Automate Email → Remittance Ingestion

We’re looking for an experienced automation consultant to help us turn a currently manual email-processing workflow into a reliable, largely autonomous system using n8n, Make, or similar tooling. Our company operates software that processes financial documents on behalf of many different customers. We receive a high volume of emails from one of our customers’ major business partners. These emails contain financial documents—typically payment/remittance files and supporting documents explaining deductions taken from payments. Today, a member of our team manually monitors the inbox, opens each email, downloads the attachments, unzips files when necessary, figures out which customer each document belongs to, determines what type of document it is, and uploads it into the appropriate part of our software. We want to automate that entire process. The desired workflow looks roughly like this: Monitor a Gmail inbox for new messages from a defined set of senders. Process each email and its attachments, including ZIP archives that may contain multiple files. Identify which files actually need to be processed. Different email/report formats have different rules; some attachments are redundant versions of the same report, while others should be ignored entirely. Classify each relevant file into the appropriate document type. Importantly, a single email can contain multiple types of documents, so classification needs to happen at the individual-file level rather than simply classifying the email. Extract a customer identifier from the email, subject, filename, PDF, CSV, or TXT content. The exact location of this identifier varies by document format. Look up that identifier in a maintained mapping table to determine which customer account the document belongs to. We do not want the system guessing based on company names; uncertain or conflicting cases should be sent for human review. Upload each file into the correct customer account and document workflow in our application through an authenticated API or other interface that our engineering team will provide. Handle duplicates safely. Retries, repeated emails, and previously uploaded documents should never result in accidental duplicate records. Record what happened so that we can trace every incoming email and attachment through the workflow: what was received, how it was classified, which customer it was assigned to, whether it was uploaded successfully, and why anything was skipped or held. Surface exceptions for human review rather than requiring a person to review every transaction. Examples include an unknown customer identifier, an unfamiliar document format, conflicting customer information, a corrupt attachment, or a failed upload. Update Gmail and/or Slack so our operations team can quickly distinguish successfully completed messages from items requiring attention. The key philosophy is automation with strong guardrails. We would rather have an ambiguous document held for review than incorrectly upload a financial document into the wrong customer's account. At the same time, the normal, well-understood cases should require no human involvement at all. We have substantial historical examples of the different emails and document formats, along with documented rules for how each should be handled. We can provide these as a test corpus so the workflow can be developed and validated against real-world examples before being enabled on live email. We’d expect the consultant to deliver the working automation, keep business rules/configuration maintainable rather than burying everything in workflow nodes, implement appropriate retry/idempotency and error handling, and provide enough documentation that our team can operate and modify it afterward. We’d also like to roll this out incrementally: test against historical examples → run against live email in “shadow mode” without actually uploading → enable for a small subset of customers → expand once accuracy is demonstrated. This is intended to be the first of many automation projects. We have a sizable backlog of operational workflows across finance, customer operations, sales, data management, and internal operations that we’d like to automate. We’re therefore particularly interested in someone who can not only implement a specified workflow, but can understand messy real-world processes, identify edge cases, design sensible exception handling, and build automations that are robust enough to run unattended in production.