Which AI agents are right for business
It's easiest to choose an agent by team and task, not by model.
- Support. The agent works out a request's topic and priority, finds the answer in the knowledge base and writes a draft, and a support rep checks it and sends it to the customer. The agent passes anything about money or outages straight to people. See how it works in How does AI handle first-line support requests?
- Sales. The agent reviews a website lead, checks whether the company is already in the CRM, creates a deal for a sales rep and drafts a reply. The rep starts the conversation already knowing the context. How leads get into your team chat: How do you get website leads straight into your team chat?
- HR and new hires. The agent answers questions about vacation, sick leave, business trips and access based on company policies and cites the source. HR stops answering the same thing, and new hires don't have to wait for a colleague to be free.
- Managers. The agent compiles a weekly report from the team's threads and prepares plans and meeting recaps. Ready-made prompts are in How can a manager use AI for reports and planning?
- The whole team. The agent recaps a long discussion, pulls out decisions and open questions, and creates tasks with owners and due dates. Anyone who missed the conversation reads only the recap.
- Engineering. The agent figures out why tests failed, finds the cause and sends a fix. How a team agent differs from a personal one like Claude Code or Cursor: Personal vs. team AI agents in a team chat app.
- Documents and finance. The agent pulls data from PDFs, spreadsheets and scans, checks amounts and puts together a summary in the thread.
In the demo above, you can see five of these agents reply right in Pachca threads.
Case studies: how companies use AI agents
Flowwow. The company has about 10,000 internal documents with product, technical and operational information. So that employees don't have to search across different systems, Flowwow deployed its own language model in a closed environment. Now people ask the bot in Pachca, and it finds the answer in about half a minute. As a next step, the company is rolling out agents that create a calendar meeting right from a thread and write a short summary of a discussion with hundreds of messages. More in the Flowwow case study.
Pachca and the agent Kai. We work with an AI agent ourselves, called Kai. People give it tasks in threads, the same way they would a person. Kai handles about 90% of content tasks for the website, and requests to engineering have been cut in half. In support, it prepares most answers to technical questions in 2 minutes instead of 30: it reviews the request, finds the answer in the knowledge base and writes a draft, and a teammate checks and sends it.
Where you still need a person
The agent prepares, a person decides. Payments, refunds, complaints, contracts, HR decisions and anything that goes to a customer on the company's behalf are checked by an employee. A good agent hands these cases to people on its own: it @mentions the right person in the thread and attaches everything it has gathered.
Keep checking the agent's work after launch, too. For example, once a week look at where people edited its drafts and add to the knowledge base. Why AI isn't replacing people yet but is changing their work: Will AI replace employees?.
How to connect an AI agent to Pachca
In Pachca, an agent works as a bot: a separate member with its own name and avatar. You add it to the chats and channels it needs, and it replies in threads when someone @mentions it or replies to it. There are three ways to connect an agent:
- No code. You create the bot in Pachca, and an n8n workflow with a language model and a knowledge base does the thinking. Step by step: How do you build an AI agent without code?
- Your own agent via the API. Developers connect an agent on any model through the API, Agent Skills, MCP or the CLI. It's all in the AI agent docs.
- A personal agent. Claude Code, Cursor or Codex work with your chats and tasks. How to connect one: How to connect your personal AI agent to Pachca.
The agent sees only public channels and the chats and threads it's been added to. The workspace admin decides who can connect integrations and agents. Remove the agent from a chat and it loses access to it; turn off the bot and the agent stops.
Frequently asked questions
Frequent, repetitive ones where the answer is in company data and the result is easy to check: answering employees from the knowledge base, lead triage, discussion recaps, draft replies in support.
A chatbot follows a pre-written script. An AI agent decides for itself how to get a task done: it reads the conversation, searches documents, creates tasks and reports back on the result.
Yes. In Pachca, you create a bot in the settings, and you can build the agent's logic in n8n without code: a language model, a knowledge base and a reply in the thread.