How to read these examples
Each scenario follows the same shape:
- The setup — who’s doing the work and what they need.
- The brief — what they actually type into the agent.
- The sources — what they attach or connect.
- What comes back — the shape of the output.
- The refinement — the one revision most people make to finish it.
Copy any of these. Adapt them to your team. They’re starting points, not rules.
1 — Weekly product update for leadership
The setup
A product manager writes the same Monday update every week. Twenty minutes that adds up to eight hours a quarter.
The brief
Outcome: A two-paragraph internal update for our product leadership team. Audience: VP of Product and engineering leads. Constraints: Plain tone. Under 250 words. Skip launch dates that aren’t confirmed. Include one risk callout at the bottom. Sources: Last week’s standup notes (attached) and Thursday’s customer call transcript (attached).
What comes back
A short update that names what shipped, what slipped, what the team learned from the customer call, and one risk worth tracking. Already in the format the leadership team expects.
The refinement
“Make the risk callout one sentence instead of two.”
Time saved per run: ~15 minutes. Saved across the quarter: ~3 hours.
2 — Backlog refinement on a stretch story
The setup
An engineering lead has a vague backlog item: “improve onboarding.” They need it broken down into clear acceptance criteria before sprint planning.
The brief
Outcome: A refined version of the attached backlog item with clear acceptance criteria. Audience: The engineering team in sprint planning. Constraints: Use our standard “Given/When/Then” format. Keep the story title short. Add three to five acceptance criteria. Flag anything that depends on another team. Sources: The current backlog item (pasted in), the related product brief from Confluence (linked).
What comes back
A rewritten story with a tighter title, three to five Given/When/Then acceptance criteria, and one flagged dependency on the design team.
The refinement
“Add a fourth criterion about the email confirmation step.”
Time saved per run: ~25 minutes. Removes a planning meeting blocker.
3 — Competitor one-pager from a fresh URL
The setup
A go-to-market lead spots a new competitor. The team needs a quick positioning brief by tomorrow’s standup.
The brief
Outcome: A one-page competitive brief covering positioning, target audience, claimed differentiators, and the gaps we can press on. Audience: Our GTM team — head of sales, head of marketing, two product leads. Constraints: One page. Plain tone. End with a three-bullet “where we can win.” Avoid speculation about their internals. Sources: Their homepage URL and pricing page URL. Our positioning doc (attached).
What comes back
A structured one-pager with their headline claims, target audience signals, three differentiators they emphasize, and three angles where the team’s positioning is sharper.
The refinement
“Tighten the ‘where we can win’ section to the two strongest angles.”
Time saved per run: ~90 minutes. Ready for tomorrow’s standup.
4 — Customer follow-up email from a long call
The setup
A customer success lead just finished a 45-minute call. They need a polished follow-up note that lists what was decided, what’s open, and what the customer should expect next.
The brief
Outcome: A customer follow-up email recapping the call, listing decisions, open questions, and next steps. Audience: Our main contact at the customer (a senior engineering manager). Constraints: Friendly but direct. Three short sections: “What we covered”, “What we decided”, “What’s next”. Include a clear date for the next check-in. Sources: The call transcript (attached). My one-line note: “they want migration support before end of Q3.”
What comes back
A short, well-structured email the lead can send after a single read-through. Tone reads like a thoughtful human wrote it.
The refinement
“Move the migration deadline into the ‘What we decided’ section.”
Time saved per run: ~20 minutes. Turns a dreaded post-call task into a five-minute review.
5 — Multi-doc analysis to support a decision
The setup
An operations lead is evaluating two vendors. They have a 30-page RFP response from each. The exec team meets in three days.
The brief
Outcome: A comparison of the two attached vendor responses across five dimensions: pricing, security posture, integration footprint, support model, and roadmap commitments. Audience: Our exec team — CFO, COO, head of security. Constraints: A side-by-side table, then a one-paragraph recommendation. Flag any claim a vendor makes that isn’t supported in the source document. Sources: The two vendor RFP responses (attached).
What comes back
A side-by-side comparison table across five dimensions, a one-paragraph recommendation, and three flagged claims that lacked support in the docs.
The refinement
“Add a row to the table for ‘data residency’ — both responses mention it.”
Time saved per run: ~4 hours. The exec team gets the recommendation they need on time.
6 — Research brief on a new topic
The setup
A strategy lead needs to understand a topic — say, “AI governance frameworks for financial services” — well enough to brief the team in 48 hours.
The brief
Outcome: A research brief on AI governance frameworks relevant to financial services. Audience: Our internal AI strategy group — five people with mixed technical backgrounds. Constraints: Eight to twelve pages. Four sections: landscape, key frameworks, regulatory drivers, recommended posture for our company. Plain prose, not slides. Include source links where possible. Sources: Our internal AI policy draft (attached). Three industry articles I’ve pasted in. Any public regulator guidance the agent can confirm.
What comes back
A structured brief with the four sections, a clear landscape view, named frameworks, the regulatory drivers behind them, and a recommended posture tailored to the company’s current position. With source links.
The refinement
“Add a one-paragraph executive summary at the top.”
Time saved per run: ~12 hours. Turns a week’s research into an afternoon’s review.
7 — Compliance pack with Model Card + Transparency Notice
The agent
Two of the compliance-oriented library agents — Model Card Writer and Transparency Notice Writer.
The setup
A product manager is shipping a new AI-powered feature. Their compliance team needs two artifacts before launch: a model card describing the AI used, and a user-facing transparency notice explaining how the feature works.
The brief
Outcome: A draft model card and a draft transparency notice, both ready for compliance review.
Audience: Model card is internal (compliance team). Transparency notice is external (end users).
Constraints: Aligned with EU AI Act guidance. No marketing language. Cite sources for any factual claim.
Sources: The feature’s PRD (attached). The model card for the underlying model (attached). The team’s existing transparency notice (used for tone and structure).
What comes back
Two clean drafts in the team’s expected formats — the model card with sections for purpose, training data, limitations, and risks; the transparency notice with plain-language explanations of what the AI does and what users can control.
The refinement
“Tighten the transparency notice — the limitations section is too long for an end-user audience.”
Time saved per run: ~4 hours. Compliance docs that used to take a week of drafting and review land in a single afternoon.
8 — Sales enablement workflow
The agent
Sales Enablement from the agent library, customized with the team’s product positioning attached as knowledge.
The setup
A go-to-market lead is preparing for a product launch. They need a talk track, a battle card against the main competitor, and a one-page playbook — all in the team’s voice.
The brief
Outcome: Three sales artifacts — a talk track, a battle card, a one-page playbook.
Audience: Account executives at our company. They’re technical but not deep in the new feature.
Constraints: Use the team’s existing positioning language. Don’t make claims we can’t back up.
Sources: The product spec (attached). The competitor’s public docs (attached). Three customer-call transcripts where this product type came up (attached).
What comes back
Three artifacts in distinct formats — a conversational talk track, a structured battle card with “what they say” / “what we say” pairs, and a tight one-page playbook with discovery questions, common objections, and a closing recommendation.
The refinement
“Add a ‘what to do if the prospect mentions price first’ section to the playbook.”
Time saved per run: ~6 hours. Three artifacts that would have been three separate drafting sessions land together.
9 — Visual deck mockup for an internal proposal
The agent
The HTML Mockup Generator from the agent library, paired with Chart Generator for the data slides.
The setup
A team lead is pitching an internal initiative. They have the argument and the data, but they want a polished one-screen layout that shows what the rollout would look like — without spending a day in a design tool.
The brief
Outcome: A clickable HTML mockup of a single-screen proposal layout. Include a hero section, three feature panels, two charts, and a call-to-action.
Audience: Senior leadership reviewing the proposal next week.
Constraints: Clean, dense, professional. No marketing language. Use the team’s brand palette (described in attached doc).
Sources: The proposal outline (attached). The team’s brand guidelines (attached). Two CSV files with the data for the charts (attached).
What comes back
A complete HTML mockup with a polished hero, three feature panels, two clean charts in the right places, and a clear call-to-action. The layout reads as something a designer might have built.
The refinement
“Move the most striking chart up into the hero section as a small inline visual.”
Time saved per run: ~5 hours. A proposal that would normally need a designer’s time becomes a self-serve deliverable.
Patterns across the examples
Every one of these scenarios has the same DNA:
- One specific outcome. Not “write something about X” — a concrete deliverable.
- A named audience. The tone changes when the audience does.
- Hard constraints. Length, format, what to exclude.
- Real sources. Files, transcripts, connected apps — never just topic words.
- One refinement to finish. Almost always one small change, not a rewrite.
When you brief an agent this way, the time savings compound. The first time, you save twenty minutes. By the tenth time, you’ve built a team-wide habit that saves hundreds of hours a quarter.
Want more patterns?
- Tips & best practices — the patterns top teams use.
- Working with assistants — the full brief-context-output loop.
- Agent library — the pre-built agents used in scenarios 7–9.
- Creating your own agent — once the library doesn’t quite fit.
- Visual agents — go deeper on Image, Chart, and HTML Mockup.
- Getting started — start with a 10-minute first run if you haven’t yet.