RecipeAgents / Guides
Agent recipe guideAI customer support agent for small business
A support agent is only useful if it's right and knows when to stop. This recipe uses Intercom Fin or Chatbase, both of which answer from documents you upload, with a system prompt that keeps the bot inside your docs and hands upset or uncovered customers to a human.
The prompt
You are support for [Company]. Answer only from the provided docs. If not covered or customer is upset, say you'll connect them to a human and collect their email.
Step by step
- Upload your docs, FAQ and help center articles.
- Set escalation rules: topics that always go to a human (billing disputes, refunds, legal).
- Embed the widget on your site.
- Every week, review the unanswered and escalated questions and write new docs for the common ones.
The weekly review is the product
The bot gets better only when your docs do. The questions it can't answer are a free list of the help articles you're missing. Twenty minutes a week reviewing them is the most valuable part of this recipe.
Intercom Fin or Chatbase?
Intercom Fin fits if you already use Intercom for support. Chatbase is a standalone option you can embed on any site. Compare current pricing on each vendor's site; both change often.
Mistakes to avoid
Letting the bot answer from general knowledge. 'Only from the provided docs' is the most important line.
Hiding the human option. Customers should always be able to reach a person.
Uploading outdated docs; the bot will confidently repeat them.
Build it in Make, Zapier or n8n
The recipe is tool-agnostic. Every step maps to a standard building block: a trigger from the app where the data starts, an AI step that sends text to a model with the prompt, a parser that reads the JSON or text that comes back, and actions in the destination apps. Zapier is the quickest to get running for linear flows. Make makes branching and loops easier to see. n8n can be self-hosted if you want to keep data on your own infrastructure. Pick the one your team already pays for.
Test it before you trust it
Run the recipe on ten to twenty real examples before you switch it on for everything. Put the inputs and outputs side by side in a sheet and mark each one right or wrong. Where it's wrong, add one specific line to the prompt that addresses the pattern, then run the same examples again. Keep a human approval step on any action that is visible to customers or touches money until the error rate is one you'd accept from a new hire.
Watch running costs as well. Each AI step is billed by the model provider and each task or operation by the automation tool. Filtering inputs before the AI step, for example skipping newsletters or tiny files, is the easiest way to keep both down. Check current pricing on each vendor's site, since it changes often.
Before you use it
Every recipe on RecipeAgents follows the same shape: a trigger, the tools, the steps, and the exact prompt. Start with the human-approval version. Once you have watched a few dozen runs and trust the output, decide which steps can run on their own. Model and tool names change often, so check current pricing and features in each tool before you build.
Frequently asked questions
- Can a small business use an AI support agent?
- Yes, if you have written docs or an FAQ to train it on and someone reviews the escalations weekly.
- What happens when the AI doesn't know?
- With this prompt it says it will connect the customer to a human and collects their email.
- How do I measure if it works?
- Track how many conversations it resolves without escalation and how many escalations were about topics your docs already cover.