Use cases · AI agent examples at work

AI agent: real-world examples. Ten Hermes Agent use cases, with their guardrails.

Ten business tasks an AI agent can take on, most of them already running on our own projects. For each: the problem, what the agent does, and the human safeguard that stays in place.

Book 20 minutes with us See the examples

Hermes Agent is an open source project by Nous Research. Kailli is not affiliated with Nous Research.

01The examples

Ten AI agent examples

Tasks that come back often, follow known steps and are quick to check. That is where an agent pays off most.

Build and deploy a website

The problem: A website needs dozens of consistent pages, checked before every release.

What the agent does: It writes the pages, runs the automated quality checks, prepares the deployment and flags what is blocking.

Guardrail: A human still decides when it goes live.

The founder's second brain

The problem: Decisions, meeting notes and next actions are scattered across email, notes and messaging apps.

What the agent does: It keeps a knowledge base up to date and retrieves the context of any project in one question.

Guardrail: The founder reviews and corrects what gets recorded.

Sales management and CRM

The problem: Follow-ups slip through the cracks and dashboards take time to prepare.

What the agent does: It reads the CRM, prepares the follow-ups to send and generates the weekly dashboards.

Guardrail: Every message is reviewed and sent by a salesperson.

SEO research

The problem: Writing without knowing whether anyone searches for the topic means publishing into the void.

What the agent does: It queries keyword data and Google results, then ranks topics by real demand.

Guardrail: Rule: prove demand before writing. Choosing topics stays a human call.

Content publishing

The problem: An article always goes through the same steps: draft, review, formatting, publishing.

What the agent does: It writes from a recipe card, formats the piece in the CMS and prepares publication.

Guardrail: Nothing is published without review.

Invoicing and admin

The problem: Invoices and deposits demand rigor, and a mistake is costly.

What the agent does: It prepares invoices and deposits in the accounting tool from approved information.

Guardrail: Human approval is mandatory before anything is sent.

Discord community

The problem: A community asks the same questions and expects quick answers.

What the agent does: It answers in Discord, finds the relevant information and carries out the team's requests.

Guardrail: It only acts for authorized people.

Server operations

The problem: Audits, updates, backups: recurring tasks that are easy to forget.

What the agent does: It audits the server, checks the Docker containers and backups, and sends alerts to your messaging app.

Guardrail: Dangerous commands wait for explicit approval.

Semi-automated Instagram account

The problem: Posting regularly takes ideas, visuals, captions and a schedule.

What the agent does: It suggests ideas, prepares visuals and captions, and keeps the posting schedule.

Guardrail: Every post is approved by a human.

Email triage, monitoring and prospecting

The problem: The inbox overflows, keeping up with the news takes time, and so does prospecting.

What the agent does: It sorts emails and prepares drafts, summarizes a daily news watch, researches prospects and writes a first draft of the message.

Guardrail: No message goes out without a human.

02Where to start

Choosing your first use case

The right first case is small, frequent and easy to check.

  • The task comes back at least every week.
  • The steps are known and can be written on a card.
  • The result can be checked in a few minutes.
  • A mistake can be fixed without damage.

The Hermes Agent installation guide explains the framework to set up before the first case, and our AI agents training teaches your teams how to hand a task to an agent.

Let's pick your first use case together, in 20 minutes.

Book 20 minutes

Written by Rémi Carbonnefounder and CTO of Kailli

Frequently asked questions about use cases

What is the best first use case for an AI agent?

A frequent task, with known steps, easy to check and with no serious consequences if it goes wrong: a weekly summary, a news watch, follow-up drafts.

Can an AI agent manage an Instagram account?

It can suggest ideas and prepare visuals, captions and a schedule. We recommend that a human approve every post before it goes out.

Can an AI agent do prospecting?

It can research prospects and write a first draft of the message. Sending stays human: it is safer for your reputation and for GDPR.

Are your examples client cases?

No. They come from our own projects at Kailli or from what we can set up with you. We don't name any clients.

Twenty minutes. And you'll know whether we can help.

No sales pitch. You tell us where you stand with AI, we tell you how far you can go.

Book 20 minutes Prepare for the call