Build a Custom Agent that delivers weekly market insights
Learn how to build a Custom Agent that delivers a weekly brief of what changed in your market, what's at risk, and what to do about it.
A Custom Agent that scans the sites you trust each week for competitor launches, pricing shifts, and category news, saves a full brief to your Market Briefs database, and posts a short summary to Slack. The team gets a regular read on what's changing in your market.
Who this is for
Product marketing, product, and strategy teams who own a recurring competitive or market update.
Tehtävä
Try this prompt
Use this example to get started, and feel free to adapt it to your team and market:
I need a weekly market brief that tracks competitor launches, pricing changes, and important category news so we can spot anything that might impact our next launch.
In this step, you'll prompt the agent to draft an initial setup, including Instructions, Triggers, and the Tools and access it needs. You'll fine-tune each of these in the next steps.
Open
Agentsfrom your sidebar and clickNew agent.In the chat, describe what you want the agent to do. You can use the Try this prompt callout above as a starting point.
The agent will draft
Instructions, suggestTriggers, and flag theTools and accessit needs.Answer any quick follow-ups, like which database or Slack channel to connect.
Helpful tips
Be specific in your first prompt: Tell the agent the cadence (weekly), the sources (competitor pages, category news), and where the brief should live (Market Briefs database). The more concrete you are upfront, the less back-and-forth later.
@-mention pages and databases instead of describing them: @-mention your Competitors page and Market Briefs database so the agent picks up the right context without guessing.
Iterate in chat, don't configure manually: It's faster to refine in conversation than to edit each setting by hand. Save the manual tweaks for the final polish.
The agent runs on its own each week, so it needs an explicit list of sites it's allowed to scan. Trusted URLs is how you set that scope.
Open
Instructionsin your agent and scroll toTrusted URLs.Add the sites you want the agent to scan each week, such as:
Competitor news pages and product blogs.
Industry publications relevant to your market.
Any data providers or analyst sites you already trust.
In
Tools and access, you can also enableWeb Searchso the agent can scan across the web.
Helpful tips
List every site upfront: If the agent hits a URL that isn't on the list, it'll pause and wait for approval, which defeats the point of a scheduled run.
Point to the specific section, not the whole domain: Instead of approving a whole domain, point to the section you actually need (like the "markets" or "news" path).
For the agent to run on its own each week and publish where the team will see it, you need to set the schedule and grant access to a few sources.
Set the weekly trigger: Open
Triggersand confirm there's a Monday 9:00 AM scheduled run that creates a new page in Market Briefs.Connect Slack: In
Tools and access, connect your channel (e.g.,#all-competition-market-intel) with
writeaccess so the agent can post the summary.Give read access to context sources: Still in
Tools and access, give the agent read access to anything that helps ground its takeaways, like competitive one-pagers or Gong transcripts.Hit
Save.
Helpful tips
Spell the Slack channel name exactly as it appears in Slack: Any mismatch and the agent won't post until you confirm the channel.
Read access is enough for internal sources: The agent only needs to reference them for context, not edit them.
Approve new tool connections inline: If a tool isn't connected yet, the agent will prompt you for it. Approve and keep going.
In this step, you'll set guardrails on what the agent pulls and what it does when something is uncertain.
In the chat, give the agent three guardrails:
Recency: Only look at updates from the last 7 days.
Volume: Cap summaries at 2 to 3 items per competitor.
Sourcing: Only include claims it can link to. If a source is missing or unclear, flag it instead of guessing.
When prompted, click
Show changesto review what the agent updated inInstructions, then hitSave.
Helpful tips
Tell the agent what to skip, not just what to include: Negative rules like "skip op-eds and rumor-stage reporting" tighten briefs faster than positive ones do.
Anchor the agent with one example: Tell the agent: "For each competitor, a good item looks like: [Competitor] launched X on [date]: link". One concrete example tends to lock in the right specificity faster than three abstract rules.
Tune scope based on how your team uses the brief: If you're missing important updates, widen the window. If briefs feel noisy, tighten the cap.
A good weekly brief is consistent. In this step, you'll define the sections your agent uses so every brief is easy to scan and share.
In the chat, tell the agent to format each brief with these sections:
Highlights: 3 to 5 bullets covering the most important changes.
Risks: What could impact your positioning or roadmap.
Opportunities: Openings the team could act on.
Positioning tweaks: Messaging or narrative adjustments to consider.
Recommended actions: Concrete next steps for Product and Marketing.
Sources: Links to every claim made in the brief.
When prompted, click
Show changesto review what the agent updated inInstructions, then hitSave.
Helpful tips
Keep section names short and predictable: The same headings every week train your team to scan in the same order, which is half the value of a recurring brief.
Tell the agent to skip empty sections: Some weeks won't have meaningful risks or opportunities. Asking the agent to omit (rather than fill) empty sections keeps the brief honest.
Mirror the Slack summary against the headings: If the brief leads with Highlights and Risks, the 2 to 3 items in the Slack post should pull from those too.
Before the first scheduled run, kick off the agent manually. A test pass surfaces the small things, like wrong sources, missing links, or a Slack channel that didn't connect, while there's still time to fix them.
Open
Triggers, clickRun agent, and select the Monday 9:00 AM scheduled run.Review the brief the agent creates and confirm:
Section headings match what you set in the previous step.
Takeaways are specific (a named competitor + dated event), not generic ("the market shifted").
Every claim links back to a source.
Check #all-competition-market-intel in Slack to confirm the summary posted with 2 to 3 highlights and a link to the full brief.
Helpful tips
Check
Activityfirst if anything looks off: It shows each run, what happened, and whether any step failed. You can ask the agent to troubleshoot with you directly from there.If the Slack post didn't go through, the channel name is usually the issue: Paste the exact channel name into the chat and ask the agent to try again.
Run a second test after any fix: Tweaking instructions can have ripple effects, so a quick re-run is the fastest way to confirm the change held.
Once the agent is running reliably, share it with the team and put a light maintenance routine in place so the brief stays useful as your market shifts.
Set expectations with the team: Tell stakeholders where the briefs live (the Market Briefs database), when they publish (Monday at 9:00 AM), and what the Slack post includes.
Be intentional about sharing: Sharing the agent also shares what it has access to. Double-check the sources and integrations look right for the audience before adding new collaborators.
Test after any meaningful change: If you add or remove sources, adjust scope, or change formatting, run a manual test before the next scheduled run.
Keep your source list focused: A smaller, curated set of trusted URLs tends to produce more relevant briefs than a long, noisy one.
More resources
For tips on keeping your agent sharp over time, check out our Custom Agent best practices guide.
For a deeper dive into Custom Agents, take our course on Notion Academy.
Emmekö vastanneet kysymykseesi?