How to create agents better than 99% of people

Edition 140 | November 30, 2025

This is what happens when you let an AI agent run without guardrails. First it leaks your valuation, next it's telling VCs you cry in the shower.

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

In today’s issue…

• How to create better agents

• Anthropic new advanced tool calling capabilities

• Agents may influence $73B in Cyber Week sales

• IRS deploying Salesforce’s Agentforce

• How agent search beats Google

…and more

🔍 SPOTLIGHT

Over the next few years, we’ll all have agents working for us in some capacity, whether we realize it or not.

They are the future of work, yet almost everyone I talk to struggles with the same questions: “What should I actually use them for?”

And I struggled with this too. I’d build an agent, get excited about the possibilities, then watch it sit unused because I automated the wrong thing. Not everything needs an agent, contrary to popular belief, and building the wrong one wastes time you’re trying to save.

So I developed a fun, dead-simple framework to spot agent-worthy tasks.

I call it B.O.T.S:

B - Boring & Repetitive

These are easy to spot and are the absolute low hanging fruit. The “I have to do this again” tasks:

• Sending follow-up emails with basic personalization

• Weekly status report compilation

• Data entry across multiple systems

• Calendar scheduling back-and-forth

• Quick test: Would you happily pay someone $20/hour to never do this task again?

O - Oops Prone

These can be especially frustrating. Tasks where small mistakes create big headaches.

• Misspelling client names in proposals

• Entering wrong data into your CRM

• Missing important details in documentation

• Forgetting to follow up with hot leads

• Quick test: Have you made the same type of mistake more than twice in the last month?

T - Time Intensive

The marathon tasks that eat your productive hours.

• Researching competitors or market trends

• Creating comprehensive meeting notes

• Outlining project proposals

• Analyzing customer feedback at scale

• Quick test: Does this task regularly take more than 30 minutes of focused work?

S - Snowballs

These can be harder to spot. These tasks seem manageable now but multiply as you grow.

• Lead qualification and follow-ups

• Customer onboarding sequences

• Invoice processing and reconciliation

• Content distribution across channels

• Quick test: Ask yourself, "What happens when I 10x this?" If the answer makes you sweat, you've found a snowball.

Categorizing your life tasks in these buckets helped me find where I could leverage an agent or leave it alone. And what’s important to remember is the agent does NOT have to do the whole task. If it just makes it easier to do, that is a great starting point.

So, now you’ve identified a few aspects of your life that could be agentified. Where do you start?

This is where most people would just pick randomly and lose traction from one of these playing out:

1. Build the most complex agent first (and burn out)

2. Build the easiest agent first (and see no real impact)

Like life, an agent is all about leverage. Finding the 80/20. Where a small amount of work or effort can create the most outsized return.

The impact x effort matrix helps you visualize where to put your effort.

High Impact, Low Effort is the goal. Some simple agent examples that fit this:

• Lead Response Agent: Auto-responds to inbound leads within 5 minutes with personalized messages

• Meeting Summarizer: Captures action items and sends follow-ups automatically

• Invoice Processor: Extracts data from invoices and logs to spreadsheet/accounting software

Where you start to lose traction are the agents that don’t do much for you, whether they are hard to build or not.

Look, I've built hundreds of agents at this point.

The ones that actually get used? They're never the fancy ones. They're the boring ones that just work.

Do a simple audit of your tasks, run it through B.O.T.S. and check where it lands on Impact x Effort. If it passes both tests, make a simple agent for it.

What's your biggest time-suck right now? Seriously, hit reply and tell me. I read every email, and I'll tell you exactly how I'd tackle it with an agent.

Let's build something boring and useful together.

—AP

P.S. - If you're thinking "but I don't know how to build agents"—that's literally why I run a community teaching people to build no-code agents. You don't need to be technical. You just need to be tired of doing the same thing over and over.