Daniel AwoboduVerified StackedHQ member

An AI Office Assistant that does the work of an admin in sorting out emails and only calling attention to the important emails that needs to be attended to. Categories: invoice(log to drive) | scheduling | General Urgent?: Sort + label need_followup? : create event in calender

I built an AI chatbot System that qualifies leads, sets appointments, creates an event in Google Calendar, verifies free slots in Google Calendar, and logs qualified leads to a Nurture list in Google Sheets. The Agent talk to visitors warmly, collect necessary details and intent and route based on the intent of the visitor.

Rebuilt this clinic's appointment reminder system from the ground up, since the old one was texting patients about slots that had already moved Core fix: every reminder gets rechecked against the live calendar right before it sends, not just when it's first queued. A short safety delay sits between the two checks so last-minute staff edits have time to land before anything goes out. Both checks are backed by a dedupe log keyed on the event, its exact start time, and the reminder type, so a rescheduled appointment automatically gets treated as new, and never re-fires its old reminder for the

Rebuilt how this real estate brokerage follows up with warm leads, since agents kept forgetting to chase and the messages that did go out felt copy-pasted. Built three connected lanes in n8n. The moment a contact enters HubSpot as a new lead, AI writes a personalized first message — referencing their preferred area, budget, and what they last asked about — and sends it through Gmail. If nobody replies, two more follow-ups go out on a set cadence, each one continuing the same email thread rather than starting fresh. The moment a genuine reply lands, everything stops immediately, and the agent

Built an end-to-end WhatsApp receipt automation for a delivery company. Drivers photograph or PDF a fuel/repair receipt to a Green API-connected number; n8n downloads it, routes image vs. PDF to separate Gemini vision calls (Gemini can't handle PDFs and images through one operation), extracts vendor/date/total/tax/category, matches the sender against an Airtable driver roster with an "Unknown Driver" fallback so unmatched numbers still get logged instead of dropped, then uploads the file into an auto-created monthly Drive folder and appends a row to the Sheet. Caught a real accuracy risk earl

A lead hits the webhook and gets a 200 back instantly, so the form platform never retries and creates duplicates. First, the free checks. One Code node reads Tally's payload and rejects renters, missing consent, and bad emails. Those leads never cost an API call. Then enrichment. Geoapify geocodes the address — gated on confidence score, not match type, because a wrong-street match still reports "building". RentCast returns ownership, square footage and roof material. PVWatts returns solar production for those coordinates. EIA returns the state electricity rate, cached 30 days so most leads

I built a voice agent fr a dental clinic with Vapi that answers inbound calls, handles FAQs, checks calendar availability and books appointments in the specified calendar. The clinic has several branches in different states, and it books each appointment to that particular state where the patient is located. The agent has 2 tools at its disposal to call which send to an n8n webhook URL: Check Availability and book appointment tools. The agent collects the branch, time and date and uses it to check availability in the calendar, then returns 3 different 1-hour slots closest to the time, date o

Three intake routes feed one shared pipeline in n8n. A website form posting to a webhook, an internal form for requests that come in by phone or in a meeting, and a Gmail inbox using plus addressing. The main decision was writing the Airtable record before the AI runs instead of after. Status goes in as Received, then classification updates it. If OpenAI fails, you still have the request. Every route normalises into the same shape and converges on one merge node. Classification is one call with a strict JSON schema, so the category cannot come back as something Airtable rejects. A code no