SAMPLE WORKFLOW
What an implementation actually looks like.
Take an industrial supplier with heavy daily quote-request volume. Email arrives, somebody parses it, somebody pulls up the catalog, somebody calls the warehouse, somebody drafts the quote, somebody sends it. By the time it goes out, the buyer has already asked two competitors. The before is the bottleneck most distributors recognize on sight. The after is what we build.
How the workflow is actually built
The orchestration runs in Make or n8n — both work. The parsing layer uses Anthropic Claude or OpenAI to read the inbound request, extract part numbers and quantities, and surface anything ambiguous for the reviewer. The catalog and pricing source is whatever you already use: Airtable for distributors who keep their catalog there, Google Sheets when that's where the truth lives, a pricing PDF for the businesses still operating that way, or a direct read from the ERP when an integration exists.
The draft assembly is a templated build in your house format — your line-item style, your terms, your sign-off. The model fills the structured slots; the template enforces the format. The draft lands in the estimator's inbox or a shared Slack channel. That's the human-review gate, and it is the only place the system hands off to a person before the customer sees anything.
The goal of the build is to move from manual quote assembly — where the senior estimator is the bottleneck on every request — to a reviewable draft your team can check and send faster. Inbound that used to wait for the next person with bandwidth gets a draft sitting in front of a reviewer instead. Human review stays in place on every customer-facing send.
Illustrative sample workflow. Actual implementation depends on tools, catalog complexity, approval rules, and team workflow. Time markers in the diagram are design targets, not measured outcomes.