.png?table=block&id=16f4768c-ac2d-8029-8b18-d20f2c1d4419&cache=v2)
Do not index
Do not index
Every integration adds value — and more support requests. Without a system in place, support can quickly overwhelm your team.
AI agents can automate support and help you scale your ecosystem.
In this guide, I’ll show you how I built an AI-powered support system at Lunch Pail Labs with Cursor, the agency swarm framework, Slack, and HelpScout. So you can automate reduce manual support efforts, avoid scaling costs, and free your team to focus on growth.
Step 1: Set up agent swarm
An AI agent swarm is a group of AI agents, each focused on a single task. This modular setup makes it easier to scale and manage processes.
At Lunch Pail Labs, I use AI agents for tasks through Slack. Every new agent is added to the swarm, ensuring AI processes stay organized and scalable.
If you don’t have a swarm, you can create one or build on an existing setup. This tutorial and my approach can help you get started.
Step 2: Map out the workflow
Before creating an agent, map out the manual process. Here’s how support worked at Lunch Pail Labs before AI:
- A question comes in through a forum or email.
- It goes to HelpScout
- For common questions, I use information from the documentation.
- If there’s no answer, I respond manually and create a ticket to update the documentation.
When reviewing your own process ask yourself:
- Where do requests come in?
- How do you answer questions?
- What resources or information do you use?
This will help you define the workflow to encode in the AI agent.
Step 3: Build the support agent
I connected HelpScout to Slack using Slack workflows and webhooks. This let new requests from HelpScout flow into Slack, which is linked to my swarm.
I didn’t connect HelpScout directly to the swarm to keep visibility into requests. You can integrate HelpScout directly if preferred.
Using Slack workflows, new requests tag the swarm. The assist agent picks up the request and enlists the support agent, which:
- Reads the request
- Looks up documentation
- Uses a browser agent if needed
- Drafts a response in HelpScout
- Alerts the assist agent, which updates Slack
I built the agent in Cursor using Python. Most of the work was handled through AI, so little coding was needed. You can follow this process even if you’re not highly technical.
What’s Next
The first version of this system took a few hours to set up and was live answering questions the same day. It runs at less than a penny per request and frees up time from repetitive support tasks.
Next, I plan to expand its capabilities. It could create Linear tasks when documentation is missing, update content, or search through the codebase and external sources to answer new questions.
If you want more insights on building systems to support your integrations, subscribe to my newsletter. I share actionable strategies and tools to help you scale
.png?table=block&id=9ba33ac6-8e12-48f6-b980-4333b612ec56&cache=v2)