Stop doing these tasks manually

Most people are losing 5+ hours a week to tasks an agent could do. Free download inside to find yours.

Most people I talk to use AI most days. They prompt, they play, they’re pretty good at it. You can probably write an email in 60 seconds that used to take you 20 minutes. Congratulations: you’ve saved enough time over a year to watch the entire Lord of the Rings extended trilogy. Twice*.

But most people stop there. They’ve got AI doing ad hoc tasks faster, and they think that’s it. Meanwhile, the boring, repetitive stuff is still being done by hand. Same task, same week, same person, forever.

You can be a perfectly competent prompter and still be leaving most of the time-saving use cases on the table. Which sucks for you.

*Calculations may be incorrect.

The shift from “user” to “architect”

The people getting hours back each week have built agents.

An agent (also called a GPT in ChatGPT, a Gem in Gemini, a Project in Claude, or an Agent in Copilot) is a reusable expert. You set it up once with instructions and any reference files it needs, and from then on it does the same task the same way every single time. No re-prompting. No re-explaining.

Think of it as training a team member who actually remembers the briefing, instead of a brand new contractor you have to walk through the job from scratch every time.

Ask yourself this question:

How many hours am I spending on tasks an agent could handle?

For most knowledge workers I know, the answer is north of five hours a week. Sometimes closer to ten. Or even 20 (argh!!).

These tasks nibble away at your week, often without you noticing. They're scattered across small pockets of time, none of them feels worth optimising in isolation, and they're often the kind of low-grade admin you don't even count as "work" anymore.

A few of the ones I see (and hear about) most often:

Meeting follow-ups. The average knowledge worker spends 20+ hours a week in meetings. If you’re still manually pulling out actions, decisions, and summaries afterwards, you’re doing work an agent should be doing. (I use Granola for the transcript, then an agent (built in Claude) for the summary in the format I actually want. Human judgement must go on top, because AI will faithfully include the off-tangent chit-chat about someone's weekend that nobody needs in the recap.)

Pre-meeting briefings. In any given week, I’m in anywhere between 5-15 sales meetings with people I’ve never met before. In the olden days (read: 2022), I’d manually go to the person’s profile on LinkedIn, check Google News, search my inbox for prior conversations with others from the same company, and look in our CRM. That’s twenty minutes per meeting, easy (which in a 15-meeting week, is over half a day’s worth of time). An agent now does the searching and summary in the format I want, with the context I care about, without me having to visit a single website or manually do a single lookup.

Editing and proofreading. I built an agent to help edit my latest book, The Energy Game. It identifies where my argument doesn’t flow and where paragraphs go flat, which I think made for a better manuscript when I submitted it to my human editors. (Important note: if you don’t give your editing agent good instructions, it will turn your human writing into something that reads like AI wrote it. Which defeats the point entirely.)

Comparison decisions. Whether it’s software through to ergonomic chairs, an agent can pull together a comparison table with the criteria you care about. (Tip: ask for red/amber/green emojis next to each dimension. This makes the trade-offs visually obvious in two seconds.)

Industry news. Instead of reading the same generic newsletters everyone else reads and then getting lost down an internet rabbit hole through all your clicking, an agent can pull personalised industry updates for your specific role, your specific topics, over whatever time period matters. I have one that runs every morning at 8am to tell me everything I need to know about the state of AI in my specific context (the psychology of human behaviour and workplace culture).

Inbox triage. If your AI is connected to your inbox, an agent can summarise what came in overnight, flag what’s important from your key people, and draft responses in your voice. You’re still the human in the loop. The agent does the grunt work.

Bonus: an agent that finds your agent opportunities

In addition to the actual building of agents (that work reliably and consistently every time), I find that the biggest issue people have is identifying where an agent could help them in the first place.

Most people are too close to their own work to see the patterns. The repetitive tasks feel normal because they’ve always been part of the job.

So the team at Inventium.ai have built one for you. It’s called the Agent Opportunity Spotter, and it does exactly what it sounds like. You tell it about your role and your typical week, and it interviews you to surface the tasks where an agent would give you the biggest return.

Scroll to the bottom to download your very own Agent Opportunity Spotter.

And if you want to actually build those agents…

Spotting opportunities is one thing. Building agents that work (and don't go off-script the moment you turn your back) is another.

That’s what we’re launching our AI Agent Bootcamp for. It’s a four-week virtual program (kicking off on July 27) where you build agents alongside Neo Aplin (Inventium’s Head of AI and my co-host on How I AI). Roughly half the day is hands-on building. You leave with multiple working agents, including a Research Assistant, a Critique Buddy, a Brainstorm Buddy, and a Workflow Analyst Agent already built and ready to use, plus at least one custom agent you’ve built yourself for your specific role.

Spots are limited and I suspect they’re going to fill fast. Book today for our virtual AI Agent Bootcamp. And if learning in-person is more your style, we are also running this program as a full day in Melbourne (July 28) and Sydney (August 4).

And as an amazing One Percent Better subscriber, use code EB250 to get a $250 discount - valid until Thursday May 21, midnight.

If you take one thing from today’s newsletter, take this: instead of thinking “I need to use AI more”, try thinking about how you can use AI less, because the agents you’ve built are doing the work for you while you focus on the things that actually need a human.

Cheers

Amantha

P.S. Tell me in the comments: what agents have you built or do you want to build?

Download your Agent Opportunity Spotter…