Agents that find leads, research prospects and run outreach pipelines. Each entry has a copy-paste prompt, difficulty, setup time and known pitfalls.
An agent running 24/7 on a Mac Mini finds prospects, verifies emails, sends outreach, makes PDF briefings, and tracks the pipeline, keeping a lessons log so it improves over time.
A HubSpot webhook alerts a 'Researcher' agent, which scrapes the prospect's site and LinkedIn for recent news. The trigger itself is a plain webhook, not an LLM.
An Elvis tribute singer says Grok Bot emails venues, negotiates prices, closes deals and reports back, and booked 30 shows in a month.
A native, clearly labelled card in the directory results.
Become the featured sponsor →An n8n builder logs WhatsApp chats to a sheet, runs a morning schedule for status=asked leads, routes Day 1/3/7 templates, and writes last_nudged_date so the same nudge never fires twice — after a bug that re-sent Day 3 every run.
A Reddit user built Pinloop, an open-source CLI, with Claude Code: it pulls postings from company hiring systems, judges each against their resume, and drops strong fits into a morning shortlist — they applied to 190 picks from ~10k reads and got 2 offers.
A second-generation HVAC and plumbing operator has Grok Bot grade every inbound call from the recordings at 5 pm each day. It works out the real booking rate (booked divided by booked plus not booked, not counting calls from techs and vendors). The last four windows were 86.8%, 87.5%, 74.1% and 89.5%.
Can a bot revive stalled quotes for a home-services company? The owner of an HVAC and plumbing company in Northern Virginia had Grok Bot send 100+ personalized follow-up emails from his work email, running it all from his phone in minutes. For each one it pulled the customer from the company's CRM, mentioned details from their actual appointment and asked where things stood. Replies came in the same day: one turned into a $2,000 sale, and a few more jobs started moving again.
He built a missed-call text-back for an HVAC-style business with Twilio and n8n, then tested it and three calls got no text. Carrier voicemail answered after about 25 seconds, so he cut the ring timeout to 18 seconds and treats any 'completed' call under 10 seconds as missed. Landlines can't get texts, so a Twilio lookup now flags them for a callback. Panicked redialers got three identical texts, so he sends only one per 30 minutes and tells the owner 'called 3x, urgent'. Every missed test call now gets a text or a flagged callback in under a minute.
A flat 30-day reorder reminder is wrong for wholesale buyers: a cafe that orders every 10 days hears too late, and a 60-day distributor gets nagged early. This n8n workflow for Shopify works out each buyer's usual cycle from the median gap between their last 10 orders, adds a few grace days, skips anyone with an order that hasn't shipped yet, and lets sales reps turn reminders off per account. Errors go to one 'email me' branch instead of failing silently.
Can a bot turn a mailer into a signed client? A property management broker traced one new client back through his Grok Bot setup. The bot made the mailer she got, built out the Google Business Profile, and wrote the blog posts in his voice that she read. When she filled out the contact form, it replied right away asking for a good time to call and drafted a text for him. It also looked up what she owns so he could talk knowledgeably on the call. She got a contract 2 hours after her inquiry: 6 new doors to manage.
She built an n8n workflow that takes a web-form lead, enriches it with Apollo, has an OpenAI agent score it 0 to 80 as Hot, Warm or Cold, saves it to Airtable without duplicates, and alerts the team in Slack and Gmail. In testing, the same lead got scores anywhere from 25 to 77 when company data was missing. So if enrichment fails, the workflow now skips scoring, marks the lead 'Not Scored', still saves it and sends a separate Slack alert.
A builder made an n8n WhatsApp assistant that answers student questions for a study-visa consultancy after hours. It answers only from the consultancy's own material on services, documents, intakes and offices, and never makes up fees, test scores or visa chances; those go to a counsellor. It collects the student's name, service, country and best time to call one question at a time, replies in the student's own language (English, Hindi or Punjabi) and saves each lead to a table. He shares what tripped him up, including the AI telling students their details were saved when they weren't, until he told it to confirm only after the save worked.
His eBay assistant, which he calls Flipper, drafts listings from a shared iCloud photo link, about 5 a day, and he checks each one. It also rewrote titles and item details on listings that weren't selling, flagged items to bundle or donate, and reorganized his store categories overnight.
Claire Vo (ChatPRD, How I AI) reviewed Claude Duet for Lenny's Newsletter. It built a sales deck in Google Docs while she was on her phone, and pulled sales pipeline data from Slack, email and her CRM. Other jobs were slower: shopping for Warriors tickets took more than 10 minutes and never closed the loop, and inbox triage was only okay.
A developer in Nepal built a WhatsApp and website-chat ordering agent for a cake shop. Customers write in Nepali, Roman Nepali and English, often mixed, and send voice notes, photos and payment screenshots. The agent proposes orders, but code checks items, dates and delivery fees, prices come from the shop's own database, and a vision model only flags payment screenshots for the owner to confirm. A staff reply pauses the bot for that customer and the owner gets a Telegram alert when it needs help. It runs in production for the client.
Scrapes 500+ leads, verifies the emails, writes a personalized icebreaker from each lead's website, and uploads everything to an Instantly campaign.
Before a client call, the agent built a meeting-prep website and a demo dashboard of the data it can pull, tested on a real use case.
Picks a contact from the database, writes a briefing, drafts a custom message, attaches portfolio assets, builds a presentation, logs it in the CRM, sends, follows up, and adds booked meetings to Calendar. It runs daily.
A context bug made the agent drop its polite template and send blunt, specific critiques of founders' sites to 40 founders. The author reports a 62% reply rate, 8 calls, and 2 paid $300 audits.
Sales Navigator list enriched in Clay, Claude writes a personalized first email, 2 follow-ups and 3 subject lines, then emails are verified and sent via Smartlead.
A builder posted plain n8n workflow-run videos on a 40-subscriber channel, and a viewer hired him for €750 paid upfront.
For a local newsletter, the agent finds businesses already running local ads, checks they serve the area, finds the contact info they publish, and saves each lead with the original ad attached. It never guesses contact details and flags anything it couldn't verify.
An n8n workflow texts anyone whose call goes unanswered, answers questions from the business's own info, and texts the owner when a lead is ready. It double-checks the call was really missed and won't double-text. It's live at a bike touring company.
A marketer connected Grok Bot to his lead tracker. It found a lead that had gone quiet, came up with an approach, and wrote a re-engagement email.
An influencer agency had about 500 new emails a day, with only 2–4 qualified brand deals among them. It set up Lindy on each influencer's Gmail inbox with custom prompts to qualify leads and send replies.
A media company trained a Zapier agent on its target advertiser personas. The agent searches the web for matching companies and contacts and drops them into Google Sheets, ready for review or for import into Salesforce or Instantly.
Zapier's tutorial sets up an agent that runs whenever a contact is added in HubSpot. It searches the web for the person and company and fills in the empty fields. A second agent checks your calendar each morning and sends you a Slack DM with background on external meeting guests.
An xAI post-sales lead runs a routine that checks every 15 minutes but only writes a prep card when a call is 15 to 35 minutes away. The card covers who's on the invite and who actually accepted, the one thing to leave the call with, a few talking points with links, and a short do-not list. Other routines list unfinished promises three times a day and surface unanswered asks where the next move is hers.
After every sales call, an agent writes it into the call log and the CRM, and before anyone gets a number to call, it checks that log first. The team once almost called a broker they had already spoken to, with three new scripts ready; since then nothing goes on the call list without that check.
xAI's enterprise GTM lead shares the prompt for a weekly media rundown on each strategic account. Before the first run it asks which industry, signal topics and roles matter. Then it searches for new webinars, podcasts, talks, exec interviews and X posts from people at the account, keeps a state file so it only reports genuinely new items, and actually consumes each one: 3 to 6 takeaways, every mention of your category and competitors with sentiment, quotes with timestamps, and possible buying signals. If nothing is new, it sends one line saying so.
A machine-tool salesman with 30 years in manufacturing uses Grok Bot as an engineering assistant. It searches more than 20,000 tooling items, cross-references competitor part numbers, checks tool fit on a customer's Swiss lathe, recommends inserts and starting speeds and feeds, and reviews part drawings so he walks into each visit prepared.
How do you give clients a reason to call you? He built an unclaimed-property bot and had it check his customer list. It found more than $8,000 waiting for 150 of his personal contacts and emailed each one how to collect it. He already has 4 client meetings from it, and clients are sending him photos of the checks they got.
How do I spend more time with customers and less on admin? An outside B2B sales rep says his Grok Bot turns his voice-dump call notes into Notion CRM updates, so he never types into Notion. Every morning it plans his driving route around existing meetings, overdue accounts and drive time. Every Monday it flags accounts with no meeting in 30+ days, and it keeps reminding him about follow-ups until he says they're done.
Could your listings show the wrong product? A side-business sports card seller's Grok Bot team found one of their live listings titled for one player at $5.99, but the photo was a different player. The catalog had matched the wrong card and nobody had looked. Now the bot checks every live title against its own photo.
He hands Grok Bot a list of leads. It's connected to Reoon and BounceBan through their APIs: Reoon checks regular emails, BounceBan scores catch-all addresses, and the bot keeps only the high-scoring ones for him to email, which protects his sending domain.