B2B outbound system: 50 qualified leads a week without manual research
B2B digital transformation consulting for mid-market companies in Russia. Two sales development reps (SDRs) dug through LinkedIn and SPARK by hand — 15 leads a week for 80 hours of work. A three-stage chain, Apollo → Clay → Instantly with AI personalisation, delivered 52 leads a week for 6 hours of team time.
2 SDRs + 80 hrs/wk = 15 leads
B2B digital transformation consulting (mid-market RU: revenue ₽100-1500M, companies 50-500 people). Target contact — CEO / COO / IT director. Average project size — ₽2-5M. Deal cycle — 3-7 months. Cold outreach is the main acquisition channel; inbound brings in under 20% of the sales pipeline.
Two SDRs worked manually in "researcher-copywriter" mode: scrolled LinkedIn, went to SPARK to check revenue, searched open sources for press releases and signals (hired a CDO, opened a branch, closed an investment round), and wrote a personal email for each. 80 person-hours per week — for 15 leads.
Reply rate hovered in the 4-6% range. Of those who replied, about a third qualified by our criteria. Cost per lead (CPL) with this process — ₽12.8k counting only SDR salary, without overhead. Scaling this meant hiring more SDRs — but the labor market is expensive, training is long, turnover is high.
Five stages: sources → data → AI copy → delivery → CRM
Simple idea — hand every repetitive SDR step to the system and leave the SDRs one job: check quality and close warm meetings. Apollo supplies the raw list matching the ideal customer profile (ICP), Clay fills in the company data, GPT-4o and Claude write variants around specific triggers, Instantly sends from warmed-up mailboxes.
1.2M contacts in Russia matching the profile: CEO / COO / IT director, 50-500 people, revenue ₽100-1500M. Segmented by industry.
SPARK revenue + RSS press releases + BuiltWith stack + new job openings — the facts each email is personalised around.
GPT-4o + Claude (via ProxyAPI.ru) generate 3 email variants from those 4-7 facts. Subject lines A/B-tested.
8 warmed-up mailboxes, rotation, a 4-step follow-up sequence. Bounce protection and spam check.
Replies go to amoCRM with sentiment tags. Positive → the SDR, negative → unsubscribed.
Apollo as the starting list
The Apollo database for the Russian market — 1.2M C-level contacts in the target segment. We segment by industry (industrial manufacturing, wholesale, retail, B2B services), exclude those already working with competitors per BuiltWith, and end up with a clean list of ~18-22k contacts per quarter of work.
Clay: 4-7 facts per lead for the AI to work with
Each contact is enriched via Clay: revenue from SPARK, recent press releases (company RSS feeds), tech stack via BuiltWith, open job postings (HR signal of growth), mentions in industry press. The goal is 4-7 concrete facts the AI can build a genuinely personal email around.
AI copy with specifics
GPT-4o quickly generates 3 variants per lead (via ProxyAPI.ru — RU card payment). Claude Opus 4.7 does the final review of the best variant, cuts boilerplate and checks every claim against the collected data. The result — an email that mentions a specific company event (new branch, CDO hire, recent release), not "let's get acquainted with your company."
Instantly — scale without blocks
8 mailboxes on different domains (main + 7 dedicated). 2 weeks of warm-up via Instantly warm-up before the main campaign launches. Rotation of 35-45 emails per day per mailbox to avoid spam filters. A/B-tested subject lines plus a 4-step follow-up sequence (after 3, 7, 14 and 28 days). Bounces cleaned automatically.
amoCRM with sentiment-tagging of replies
Each reply goes through a mini-LLM router: positive → lead to amoCRM with "ready for call" tag and straight into the SDR queue; neutral → lead tagged "needs more materials"; negative → unsubscribed and removed from the sequence. SDRs now only handle closing meetings.
Best-of-breed tools, RU card payment
Database of 1.2M C-level contacts in Russia, segmented against the ideal customer profile
Data enrichment: SPARK + RSS + BuiltWith + job postings, 4-7 facts per lead
Cold email outreach service: warm-up, rotation, cadence, bounce protection
Fast generation of 3 email variants and a subject line per lead, from the collected facts
Final review of top variant: boilerplate, facts, tone
Proxy to OpenAI / Anthropic with RU card payment — no VPN/foreign accounts
Source of company revenue and financial statements by tax ID
Tech stack: which CMS, analytics, hosting, whether competitors are already in
CRM for managing qualified leads, sentiment tags, SDR queue
What changed in 6 weeks
+247% qualified sales pipeline
thanks to AI personalisation on 4-7 concrete facts
including all subscriptions and AI tokens
just quality control + closing meetings
setup + recurring (subscriptions and maintenance)
The real change isn't "more leads", it is what the SDRs now spend their day on. The SDR used to be an operational role (research, copy, sending); now it is a consultant at closing meetings. The team didn't grow, it shifted toward the higher-value part of the funnel.
A second effect — new niches can now be tested without growing the team. Launching a campaign in a new industry is a change of Apollo filters and a rebuild of the Clay data set — days, on the existing team.
When an automated outbound system beats manual SDRs
Universal pattern — "selling to C-level in mid-market+, deal cycle > 1 month, average ticket > ₽500k". Wherever the economics tolerate ₽3-5k per qualified lead and AI personalization delivers a quality edge over templated mailings:
- → B2B SaaS with a ₽500k+/yr ticket — a well-defined customer profile and clear buying triggers
- → Consulting / agencies — digital transformation, automation, custom development
- → Corporate services — legal, audit, valuation, HR consulting for mid-market
- → Industrial tech — IoT, MES, WMS — where the target contact is the production / IT director
- → FinTech B2B — acquiring, factoring, banking products for mid-sized business
- → Any B2B service where "hire another SDR" means hiring 6+ people to reach the desired output
- A workshop to define the ideal customer profile, then Apollo segmentation against those criteria
- Clay enrichment: 4-7 facts per contact for the AI to write from
- GPT-4o prompts + Claude review + A/B subject framework
- Mailbox infrastructure: 8 boxes, 14-day warm-up, follow-up sequence configured
- amoCRM integration: sentiment tags, SDR queue, sync with Instantly
If SDRs are busy with research instead of closing meetings — this turns around in 6 weeks
₽420k setup + 95k/mo. Mailbox warm-up and first qualified leads — by week 6. Suitable for B2B services with ticket > ₽500k and deal cycle > 1 month. Includes team training on running the system.
Related cases
Keitaro + CAPI: conversion attribution 62% → 96%
Keitaro tracker as single source of truth. CAPI Gateway via nginx + Lua dedup. Server-side fan-out with…
Speech-to-Creative: speech recognition + language swap + lip-sync
Whisper transcribes speech → GPT-4o translates into 5+ languages and rewrites the lines for the on-screen…
GMBSPYLAB — collecting the public FB Ad Library into my own Telegram feed
Async Python and Postgres 16: 30–75k raw cards a day, perceptual dedup leaves 5–12k unique ones, arq queues…
The 9,900 ₽ audit — with a concrete report and quote
I'll tell you what to deploy in your business first, what the payback looks like, and whether you need AI for the task at all (sometimes you don't).
Or just send your question — I reply within 2 hours