Find a reason to reconnect with HubSpot leads.

This walkthrough is for sales teams keeping enquiries in HubSpot.

AI lead generation automation for leads you already earned

A prospect asked for weekend installation. You could not offer it then. Now you can. Build a workflow that notices the change and finds the people it actually helps.

This recipe recovers existing enquiries. The trigger is a newly approved capability, not a daily instruction to chase everyone. The sales owner checks the match and decides whether to reconnect.

1. Build a “what changed” register

Create a Google Sheet with offer_id, capability, limits, evidence_url, approved_by, status. Export a small HubSpot segment of stalled enquiries with record IDs, the original request, owner, last contact and contact restrictions. Use only leads your team may contact.

Lead L01: Asked for weekend installation; paused because unavailable.
Lead L02: Asked for a lower price; no installation preference recorded.
Offer O17: Saturday installation now offered in Zone A.
Limit: Survey required before a date or price is confirmed.
Approval: sales manager; status=ready.

Keep contact addresses in HubSpot. Store the customer’s actual request, not a model’s guess about their budget or intent.

2. No-code path: a new offer starts the search

Build a Zap triggered by a new or updated Google Sheets row. Add a Filter requiring status=ready and a human approved_by value. Retrieve the pilot lead rows from your staging sheet, then use Looping by Zapier to evaluate each against the new offer with your AI assistant action.

Match OFFER to the recorded unmet request in LEAD.
Return lead_id, offer_id, request_quote, matching_offer_quote,
fit (yes/no/unclear), missing_information and draft under 70 words.
Only explicit overlap is a match. Do not invent interest, discounts,
availability or a previous conversation. Include the offer limits.

Filter out no; send unclear to a research queue. For yes, check HubSpot’s current record and log lead_id + offer_id in Zapier Tables before creating another proposal. Attach the two supporting quotes to the row. Start with five leads rather than looping through the entire CRM.

3. Let the sales owner approve the reason, not just the wording

L01 is a candidate if their location is Zone A; missing location means hold. L02 is not a match. The useful message explains the new Saturday option and asks whether a survey would help. It promises no date or discount.

Add Human in the Loop → Request Approval, with the offer, lead record, quotes and exact draft. This is a paid Zapier approval feature. On approval, create a HubSpot task for the owner to send the message. On rejection or timeout, stop. The owner checks recent replies and contact restrictions immediately before sending and records the outcome.

4. Advanced path: notice the prospect’s change too

Have a scheduled worker load a small allowlist of company news pages supplied by the sales owner. A Browser Use agent reads each page, captures the dated passage and compares it with the previous snapshot. A new location announcement might make an old multi-site enquiry relevant again.

Give the agent read-only browsing, a domain allowlist and a ten-step budget. Do not give it email or CRM write credentials. Browser Use lifecycle hooks let a developer capture evidence and pause execution; enforce access limits in the browser environment as well as the prompt.

A second worker joins the captured change to the original enquiry and approved offers, then creates the same review packet. The sales owner verifies the source date and relevance before any outreach. A page change without a useful match produces no message.

5. Test silence as a successful outcome

Run a lead who opted out, an offer outside the lead’s service area, an undated news page and the same offer twice. Expected results: exclude, reject, hold and reuse the existing proposal. Never treat a search failure as “no restrictions.”

Track approved matches, useful replies and time spent reviewing false matches. If most matches need invented context, improve the enquiry notes or stop the automation. A quiet queue can mean the filter is working.

Scope: An illustrative build recipe with synthetic examples, not a live-tested integration or a claim of savings. Start with test records. Connector access, API usage and approval features may require paid plans.

AI connection: Use your AI assistant of choice (e.g. ChatGPT, Claude, Gemini). For automation, configure its supported provider API or connector; chat access alone does not supply an API connection. Keep approval outside the model.