without a person is a risk, whatever the brief says.
Do I need technical skills to brief an AI marketing agent?
No. The brief is plain language. What matters is that it is measurable, sourced and explicit about approvals.
Bring your next quarterly goal and watch MIRA plan it: book a demo.