WorkBookmeetings.io
Our own outbound, run by ten agents
Our own product, not a client’s. Ten agents, one per stage, find the right companies, research them, write to them, follow up and book the meeting.
bookmeetings.io

- Made by
- TechBuilds, for ourselves
- Industry
- Outbound sales
- Built
- AI agent system
- Runs on
- Web
- Status
- Live, running our own outbound
- Services
- AI agents, Workflow automation
01The problem
Outbound is a chain of jobs. Find the right companies, research them, write to them, follow up, book the meeting. Tools existed for single steps, but nothing ran the whole chain, so a person did, and the chain stopped whenever they were busy.
02The context
We built it to run our own outbound first. That set the bar: it had to send mail we would put our name to, from domains that would not get our main domain flagged, and it had to keep running with nobody watching it.
03What we built
Ten agents, one per stage: lead finder, signal detector, researcher, enricher, evaluator, personalizer, outreacher, sender, reply handler and closer. It reads each prospect’s own website before writing. It pulls from Google, LinkedIn, Apollo, Hunter and Reddit, follows up across email and LinkedIn, handles the reply and books the call.
Lead finder and signal detector
Pulls candidates from Google, LinkedIn, Apollo, Hunter and Reddit and looks for a reason to write now.
Researcher, enricher, evaluator
Reads the prospect’s own site, fills in what is missing and decides whether they are worth a message at all.
Personalizer and outreacher
Writes from what it read, and follows up across email and LinkedIn.
Sender, reply handler, closer
Sends from dedicated warmed domains, reads the reply and books the call.
04Walkthrough
- 01Lead finderPulls from Google, LinkedIn, Apollo, Hunter and Reddit.
- 02Signal detector
- 03ResearcherReads each prospect’s own website before anything is written.
- 04Enricher
- 05Evaluator
- 06Personalizer
- 07OutreacherFollows up across email and LinkedIn.
- 08SenderMail goes out from dedicated warmed domains, so the main domain stays clean.
- 09Reply handler
- 10CloserBooks the call.
bookmeetings.io

The ten stages, named as they are in the product, beside the live site. The product itself sits behind a sign-in.
05Engineering decisions
The choices you can see the result of in the product.
One agent per stage, not one agent for everything
Each stage has one job, its own inputs and its own output. A failure at one stage is visible as that stage’s failure, and a stage can be changed without retraining the whole chain.
Read the site before writing a word
The researcher reads each prospect’s own website before anything is drafted, so the message is about them and not about a list.
Warmed sending domains
Mail goes out from dedicated domains that were warmed before use, so the main domain stays clean whatever the reply rate does.
An evaluator that can say no
A stage exists whose only output is whether to continue. Most of what the finder returns should not get a message, and the chain is built to drop it.
06What it changed
Ten agents run our outbound end to end without a person in the loop, and the main domain stays clean.
About the figures
Bookmeetings.io publishes its own figures on its own site. They are its marketing, not something measured for this write-up, so they are not repeated here.
07For a business like this
This is an AI agents build. Here is how we run one.
- AI agentsAgents that answer customers and staff, and hand over to a person when they should.A sensible first step: Conversation reviewExplore this service
- Workflow automationInvoices, reports and data entry between your tools, done automatically.A sensible first step: Shadow the workExplore this service
Every build runs the same way: a free call, a written scope with a fixed price, a working version to try each week, and a handover of the code, infrastructure and data to your accounts. The process, in full.
