An AI-powered proposal writer for an art fabrication workshop — replacing a manual, time-consuming drafting process with an autonomous agent that researches clients, structures scope and pricing, and delivers review-ready proposals in minutes.
The workshop director was writing every proposal from scratch — pulling together project details, pricing, timelines, and scope into documents manually each time. The process was slow, inconsistent, and pulled focus away from actually running the business.
<aside>
Make.com, Airtable, Gmail.
</aside>
<aside>
SOP — Proposal Writer AI Agent
System Architecture & Maintenance
</aside>
This was built as an AI agent system in Make.com. The agent is given a system prompt, context documents, and a set of tools — small single-purpose scenarios the AI can call at each step in the sequence.
Context documents give the LLM company knowledge — services, pricing, past projects, and previous proposals. The agent draws on this base to draft proposals that match the company's tone, structure, and specifications, drastically reducing manual effort.
Each tool is a miniature automation — sometimes just a single module — that the LLM calls to perform a specific task.
The agent is triggered when an email lands in a dedicated inbox containing a project brief. The email data is sent to a webhook and parsed by the AI model.
The agent uses Perplexity AI to research the client to enable highly personalised proposals.
Creates a new project record in Airtable, linked to the relevant client and contact.
Updates the client record with any new information surfaced during intake.
Searches Airtable for an existing client match. If found, links the new project to them. If not, creates a new client record — preventing duplicates automatically.
Drafts the full proposal as a formatted Notion page, ready for human review.
