Dear Techies,

You know the one.

It's on the calendar. You glance at it five minutes before it starts. You have no idea what it's actually about, who's in it, or what you were supposed to have prepared.

So you wing it. Again.

The frustrating part isn't the meeting. It's that two minutes of preparation would have made the whole thing better. You just never seem to have those two minutes.

That's the problem an AI agent can solve. And the only thing standing between you and that solution is knowing how to ask for it clearly.

That's what this newsletter is about.

Most conversations about AI agents focus on the tools. Which platform. Which integrations. Which model.

The tool matters less than you think.

What actually determines whether your agent works is how clearly you can describe what you want. That's the whole skill. And it's not a technical skill. It's a communication skill.

Here's a way to think about it. Imagine you've just hired someone brilliant, motivated, and completely new to your world. They don't know your job, your priorities, or your preferences. But they'll do exactly what you ask, thoroughly and without complaint.

What would you tell them?

The people who get great results from agents aren't necessarily more technical. They're more precise. They've learned to describe a task the way a good manager briefs a new hire: clear goal, relevant context, specific output.

That's the shift. From "how do I set this up" to "how do I explain this well."

There's a simple structure that makes agent instructions much easier to write.

Name the role. Tell the agent what kind of helper it is. This shapes how it approaches the task.

State the goal. What should it accomplish? One clear sentence.

Specify the output. What should the result look like? How long? What should it include?

Set the constraint. What should it avoid? What are the limits?

You don't need all four every time. But when an agent isn't doing what you want, one of these is almost always missing.

Let's take this somewhere entirely outside work.

Say you're a parent with three kids and school pickup is chaos. Different finish times. Different days for clubs. Different things to remember for each child. You carry all of it in your head.

Here's an instruction you could give an agent today:

"You are a family logistics assistant. Each morning, check my Google Calendar for any children's events, school pickups, or after-school activities happening today. Create a short daily summary that lists each child, what time they finish, where they need to be collected from, and anything I need to bring or remember. Keep it to six lines or fewer. Send it to my email by 7:30am."

Four sentences. Role, goal, output, constraint.

The agent doesn't need you to explain what school pickup is. It reads your calendar, finds what's relevant, and puts it in the right shape.

You write the instruction once. It runs every morning.

Let me be direct, because the video demos are designed to work.

Real life has messier calendars. Events with vague titles. Meetings with no description. Attendees listed by email address rather than name.

When the agent doesn't have enough information in the calendar event, the brief it produces will reflect that. Shorter. More generic. Less useful.

That's not the agent failing. That's the agent showing you something useful: the calendar event itself needs more detail.

A meeting titled "Sync" with no description will produce a weak brief. A meeting titled "Q3 content review: discuss performance data, plan next quarter" will produce a genuinely helpful one.

The agent is only as good as the information it can access. Before tweaking the instructions, check the source.

Write the instruction for one task you do regularly that involves gathering information and acting on it.

Don't build the agent yet. Just write the instruction. Use the four-part structure as your guide.

Here's what three finished versions look like:

"You are a client prep assistant. Each Monday, check my calendar for client calls happening this week. For each call, create a three-line brief: what the client does, what we last discussed if it's in my notes, and one question worth asking. Send the briefs to my email by 8am."

"You are a weekly priorities assistant. Every Friday afternoon, check my task list and calendar for next week. Summarise my three most important commitments and flag any scheduling conflicts. Send it to my email."

"You are a research assistant. When I forward you an email from someone I'm about to meet, find their LinkedIn summary and recent public work, and send me a short background note before the meeting."

Each one has a clear role, a trigger, a defined output, and a delivery method.

Write yours. The instruction you write today is the agent you'll build next.

The video for this episode shows the full live build, including the moment where the agent reads the calendar event and reasons through what to include in the brief.

That decision-making process, watching the agent think in real time, is something words on a page can't capture. That's what the video is for.

Watch the full build here : https://www.youtube.com/watch?v=qXItMYCBxsY

If someone you know is always running into meetings underprepared, forward them this. It might save them a painful conversation next week.

Stay Savvy,

Ijeoma | Tech Savvy Starts Here

P.S. Did you know I wrote a book? Tech Savvy Starts Here is available on Amazon—a practical, engaging guide for families and educators helping kids build confidence with technology. Check it out here.

Enjoyed this edition?
Forward it to a friend or colleague who will enjoy it as well.
Missed something? Catch up in the newsletter archive.

🧠 Keep learning. | 💬 Keep questioning. | 💥 Keep growing.