Figures from the campaign send logs as of 29 September 2026.
Six agents, one shared folder
Each agent reads the file the previous agent wrote and writes the next one. Nothing is hidden inside a tool: you can open any step in a spreadsheet and check it.
ICP Builder
Drafts and scores ideal-customer profiles.
01_icp_list.csvProspect Extractor
Finds names, titles, emails and LinkedIn.
02_raw_prospects.csvList Qualifier
Scores fit and splits tier A and B.
03_qualified.csvEmail Drafter
Writes a personal sequence per prospect.
04_email_drafts.csvEmail Sender
Sends what's due today. Stops on any reply.
05_send_log.csvAnalytics
Reports what worked, by segment and email.
06_analytics.mdBuilt on Claude and Hermes agents, with Apollo and Exa for data, Apify for email discovery, and a daily scheduled run on the founder's own computer.
Selling pipeline to SaaS sales leaders
From a database search to a short list
people matched the chosen profile in Apollo. Searching is free, so the agent cast wide, then paid to enrich only the final 50.
A second batch of 47 funded-startup leaders came from Exa web search. Ten customer profiles were drafted before one was chosen.
Seven emails, one new reason each
Email 5 goes out only when the agent can name five of the prospect's own in-market accounts, for example lenders and insurers that grew headcount 10%+ in six months for a cloud-communications company. It found 120 such accounts for 24 prospects. The other 23 get six emails, so no one receives a weak one.
Four days in
- Asked for a discovery call 1 company
- Replied "no", stopped with colleague 1 company
- Hard bounces, stopped automatically 7 of 94
The sender agent read the inbox before the day's run, stopped the sequence for that prospect and a colleague on the same domain, and flagged it for the founder to answer personally.
Selling office lunches across Noida
An offer you can do in your head
Local businesses aren't in B2B databases, so the agents searched the web
131 contacts across ten segments, ranked by how likely they are to feed 50 people a day: IT offices without a cafeteria and BPO shift teams first.
Fourteen emails, paced for a repeat purchase
25 new contacts start each day, so volume ramps evenly and the new domain builds a sending reputation. After three days: 75 emails sent, 0 bounces, 0 complaints.
What broke, and how we fixed it
The sender is the first bottleneck
A transactional email service wasn't suited to cold email, and workspace security blocked a Gmail integration. A mailbox on our own domain worked on the third try.
Verify every address first
7 of 94 SaaS emails bounced, mostly addresses that no longer exist. Verification before sending is now a fixed step for every new batch.
Stop the whole company, not one person
When one person says no, their colleagues on the same domain stop too. It protects the brand and the domain.
Automation still needs a check
Two scheduled runs missed a day. Moving the start dates recovered every email without doubling anyone up.
What changes between a B2B service and a local offer
| A · SaaS pipeline service | B · Office meals | |
|---|---|---|
| Buyer | VP Sales, CRO | Admin, accounts, CEO |
| Data | B2B database first | Web search first |
| Cadence | 7 emails in 7 days | 14 emails in 30 days |
| Personalisation | Company trigger + five named buyers | Company name + per-plate maths |
| Sender | Founder's mailbox | Brand mailbox |
| Early result | 1 discovery call requested | 75 delivered, 0 bounces |