AI Agents for Small Business: Automate Without Coding

• Why Lili tested all 4 tools with the exact same real-world task: backlink outreach

• What the task involved and why it's a perfect benchmark for agent quality

• An intro to Lindy, Manus, Relevance AI, and Hyperagent before we go deeper

• How each tool performed on the same test task

• What Lindy is good for and where it fell short

• Why Manus works well on the go but got messy, and why Relevance looks better than it performs

• Why Hyperagent was the only tool that completed the full task end to end

• What makes it stand out: live progress view, inbox connection, and output quality

• The one downside to watch out for, and why Lili still uses it every day

• A first look around the Hyperagent dashboard

• The core concepts you'll work with: agents, threads, tools, and invocations

• How these map to other platforms if you're using Lindy or Manus instead

• What the system prompt is and why it's the most important part of any agent

• How projects work as folders that let related agents share context and data

• When to use a project and when a standalone agent is enough

• The difference between an agent and a thread

• When to continue in the same thread and when starting fresh saves you money

• Why long threads slow your agent down and cost more credits

• How to pick a cheaper model for simple tasks and why it matters for your bill

• What invocations are and how they control when your agent runs

• Why "context stays in one continuous thread" can quietly drain your credits

• How to connect your agent to Gmail, Slack, Drive, Airtable, and other tools

• The difference between tools and skills, and why skills make your agent more reliable

• How to add tools at agent level or directly inside a thread

• How the library stores everything your agent produces, in every version

• What memories, skills, and rubrics are and how they make your agent smarter over time

• When to add a memory, when to write a rubric, and how learning reduces the work you repeat

• How Lili's Sales Prospector burned through credits on a task that was too open-ended

• The 3 things she should have done differently from the start

• Why scheduling an agent too early is one of the most common and costly mistakes

• What "human in the loop" means and why it's your safety net especially at the start

• How to set checkpoints so the agent pauses before doing anything risky

• When you can start loosening control as trust builds

• A one-slide recap of how Lindy, Manus, Relevance AI, and Hyperagent compare

• How each performed on the same test task and which one came out on top

• Skip here from chapter 9 if you already know which platform you want to use

• A quick recap of the dashboard concepts covered in section 2

• Agents, threads, projects, integrations, learning, and credits summarised in one go

• Silly Lili's unfortunate Zoom moment makes an appearance

• Why you don't need to write an essay to get a great agent setup

• How global memories replace most of the context you'd otherwise repeat every time

• When to let the agent figure things out versus when to give it an example

• How to plan an agent before you build it: what it is, what form, what cadence

• Live setup of the inbox triage agent step by step in Hyperagent

• How to define what's important and what to ignore for your specific business

• Live setup of the content repurposer that turns one comparison page into a blog post, LinkedIn post, and newsletter section

• How to use HITL checkpoints to review copy before the agent generates images

• Why giving the agent an example output saves credits and guesswork

• A walkthrough of the Sales Prospector setup, including the Hunter.io integration for contact enrichment

• What Lili got wrong and what it cost her, so you don't repeat the same mistakes

• Why she scrapped the cold outreach approach entirely and what she learned from it

• How to update an agent without touching the system prompt yourself

• When to use plan mode versus execute mode and why it matters for complex tasks

• A live example of Lili changing the inbox summary agent in under a minute

• Why you don't always need an agent and when Claude with a skill is a better fit

• A real example: the offer writer skill that handles Lili's client replies without burning agent credits

• The simple question to ask yourself before reaching for an agent

• Why starting a new thread for unrelated tasks keeps your agent sharp and your bill lower

• How to set output limits so the agent doesn't keep going when you only needed 20 results

• The exact phrasing that stops an agent from running indefinitely

• Why you should test an agent manually several times before scheduling it

• How HITL checkpoints stop the agent from burning credits in the wrong direction

• How switching to a cheaper model for simple tasks can save money without affecting quality

• Why you shouldn't overengineer the agent early on and what to do instead

• How to fine-tune through conversation rather than rewriting the system prompt

• Progress over perfection: how to iterate quickly and trust the process

• The 5 types of tasks that aren't worth delegating to an agent

• When an LLM is faster, cheaper, and frankly better for the job

• The simple rule of thumb: agent for doing, LLM for thinking

• Catch yourself repeating things and save them as a memory instead

• Turn your weekly rituals into skills so the agent always knows how to handle them

• Define what "good" looks like with rubrics so the agent self-checks before showing you the output

Questions are:

1. Is there a free way to try this before I pay?

2. I'm not technical at all. Is this really doable for me?

3. Can the agent send emails or make changes without me approving them first?

4. Is it safe to connect the agent to my email inbox?

1. Can one agent do several different jobs, or do I need a separate agent for each?

2. Is the agent also working when my computer is switched off?

3. Can the agent learn and get better over time?

4. Can I have the agent work with my existing tools like Gmail, Slack, or Google Sheets?

5. Is my data safe when I connect my tools like Airtable to an agent?

• What a voice agent actually does and how it differs from a text-based agent

• When it makes sense for a small business: call handling, lead qualification, after-hours coverage

• Which tools Lili tested and which one she'd recommend today

• Lili's final advice: start with one simple agent today and build from there

• Where to leave a review, ask questions, and find Lili after the course

• A send-off only Silly Lili could deliver