An n8n workflow watches Gmail, extracts lead fields with an LLM, and appends each lead to a spreadsheet row.
“Turn inbound emails into structured lead rows with n8n and an LLM” is an intermediate Sales AI agent workflow from Any. It connects n8n, Gmail, Google Sheets and ChatGPT in 5 steps. Setup takes about Not stated and running costs are about Not stated.
Inbound lead emails arrive as free-form text, so pulling out names, companies, and contact details by hand is slow and inconsistent. A simple automation can extract the fields, but empty values need a deliberate placeholder so deduplication keys stay unique and two different people never collapse into one row.
In n8n, add a Gmail trigger node and point it at the inbox or label where inbound lead emails land. Decide whether it polls on a schedule or fires on new mail, and confirm the node outputs the raw message body and sender details for the next step.
Add an LLM node after the trigger and give it a fixed schema: name, company, email, phone, and any other fields you track. Instruct it to return a single JSON object and to use a specific sentinel such as UNKNOWN for any field the email does not contain, rather than N/A or an empty string.
Add a Set or Code node that trims whitespace, lowercases emails, and replaces any missing or blank value with the same UNKNOWN sentinel. Build the dedupe key from the normalized fields so that two leads with different missing data cannot produce identical keys.
Query the target Google Sheet for an existing row matching the dedupe key. If a match exists, route to an update or skip branch; if not, append a new row. Keep the sentinel consistent in the sheet so future runs compare like with like.
Send rows that contain UNKNOWN in critical fields to a review queue or a separate tab so a human can fill them in. Periodically inspect merged or skipped records to confirm the sentinel logic is holding and adjust the extraction prompt if the model drifts.
You are processing an inbound lead email in an n8n workflow. Read the email body below and return one JSON object with the keys name, company, email, phone, and notes. Use the exact string UNKNOWN for any field that is not clearly present in the email; never use N/A, null, an empty string, or a guess. Do not merge information from signatures into the wrong field. Return only the JSON object with no commentary. Email body: [EMAIL BODY].
Given the extracted lead JSON below, produce a normalized dedupe key. Lowercase and trim all string values, replace any missing, blank, or placeholder value with the literal token UNKNOWN, then join the values of email, phone, name, and company with pipe characters in that order. Return a JSON object with the fields normalized and a single key field containing the joined string. Do not drop fields that are UNKNOWN, since that would make different leads collide. Input: [EXTRACTED LEAD JSON].
An n8n workflow watches Gmail, extracts lead fields with an LLM, and appends each lead to a spreadsheet row.
You need n8n, Gmail, Google Sheets and ChatGPT.
Plan for about Not stated. It is rated intermediate. The guide walks through 5 steps, starting with “Connect the Gmail trigger”.
Expect about Not stated in running costs (Under $25/mo range).
Inbound lead emails arrive as free-form text, so pulling out names, companies, and contact details by hand is slow and inconsistent. A simple automation can extract the fields, but empty values need a deliberate placeholder so deduplication keys stay unique and two different people never collapse into one row.
Score every demo request against your ICP, update HubSpot and ping the right rep in Slack within a minute of the form submit.
Spot order-status questions in the support inbox, look up the order in Shopify and reply with live tracking and an honest ETA.
An hour before each discovery call, get a one-page brief on the account, the attendee and recent news, saved to Salesforce.
Built this workflow? Show it is featured on AgentUseHub with a live upvote count that links back here. About badges
