https://collinsautomates.com/2026-10-28https://collinsautomates.com/media/site/collins-640.jpghttps://collinsautomates.com/media/reel/poster.jpgClient results reel: eight automations and what the clients saidA short reel of automations I have built in n8n and Make.com, followed by clients in their own words. It covers a subscription sign-up and booking flow, a Companies House lead trigger system, a scanned document pipeline, an AI email triage agent, AI personalised letters, a voice agent repair, a cited RAG assistant and a social listening system. Then a client speaks on camera about the work, followed by written reviews. The pieces in this portfolio go into each build in detail.https://collinsautomates.com/media/reel/client-results-reel.mp41452026-10-08 https://collinsautomates.com/work/2026-10-28 https://collinsautomates.com/reviews/2026-10-28 https://collinsautomates.com/work/n8n-revenue-gps/2026-10-28https://collinsautomates.com/media/n8n-revenue-gps/cover.jpghttps://collinsautomates.com/media/n8n-revenue-gps/cover.jpgRevenue GPS Dashboard from Stripe, Zoho Books, GoHighLevel and ClickUpn8n workflow that gives an agency founder one daily view of the money. Every morning it pulls recurring revenue from Stripe, invoices from Zoho Books, pipeline from GoHighLevel and client agreements from ClickUp, checks each source, and reconciles them against each other. Code calculates MRR, receivables, pipeline and 30/60/90 day forecasts split into confirmed and speculative. Gemini writes a daily brief from verified figures only, and a guardrail checks every number in it before it is published. In the recorded run: 12 steps in 22 seconds, 32 of 32 numbers verified, one unbilled client caught.https://collinsautomates.com/media/n8n-revenue-gps/video.mp41292026-10-28 https://collinsautomates.com/work/make-real-estate-follow-ups/2026-10-08https://collinsautomates.com/media/make-real-estate-follow-ups/cover.jpghttps://collinsautomates.com/media/make-real-estate-follow-ups/cover.jpgMake.com Real Estate Follow-Ups: Buyers, Sellers and ListingsThree Make.com scenarios for a real estate company, live and in daily use. Staff enter buyer and seller details in a Google Sheet and mark the row Ready. Make finds the listing file in Google Drive, emails the buyer and the seller with it attached, copies the office so every send is confirmed, and logs the time to a master sheet. From that log, buyers are followed up every three days and sellers reminded every seven. Close to five hours of daily follow-up work now takes minutes, and no buyer is missed.https://collinsautomates.com/media/make-real-estate-follow-ups/video.mp42872026-10-08 https://collinsautomates.com/work/make-gohighlevel-call-notes/2026-10-08https://collinsautomates.com/media/make-gohighlevel-call-notes/cover.jpghttps://collinsautomates.com/media/make-gohighlevel-call-notes/cover.jpgMake.com + GoHighLevel: Sales Calls Turned into AI Action PlansBuilt for a client running a sales team that makes 50 to 100 calls. When a call ends in GoHighLevel, a webhook fires, Make pulls the recording from Twilio, AssemblyAI transcribes it, and OpenAI turns the transcript into an action plan and a ready-to-send follow-up message. Both are saved as a note on the GoHighLevel contact. Reps stopped replaying recordings to remember what was agreed: they open the contact, read the plan, and copy the follow-up message.https://collinsautomates.com/media/make-gohighlevel-call-notes/video.mp41292026-10-08 https://collinsautomates.com/work/n8n-companies-house-triggers/2026-09-30https://collinsautomates.com/media/n8n-companies-house-triggers/cover.jpghttps://collinsautomates.com/media/n8n-companies-house-triggers/cover.jpgCompanies House B2B Lead Trigger Systemn8n system that turns the UK company register into four buying-signal triggers: first accounts just filed, accounts due soon, newly incorporated, and overdue. A 5.69 million company snapshot sits in a database kept current with the daily accounts files, so each trigger is a query rather than an API crawl. Every candidate is then re-checked on the live register, filtered by SIC code and director count, and scored 0 to 100 with written reasons. In the live test, 317 companies were checked and 296 qualified, with dormant accounts and strike-off notices flagged.https://collinsautomates.com/media/n8n-companies-house-triggers/video.mp41282026-09-30 https://collinsautomates.com/work/n8n-gtm-digital-workers/2026-10-08https://collinsautomates.com/media/n8n-gtm-digital-workers/cover.jpghttps://collinsautomates.com/media/n8n-gtm-digital-workers/cover.jpgn8n AI Digital Workers: Cited Pre-Call Briefs and Safe Follow-UpsTwo revenue digital workers built in n8n. The first prepares a rep for a call: it loads the account and its research signals, Gemini writes a three-part brief, and plain code then checks every citation against what was actually retrieved and strips anything invented. An account with nothing on file gets an honest refusal. The second runs after the call: it extracts commitments, flags any with no agreed date, updates the CRM record and drafts the follow-up into an approval queue. There is no send step anywhere in the workflow.https://collinsautomates.com/media/n8n-gtm-digital-workers/video.mp43612026-10-08 https://collinsautomates.com/work/n8n-rag-patient-assistant/2026-10-03https://collinsautomates.com/media/n8n-rag-patient-assistant/cover.jpghttps://collinsautomates.com/media/n8n-rag-patient-assistant/cover.jpgn8n AI Patient Assistant: Cited RAG, OCR and Eligibility EngineTwo n8n workflows for a fertility clinic brief. The first answers patient questions in English, Arabic and Hebrew from approved clinic content only, with a citation on every line, and refuses with a handoff when the question falls outside that content. The second accepts scanned lab reports that contain no selectable text, OCRs a copy while the original is stored untouched, extracts the values, and pre-screens the case against eight criteria in plain code. The model never decides eligibility, and invented citations are stripped before a patient sees them.https://collinsautomates.com/media/n8n-rag-patient-assistant/video.mp41702026-10-03 https://collinsautomates.com/work/make-social-media-posting/2026-10-08https://collinsautomates.com/media/make-social-media-posting/cover.jpghttps://collinsautomates.com/media/make-social-media-posting/cover.jpgMake.com AI Social Media Posting: Instagram, LinkedIn, FacebookA Make.com scenario that turns one upload into three platform-ready posts. Drop an image or video into a Google Drive folder and the scenario picks it up, Google Gemini writes a separate caption for Instagram, LinkedIn and Facebook, and Make publishes each one to the right account. Instagram will not accept a Google Drive link, so the file is re-hosted on Dropbox first and the Dropbox share link is what Instagram receives. One upload, three posts, each written for the platform it lands on rather than one caption pasted everywhere.https://collinsautomates.com/media/make-social-media-posting/video.mp4772026-10-08 https://collinsautomates.com/work/n8n-whatsapp-ai-agent/2026-10-10https://collinsautomates.com/media/n8n-whatsapp-ai-agent/cover.jpghttps://collinsautomates.com/media/n8n-whatsapp-ai-agent/cover.jpgn8n WhatsApp AI Agent: Auto-Replies to Text and Voice NotesAn n8n AI agent that answers a business's WhatsApp messages, including voice notes. Each message is checked first: text goes straight to the agent, voice notes are downloaded from WhatsApp and turned into text, and anything else gets a polite fallback reply. The agent answers from a business prompt that holds the menu, prices, hours and booking rules, and chat memory keeps each conversation in context. In the demo it is the virtual host for a restaurant, answering menu and opening-hours questions in seconds.https://collinsautomates.com/media/n8n-whatsapp-ai-agent/video.mp4302026-10-10 https://collinsautomates.com/work/n8n-telegram-calendar-assistant/2026-10-10https://collinsautomates.com/media/n8n-telegram-calendar-assistant/cover.jpghttps://collinsautomates.com/media/n8n-telegram-calendar-assistant/cover.jpgn8n AI Calendar Assistant: Manage Google Calendar from TelegramAn n8n AI agent that runs a Google Calendar from Telegram messages. Ask what is on today, move a meeting, book a call or block time for dinner in plain language, and the agent uses four Google Calendar tools to read, create, update and delete events. Before booking it checks for clashes, and when a slot is taken it suggests another time instead of double-booking. Every change is confirmed in the chat and shows up in Google Calendar straight away, so the calendar app never needs to be opened.https://collinsautomates.com/media/n8n-telegram-calendar-assistant/video.mp4362026-10-10 https://collinsautomates.com/work/n8n-whatsapp-restaurant-orders/2026-10-12https://collinsautomates.com/media/n8n-whatsapp-restaurant-orders/cover.jpghttps://collinsautomates.com/media/n8n-whatsapp-restaurant-orders/cover.jpgn8n WhatsApp AI Agent: Restaurant Orders, Bookings and Voice NotesAn n8n AI agent that runs a restaurant's WhatsApp. Customers ask for the menu, place an order or book a table, by text or by voice note. The agent reads the live menu from a Google Doc, confirms the order and the total, writes it to a Google Sheets order log, and checks Google Calendar before confirming a reservation. Voice notes are downloaded and transcribed with OpenAI first, so a spoken order is handled like a typed one. Staff edit the menu in a document and see every order in a sheet.https://collinsautomates.com/media/n8n-whatsapp-restaurant-orders/video.mp4472026-10-12 https://collinsautomates.com/work/n8n-telegram-inventory-agent/2026-10-12https://collinsautomates.com/media/n8n-telegram-inventory-agent/cover.jpghttps://collinsautomates.com/media/n8n-telegram-inventory-agent/cover.jpgn8n AI Inventory Agent: Check Stock and Reorder from TelegramAn n8n AI agent named Stocky that manages office supply stock from a Telegram chat. Ask what is running low and it reads live stock levels from a Google Sheet, compares each item with its reorder amount, and lists what needs topping up. Reply to reorder, and it calculates the new stock levels, writes them back to the sheet and ticks a Reorder Sent box for each item, then confirms in the chat. The sheet stays the single record of stock, and nobody has to open it to know what to buy.https://collinsautomates.com/media/n8n-telegram-inventory-agent/video.mp4922026-10-12 https://collinsautomates.com/work/n8n-receipt-ocr-tracker/2026-10-10https://collinsautomates.com/media/n8n-receipt-ocr-tracker/cover.jpghttps://collinsautomates.com/media/n8n-receipt-ocr-tracker/cover.jpgn8n AI Receipt Tracker: Telegram Photo to Google Sheets with OCRAn n8n expense tracker that starts with a photo. Send a receipt to a Telegram bot and the workflow saves the image to Google Drive, reads it with Mistral OCR, and uses Gemini to pull out the date, store, tax number, total, tax amount and payment method. Each receipt becomes one row in Google Sheets with a link back to the original image, ready for bookkeeping, expense claims or tax time. Nothing is typed by hand, and the original receipt stays one click away from its row.https://collinsautomates.com/media/n8n-receipt-ocr-tracker/video.mp4422026-10-10 https://collinsautomates.com/work/make-product-listing-generator/2026-10-10https://collinsautomates.com/media/make-product-listing-generator/cover.jpghttps://collinsautomates.com/media/make-product-listing-generator/cover.jpgMake.com AI Product Listing Generator: Etsy Images and DescriptionsA Make.com scenario that turns a Google Sheet of product ideas into ready-to-list products. Each row holds an idea, a prompt and an example image. Make sends the example photo and prompt to OpenAI's image model to generate a new product image, saves it to Google Drive and writes the link and date back to the sheet. ChatGPT writes the Etsy listing: title, detailed description, personalisation notes and 13 relevant tags. One sheet drives every product, so a whole collection is produced by filling in rows.https://collinsautomates.com/media/make-product-listing-generator/video.mp4482026-10-10 https://collinsautomates.com/work/n8n-telegram-ai-assistant/2026-10-10https://collinsautomates.com/media/n8n-telegram-ai-assistant/cover.jpghttps://collinsautomates.com/media/n8n-telegram-ai-assistant/cover.jpgn8n Telegram AI Assistant: Text and Voice Messages with MemoryA personal AI assistant that lives in Telegram, built in n8n. Every message is checked as text or voice. Voice messages are downloaded and transcribed with OpenAI, then both kinds go to an AI agent with chat memory, and the answer comes back in the same chat, with a typing indicator while it works. In the demo it explains what it can do, then writes a Python depth-first search on request. The same pattern becomes a team helpdesk, an internal knowledge bot or a client-facing assistant by changing the prompt and adding tools.https://collinsautomates.com/media/n8n-telegram-ai-assistant/video.mp4322026-10-10 https://collinsautomates.com/work/make-marina-agent/2026-09-23https://collinsautomates.com/media/make-marina-agent/cover.jpghttps://collinsautomates.com/media/make-marina-agent/cover.jpgAI Email Triage Agent to Gmail DraftsMake.com scenario that triages a front-office inbox with an AI agent and never sends anything itself. Each inbound email is deduplicated in a data store, passed to the agent over HTTP, and returned as structured JSON: intent, entities, draft reply, confidence, citations and an escalate flag. Confident answers become drafts for staff to approve, anything flagged goes straight to a human, and every message writes an audit row showing the facts it used. The knowledge base is a Google Sheet the office staff edit themselves, with no scenario changes.https://collinsautomates.com/media/make-marina-agent/video.mp41522026-09-23 https://collinsautomates.com/work/n8n-deal-screener/2026-09-24https://collinsautomates.com/media/n8n-deal-screener/cover.jpghttps://collinsautomates.com/media/n8n-deal-screener/cover.jpgAI Real Estate Deal Screener with Underwritingn8n workflow that screens inbound real estate deal emails end to end. AI extracts address, price, rehab, ARV, rent, taxes and insurance from messy broker and wholesaler emails, and returns null rather than guessing. Validation normalises the numbers and stops incomplete deals. Deterministic code runs the underwriting: all-in cost, cash to close, NOI, cash flow, DSCR, CoC, ROI, cap rate and the 70% rule, then ten buy box checks read from a sheet. AI only writes the explanation. Five deals in one run: 2 PASS, 2 FAIL, 1 NEEDS DATA.https://collinsautomates.com/media/n8n-deal-screener/video.mp41352026-09-24 https://collinsautomates.com/work/make-email-outreach/2026-10-08https://collinsautomates.com/media/make-email-outreach/cover.jpghttps://collinsautomates.com/media/make-email-outreach/cover.jpgMake.com AI Email Outreach: 100+ Personalised Emails a DayA Make.com outreach system that turns a Google Sheet of leads into personalised emails. For each row, OpenAI writes a message from that lead's own details, the right file is found in Google Drive and attached, Gmail sends it, and the row is updated so nobody is contacted twice. It sends 100+ personalised emails a day and saves over two hours a week of manual outreach, with every email reading like it was written for one person because it was.https://collinsautomates.com/media/make-email-outreach/video.mp4962026-10-08 https://collinsautomates.com/work/n8n-document-pipeline/2026-10-06https://collinsautomates.com/media/n8n-document-pipeline/cover.jpghttps://collinsautomates.com/media/n8n-document-pipeline/cover.jpgn8n OCR Document Pipeline: Extraction, Validation and IndexAn n8n pipeline that turns scanned documents with no text layer into structured, searchable records. Four scans went through one run: an English invoice, a contract page, a Spanish delivery note and an expense claim with handwritten fields. Each original is stored untouched, OCR runs on a copy, AI classifies and proposes fields, then deterministic code validates them, re-checking totals against numbers actually present in the text and catching impossible dates. Every document ends as a searchable PDF plus one index row, with its own review flags.https://collinsautomates.com/media/n8n-document-pipeline/video.mp41332026-10-06 https://collinsautomates.com/work/make-subscription-booking/2026-10-03https://collinsautomates.com/media/make-subscription-booking/cover.jpg https://collinsautomates.com/work/n8n-lead-mail/2026-09-23https://collinsautomates.com/media/n8n-lead-mail/cover.jpghttps://collinsautomates.com/media/n8n-lead-mail/cover.jpgAI Personalised Letter & Direct Mail Pipelinen8n automation that turns one row in a Google Sheet into a print-ready, personalised sales letter. For each lead marked in the sheet it screenshots their website with cookie banners removed, uses AI to name their 2 to 3 key services and pick the best fitting testimonial out of 10, fills a Google Docs template with the company, address and Herr or Frau salutation, exports a PDF to the print folder, and writes the links and status back to the row. It runs unattended at 07:00, and one CONFIG node moves the whole build to your own Google and OpenAI accounts.https://collinsautomates.com/media/n8n-lead-mail/video.mp4952026-09-23 https://collinsautomates.com/work/n8n-social-listening/2026-09-16https://collinsautomates.com/media/n8n-social-listening/cover.jpghttps://collinsautomates.com/media/n8n-social-listening/cover.jpgAI Social Listening & Lead Alerts to Slackn8n automation MVP that monitors Reddit, guitar forums, YouTube and RSS for buyers who want to switch, built on a $500 fixed budget. Each conversation is classified by an AI model, scored 1 to 10 for switching intent, and posted to Slack with a direct link and a ready draft reply for a human to send. A free keyword prefilter ahead of the AI keeps running cost at $1.21 a month instead of $10.49. Brands, keywords, the AI prompt and the score threshold are all edited in Airtable, with no workflow changes and every API key on the client's own accounts.https://collinsautomates.com/media/n8n-social-listening/video.mp4742026-09-16 https://collinsautomates.com/work/n8n-white-label-template/2026-10-06https://collinsautomates.com/media/n8n-white-label-template/cover.jpghttps://collinsautomates.com/media/n8n-white-label-template/cover.jpgn8n White-Label Template: New Client Onboarded by Config ChangeOne n8n intake workflow built to be reused across an agency's clients. The nodes never change. Client name, AI vocabulary, escalation threshold, CRM, logging and alert destinations all live in a single config node, so onboarding the next client is an edit rather than a build. Recorded running live: the same pipeline is swapped from a dental clinic to a real estate agency with zero nodes added or connections moved, then deliberately broken to show three retries, a caught error, an alert carrying the client and execution id, and a clean response still returned to the caller.https://collinsautomates.com/media/n8n-white-label-template/video.mp41812026-10-06 https://collinsautomates.com/work/n8n-wholesale-catalogue/2026-10-06https://collinsautomates.com/media/n8n-wholesale-catalogue/cover.jpg https://collinsautomates.com/work/n8n-dunning-repair/2026-10-06https://collinsautomates.com/media/n8n-dunning-repair/cover.jpghttps://collinsautomates.com/media/n8n-dunning-repair/cover.jpgn8n Workflow Repair: Dunning Pipeline Silently Skipping InvoicesA German trade company ran its dunning process on self-hosted n8n: open invoices from Lexware Office, customer data from Hero Software, reminder letters drafted by an AI agent. Every run finished green while €61,480 of overdue invoices was never chased and other customers were reminded for money they had already paid. I traced the workflow node by node and found four separate faults stacked on top of each other: pagination, German date parsing, the wrong amount field, and an AI agent retyping figures. All four are fixed, with monitoring around the steps that stay human.https://collinsautomates.com/media/n8n-dunning-repair/video.mp4812026-10-06 https://collinsautomates.com/work/n8n-retell-voice-agent/2026-09-15https://collinsautomates.com/media/n8n-retell-voice-agent/cover.jpghttps://collinsautomates.com/media/n8n-retell-voice-agent/cover.jpgn8n Migration & Retell AI Voice Agent Debugging (60 Zapier Zaps)An SMB was running 60 Zapier workflows and a Retell AI voice agent that dropped calls on transfers and after hours. I migrated every workflow to self-hosted n8n with retry, backoff and dead-letter alerting, then traced the voice agent end to end across webhook respond mode, container timezone, E.164 formatting and unwired Twilio error outputs. Four faults, all reporting healthy. Calls abandoned fell from 25.8% to 0.5%, transfer success rose from 63% to 99.2%, and p95 tool response dropped from 7.4s to 480ms.https://collinsautomates.com/media/n8n-retell-voice-agent/video.mp4682026-09-15