Your challenge this week
You have learned what makes an agent, how to design one, and how the build process works in Relay. Now it is time to do it yourself. Your goal: build one working agent in Relay with at least one skill that runs on a trigger. This is not a hypothetical exercise. By the end, you should have an agent that can run a test workflow and produce a real output (an email, a document, a report) that you can review and improve.Challenge 1: Pick Your Task
Challenge 1: Pick Your Task
Choose a task from your actual work that fits the agent criteria. It should be:Not sure where to start? Here are some ideas based on what your cohort discussed in the session:Competitive analysis report: Scan competitor LinkedIn profiles and company pages for recent activity. Summarise product updates, customer wins, and key messaging themes into a report. Email it to yourself every Monday at 8am. Connect LinkedIn and Gmail.Meeting preparation briefing: Research upcoming meeting attendees, pull related documents and recent email threads, and compile a one-page briefing. Trigger it 24 hours before each meeting in your calendar. Connect Google Calendar, Gmail, and Google Drive.Weekly project status digest: Pull updates from email threads, Slack channels, and project tools. Compile a summary of where things stand across active projects. Run it every Friday at 3pm. Connect Slack, Gmail, and your project management tool.Document filing and organisation: When a signed contract arrives from DocuSign, automatically file it in the correct Google Drive folder and rename it to match your naming convention. Trigger it when an email from DocuSign with “Completed” in the subject lands in your inbox. Connect Gmail and Google Drive.
Repeatable
Something you do regularly: weekly, daily, or every time a specific event happens
Definable
Something you can describe in clear steps with a specific trigger and output
Low risk to start
Something where a mistake does not cause real damage. You can add higher-stakes tasks later
Clear output
Something with a tangible deliverable (an email, a report, a filed document) that you can review
Challenge 2: Write Your Agent Job Description
Challenge 2: Write Your Agent Job Description
Before you open Relay, write down:Job title: What would you call this role?Purpose: One sentence on what this agent does and why.Responsibility (What / When / How):
Tools needed: Which apps will you connect? (Gmail, Google Drive, LinkedIn, Slack, Google Calendar, etc.)
| Your agent | |
|---|---|
| What | What does it deliver? What does the output look like? |
| When | What triggers it? A schedule? An event? |
| How | What tools does it use? What steps does it follow? |
| Who reviews | Who checks the output before it goes live? What level of oversight does this need? |
Challenge 3: Build It in Relay
Challenge 3: Build It in Relay
Open Relay and follow this sequence:
Create a new agent
Give it the job title and description from your job description above. Select the tools it needs from the connection menu.
Add your first skill
Pick from the suggestions or write your own prompt. Be specific: include the what, when, and how. If Relay asks clarification questions, give it the most detailed answers you can.
Review the flowchart
Check the generated flowchart. Does the trigger look right? Are the steps in the correct order? Does the output go where you want it?
Challenge 4: Reflect and Refine
Challenge 4: Reflect and Refine
After your first test cycle, ask yourself:Did the output match what you expected? If not, was the issue in your brief or in the agent’s execution?What feedback did you give? Was it specific enough to produce a better result?Would you trust this to run automatically? If not, what would need to change before you would?What is the next skill you would add to this agent? Think about what other responsibilities could sit under the same role.
Challenge 5: Stretch Goal
Challenge 5: Stretch Goal
If your first agent is running well and you have time, try one of these:Add a second skill to the same agent. Give it another responsibility that uses the same tools but runs on a different trigger or schedule.Add a human-in-the-loop step. Insert a review point in the workflow where the agent pauses and waits for your approval before taking the final action (sending an email, filing a document).Share your agent design with a colleague. Walk them through the job description and the flowchart. See if they can understand what the agent does without you explaining it. That is the test of a well-designed brief.
As shared in the live session, building the first flowchart is about 10 percent of the work. The real value comes from testing, refining, and expanding over time. Treat this as the foundation of an agent you will keep improving, not a one-off exercise.