A sales pipeline that finds, qualifies and follows up on its own
Finding the right businesses, researching each one and following up consistently is the part of sales that eats a small team's week. We built our own pipeline that does that work around the clock, so our time goes on conversations.
The problem
For a small studio, sales is a grind of small jobs: finding businesses that might need you, reading their website, working out who runs it, writing something personal, and remembering to follow up. Done by hand, it's the first thing to slip when client work gets busy.
We wanted that work done consistently every day, with a person stepping in only where a person actually adds something: a call, or a reply to someone who's interested.
How it works
- FindOwner-run businesses from Companies House, OpenStreetMap and imported lists.
- ResearchReads each website for what they sell, who runs it and how to reach them.
- QualifyAI scores each fit from 0 to 100 and rejects poor fits like chains and retail-only shops.
- PersonaliseWrites a first line about their business, not a template.
- Follow upA short, spaced sequence that stops the moment anyone replies.
Real data from the running system. Company names, emails and summaries are blurred.
- Targeting that filters itself. It looks for companies with one to four directors and skips dormant or very new companies before any AI time is spent.
- Messages that read like one person writing. Plain text, sent in the recipient's working hours with natural gaps, never in bulk.
- A task list for the human parts. Calls and warm replies land on a Your tasks page, and a push notification goes to a phone.
Handling replies
Every reply is read and sorted by AI:
- Interested: answered within about ten minutes with a case study and a booking link, and we get a phone alert.
- A question: an answer is drafted for us to check and send.
- Out of office: the sequence pauses until they're back.
- "Talk to my colleague": the new contact is added with a note about who referred them.
- Not interested or unsubscribe: never contacted again.
It learns what works
- Built-in A/B testing. Each message has several versions competing on positive replies. Most volume goes to the likely winner while new ideas still get tried. Clear losers are retired, and AI writes new challengers based on what's winning.
- Send times. Once there's enough data, it re-plans each night towards the days and hours that get replies.
- Targeting and volume. Reply rates by country and sector reorder the queue, and new leads are found faster or slower to keep a steady supply.
- Meetings count most. Marking a lead as a meeting or a win teaches it which messages lead to real conversations, not just replies.
Kept safe and honest
- Nothing goes out unchecked at first. It starts in a draft-only mode, then asks for approval on the first messages in each channel before running on its own.
- A watchdog every 30 minutes checks email authentication, blocklists, bounce and complaint rates, and stalled sending. It pauses things itself and sends an alert if anything looks wrong.
- Compliance built in. Every message identifies us, has an opt-out that is honoured forever, and it only contacts countries where business-to-business outreach is allowed.
- Recovers on its own. Crashed parts restart automatically, and all data is backed up nightly.
Is your follow-up the first thing to slip?
Tell us how you find and chase new business today. In 15 minutes we'll tell you honestly which parts are worth automating.