The hardest part of adopting an AI agent is the first day. The phone has a new contact. You sent it "hey." It said "hey" back. Now what? Most users bounce off in the first hour because they cannot think of what to say next; the previous AI products they have tried were chat surfaces, and a chat surface is good for asking long questions, not for the small operational asks that make a personal agent earn its place.
The fastest way through that bounce is a starter list: ten specific small tasks that demonstrate the agent's core capabilities. Each one is 30 seconds. By the end of the first day, you have used the agent ten times, and you have a sense of what it can do. By the end of the first week, you stop thinking about it.
This post is the starter list, in the order we recommend. Each entry includes what to text, what the agent does, and what you learn.
1. Set up the morning brief
Text: "Send me a morning brief every weekday at 7am on WhatsApp."
What happens: The agent confirms the schedule and explains what the brief will include. Tomorrow morning at 7am, you get a paragraph covering today's calendar, urgent inbox, tasks due, weather, and any anomalies.
What you learn: Scheduled tasks. The agent does not just respond to your messages; it can fire on its own schedule.
2. Save a core memory
Text: "Save this: my wife is allergic to peanuts. My daughter is named Sam, she's 7."
What happens: The agent confirms two memories saved. From now on, restaurant suggestions factor in the allergy. Calendar entries about Sam recognize who she is.
What you learn: The agent has durable memory. What you tell it persists.
3. Connect your inbox
Text: "Connect my gmail."
What happens: The agent sends a one-time OAuth link. You authorize. After 30 seconds, the agent has read access to your inbox.
What you learn: Real-data integrations. The agent is not just a chat surface; it can read your real stuff.
4. Ask for a summary
Text: "What's in my inbox right now that needs a reply this week?"
What happens: The agent scans your unread, prioritizes, and replies with 3-5 specific items. Names, not subjects. Action-oriented.
What you learn: The agent applies real judgment to your real data. It is not just listing emails; it is curating.
5. Book something
Text: "Find me an opening for a haircut next Saturday morning. Felipe's on Hudson if they're open."
What happens: The agent checks Felipe's hours, suggests two times in your calendar's open windows, and offers to book. You confirm; the booking happens.
What you learn: The agent acts. It does not just suggest.
6. Run a small everyday task
Text: "Play the morning playlist on the kitchen speaker."
What happens: Music starts. The agent reports back ("Playing 'Morning Playlist' on the kitchen speaker, 47 tracks, 2 hr 12 min").
What you learn: The agent operates on your real-world devices. The phone is the launcher; the agent is the operator.
7. Ask for the morning brief on demand
Text: "What's on today?"
What happens: The agent generates the same brief it would have sent you at 7am, but right now. The output reflects whatever just changed since the scheduled brief fired.
What you learn: Scheduled tasks and on-demand tasks share the same logic. You are not at the mercy of the schedule.
8. Save a skill
Text: "When I ask you to draft an email to a client, always start with 'Hey [name],' and end with 'Best, K' as my signature."
What happens: The agent saves a skill. The next time you ask for an email draft, it uses your salutation and signature without you having to mention them.
What you learn: The agent learns how you like things done. Procedural memory is the second-most-important memory layer.
9. Ask a research question
Text: "What are the three most-discussed AI agent products in 2026? One paragraph each, with citations."
What happens: The agent runs a web scan, synthesizes, and returns three paragraphs with cited sources. Total time: under a minute.
What you learn: The agent does research, with the discipline of citing sources. It does not invent.
10. Set up a small reminder
Text: "Remind me to call Mum tomorrow at 11."
What happens: The agent confirms and sets a one-shot scheduled task. Tomorrow at 11, you get the reminder.
What you learn: Reminders are first-class. The agent's scheduled-task system handles cron, intervals, one-offs, and ISO timestamps without you knowing the difference.
- 1Morning briefDaily routine.
- 2Core memoriesWhat is true about you.
- 3Real inboxOAuth into Gmail.
- 4Inbox triageReal judgment on real data.
- 5Real bookingThe agent acts.
- 6Everyday deviceSmart-home + music.
- 7On-demand briefSame logic, different time.
- 8SkillHow you like things done.
- 9ResearchWeb scan with citations.
- 10ReminderOne-shot scheduled task.
What you have at the end of day one
Ten interactions. The agent now has core memories about your family, has read your inbox, knows your morning routine, has booked one real thing, has played music in your kitchen, has written a research summary, and has a skill saved about how you write emails.
The next day's morning brief reads differently. It references the inbox triage, your scheduled events, the research you asked about, the things you said about your family. The agent feels like it has been paying attention because it has.
The third day, you stop thinking about what to ask. You just text things as they come up. That is when the system has converted.
- Day 1: Send 'hey.' Receive 'hey.' Pause.
- Day 2: Try to remember what you wanted to ask. Forget.
- Day 7: Open ChatGPT instead. Forget about Ajeris.
- Day 1: Ten interactions, agent has real context about you.
- Day 2: Morning brief lands. Conversation continues.
- Day 7: Texting the agent is muscle memory.
Frequently asked
Do I have to do all ten on day one?
What if I do not have a Hue / Spotify / Gmail?
What is the right second-day list?
Can I undo any of these?
What if the agent gets something wrong?
Try it
The first message is "hey." The next ten are above. By the end of week one, the system has converted.
