The Ability Email Outreach Engine: 1,000 human-quality emails a day, without living in spam

Executive summary

Your instinct is right: small, spread-out volume across many inboxes, human-like variation, verified data and tight protocols is exactly how serious outbound teams operate in 2026. The plan below keeps your architecture and fixes the five things that would otherwise sink it.

Planning math (per business day)

Reply and meeting ranges are planning assumptions, not promises. They will be replaced with Ability's real numbers after the first 5,000 sends.

Review of your plan

Point by point against your brief, what stays, what changes and why.

Domain health audit: abilityglobal.io

Live DNS lookups run on 16 Sep 2026 against public resolvers. This is what receiving servers see today.

Fix abilityglobal.io, step by step

Target: 10/10 on mail-tester, SPF, DKIM and DMARC all passing and aligned, DMARC at reject within 45 days. Owner: whoever holds the Microsoft 365 admin and Cloudflare logins.

1. Freeze any bulk or cold sending from abilityglobal.io. From today this domain sends only person-to-person mail: clients, partners, warm replies, proposals.
2. Enable DKIM in Microsoft 365. Go to security.microsoft.com › Email & collaboration › Policies & rules › Threat policies › Email authentication settings › DKIM. Select abilityglobal.io, toggle on, and copy the two CNAME values shown.
3. Add the two CNAMEs in Cloudflare (DNS only, grey cloud, not proxied). They will look like this; use the exact values Microsoft gives you:

   selector1._domainkey  CNAME  selector1-abilityglobal-io._domainkey.<tenant>.<x>-v1.dkim.mail.microsoft
   selector2._domainkey  CNAME  selector2-abilityglobal-io._domainkey.<tenant>.<x>-v1.dkim.mail.microsoft

   Return to the DKIM page and switch "Sign messages for this domain with DKIM signatures" to Enabled. Status should read "Signing DKIM signatures for this domain". PowerShell alternative: Set-DkimSigningConfig -Identity abilityglobal.io -Enabled $true, and confirm 2048-bit keys with Get-DkimSigningConfig.
4. List every system that sends as @abilityglobal.io (website contact form, HRMS, CRM, invoicing, Novara HR notifications). Each one either gets its own SPF include and DKIM, or moves to a subdomain such as notify.abilityglobal.io. Keep SPF under 10 DNS lookups.
5. Point DMARC reports somewhere you can read. Sign up for a free DMARC monitor (Postmark DMARC digests, EasyDMARC or dmarcian) and replace the record:

   _dmarc  TXT  "v=DMARC1; p=quarantine; rua=mailto:<your-monitor-address>; adkim=r; aspf=r; fo=1"

6. Watch reports for 2 to 4 weeks. When 100% of legitimate sources pass aligned, move to p=reject.
7. Add TLS reporting and MTA-STS (Cloudflare or your DMARC tool can host the policy file):

   _smtp._tls  TXT  "v=TLSRPTv1; rua=mailto:<your-monitor-address>"
   _mta-sts    TXT  "v=STSv1; id=20260916"

   Start the MTA-STS policy in mode: testing, then enforce.
8. Verify the domain in Google Postmaster Tools (TXT record) to see Gmail spam rate and compliance status.
9. Run the tests. Send from your mailbox to mail-tester.com (target 10/10), check MXToolbox Domain Health and Blacklist, and send to a personal Gmail and Outlook.com to confirm inbox placement and "dkim=pass header.d=abilityglobal.io" in the headers.
10. Clean up trust signals. Add a business address to the website footer, keep a professional email signature with no images or tracking links, and make sure the Microsoft 365 tenant has outbound spam policies at defaults (no forwarding rules to external addresses).
11. Optional after 45 days: BIMI with an SVG Tiny logo; a VMC or CMC certificate is needed for Gmail to show it.

Domain and inbox architecture

Your structure, sized for 1,000 per business day with headroom and a warm reserve so a burned domain never drops daily volume.

The 2:1 Microsoft to Google split is a starting assumption based on who Ability sells to. After the first 5,000 verified leads, Claude tags each lead's mail provider from its MX record and the split is rebalanced to match.

Domain naming rules

• Brand-adjacent, short, readable aloud, .com first, then .io or .co. No numbers, no stacked hyphens, nothing that looks like phishing.
• Pattern examples to check for availability: abilityteams.com, getability.io, withability.co, abilityhq.com, meetability.com, teamability.io, abilityengineering.com, abilitycyber.com, abilityops.com, joinability.com.
• Every sending domain 301-redirects to abilityglobal.io, so a prospect who types it in lands on the real site.
• Prefer domains at least a few months old if a clean aged option is available; otherwise register fresh and respect the 3-week warmup.
• Two inboxes per domain on real-person names used by real team members or clearly assigned personas who can take a call, for example firstname@ and rachel@. Display name "First name, Ability" style. Each inbox has a profile photo and a plain-text signature.

Lifecycle of a sending domain

Sending domain setup, for each of the 22 domains

1. Register the domain (GoDaddy is fine, Cloudflare Registrar is cheaper at renewal). Turn on auto-renew and WHOIS privacy.
2. Move DNS to Cloudflare so all domains are managed in one account with API access. Your infrastructure provider may do this for you.
3. Create the inboxes on native Microsoft 365 or Google Workspace tenants, not shared SMTP. If using a provider such as Mailscale, confirm in writing that inboxes are native Microsoft or Google and on a tenant dedicated to Ability.
4. Publish records. Microsoft 365 domain:

   @                     MX     0 <domain-with-dashes>.mail.protection.outlook.com
   @                     TXT    "v=spf1 include:spf.protection.outlook.com -all"
   selector1._domainkey  CNAME  (value from Microsoft Defender DKIM page)
   selector2._domainkey  CNAME  (value from Microsoft Defender DKIM page)
   _dmarc                TXT    "v=DMARC1; p=none; rua=mailto:<monitor>; fo=1"

   Google Workspace domain:

   @                     MX     1 smtp.google.com
   @                     TXT    "v=spf1 include:_spf.google.com -all"
   google._domainkey     TXT    (2048-bit key from Admin console › Apps › Gmail › Authenticate email)
   _dmarc                TXT    "v=DMARC1; p=none; rua=mailto:<monitor>; fo=1"

5. Microsoft only: enable SMTP AUTH per mailbox (it is off by default on newer tenants) or connect via OAuth in the sequencer.
6. Set up forwarding: Cloudflare redirect rule, https://domain/* → https://abilityglobal.io (301).
7. Custom tracking domain: create a CNAME for the sequencer's tracking host but keep open and click tracking off for cold campaigns.
8. Mailbox hygiene: profile photo, signature, time zone, and no auto-forwarding rules. Send a few genuine emails between team members in week 1.
9. Connect to PlusVibe, tag each inbox with its pool (MS, GW, Reserve) and domain, and start warmup at 5 per day with ramp-up enabled and reply rate around 30% to 40%.
10. Test: mail-tester 10/10 per domain before the domain is allowed into any campaign. After 14 days of clean DMARC reports, set p=quarantine.

Recommended platform stack

One tool per job, chosen for deliverability, cost at 1,000 per day, and whether Claude can connect to it.

Prices are from 2026 vendor and review pages and change often. Confirm on each vendor's pricing page before buying.

The full system mechanism

Ten stages, each with a clear owner (AI, tool or human) and a hand-off record in the CRM. Your a to j, in order.

Data record that travels through every stage

lead_id, first_name, last_name, title, seniority, company, domain, country, city, timezone,
industry, employees, offer_line (AI-Eng | Cyber | Banking-Ops | Talent | OrviQ | Novara HR | BizAi),
icp_score (0-100), signal_type, signal_detail, signal_date, personal_line, pain_hypothesis,
email, verify_status, verify_source, mx_provider (microsoft | google | gateway | other),
campaign_id, variant_id, sending_inbox, step, last_sent_at, reply_class, reply_at,
owner, meeting_at, meeting_outcome, next_step, crm_stage, suppression_flag, source

a. Lead generation

Campaign segments (one offer per campaign, never a menu of services)

Daily sourcing workflow

1. Claude pulls from Apollo through the connector using saved ICP filters per segment (title, seniority, industry, headcount, geography), about 550 raw contacts per business day.
2. Apify actors collect signals for those companies: careers pages and job boards, news mentions, company website "about" and "services" text.
3. Claude scores each lead 0 to 100 (fit 40, seniority and budget 25, signal strength 25, data quality 10). Below 55 is dropped; 55 to 70 gets a segment-level email; above 70 gets a researched personal line.
4. Dedupe and suppress against: existing clients and partners, open deals, the LinkedIn warm list being worked by hand, anyone contacted in the last 90 days, and the global unsubscribe list.
5. Hand to verification as a CSV or directly through the n8n flow.

b. Recipient check and verification

After verification Claude looks up each company's MX record and tags mx_provider. PlusVibe then routes Microsoft and gateway recipients to the Microsoft pool and Google recipients to the Google pool. Target: under 2% bounce per campaign and under 1% on the main pools.

c. Content and human-like variation

Copy rules for every cold email

Three layers of variation

1. Spintax on greetings, openers, bridges, calls to action and sign-offs, so each email is a different string of text.
2. Variants: 3 to 4 genuinely different angles per campaign step (problem-led, signal-led, proof-led, question-led), rotated by PlusVibe.
3. Personal line written by Claude from the lead's signal for leads scoring above 70.

Before any variant goes live, Claude runs a copy-quality gate: word count, reading level, banned words, link count, spam-word scan, no em dashes, and a "would a busy CISO read past line one" critique.

Sample sequence: AI and engineering teams (US CTO)

Sample opener: cyber security (UAE CISO)

d. Sending behavior and inbox rotation

Ramp schedule (per business day)

Each step up happens only if the previous week passed every KPI in section 19. A failed week repeats.

e. Deliverability protocols and test gates

f to h. Replies, triage and follow-ups

Every reply lands in the PlusVibe unibox. The AI reply agent classifies it; a webhook sends it to n8n, which calls Claude for a draft, writes to the CRM and alerts the right person.

Follow-up architecture

• Core sequence: day 0, day 3, day 7, day 14 (business days), same thread, a new angle each time, stop on any reply.
• Post-reply follow-ups: interested but no booking gets a human follow-up at day 2 and day 5, then weekly for 3 weeks.
• Nurture: "not now" leads get one relevant, non-salesy note every 60 to 90 days (a relevant insight, a new product launch), from the same person.
• Cross-channel: for scores above 80, the same person sends a LinkedIn connection request after email step 2, following Ability's LinkedIn doctrine (no pitch in message one).

i to j. Meetings, scope and next steps

1. Booking. Cal.com or Calendly link with round-robin across meeting owners by segment (for example Management for US CTO and CISO, Rachel for UAE). Confirmation and a reminder 24 hours and 1 hour before, sent from the main abilityglobal.io calendar.
2. Move the relationship to the main domain. The meeting invite comes from the owner's @abilityglobal.io address. From here the conversation lives on the brand domain.
3. Pre-meeting brief (Claude, 24 hours before). One page: company snapshot, the signal that triggered outreach, the email thread, likely pains, relevant Ability offer and bench profiles, 5 discovery questions, and red flags.
4. The call. Recorded and transcribed with consent (Teams or Fireflies).
5. Post-meeting package (Claude, within 1 hour). Summary, needs, budget and timeline signals, decision makers, a thank-you email draft, and a first-pass scope outline: objectives, team shape, phases, assumptions, open questions.
6. Human review. Meeting owner edits and sends the follow-up the same day; delivery lead reviews the scope within 2 business days.
7. Proposal. Claude builds the proposal document or deck from the approved scope in Ability's format (light theme, Ability logo, PDF copy).
8. CRM stages: Replied › Meeting booked › Met › Scoping › Proposal sent › Negotiation › Won or Lost, with a next step and date on every open deal. Claude flags any deal with no activity for 7 days.

The Claude layer: connectors, skills and scheduled tasks

Claude does the work that needs judgment and writing. Deterministic plumbing (sending, webhooks, CRM writes) stays in PlusVibe and n8n so nothing depends on a chat window being open.

Connectors to set up in this project

Custom Claude skills to build

Scheduled tasks

• 07:00 PKT daily: deliverability report (bounce, replies, placement, any inbox over thresholds) with recommended pauses.
• 08:00 PKT daily: lead batch for the day: source, score, research, write personal lines, output a verified upload file.
• Twice daily: reply digest of anything the AI agent marked uncertain, plus unanswered interested replies.
• Monday: weekly campaign review (variant winners and losers, segment performance, next week's copy tests) and DNS and blocklist sweep of all 22 domains.
• 1st of month: ICP refresh from won and lost deals, and inbox retirement review.

Claude API inside n8n (real time)

PlusVibe webhook (reply received)
  → n8n: fetch thread + lead record
  → Claude API: classify (8 classes) + confidence + draft reply
  → if confidence ≥ 0.85 and class in [OOO, not interested, unsubscribe]: act automatically
  → else: post to Slack #replies with draft + Approve / Edit buttons
  → on approve: PlusVibe API sends reply from same inbox; HubSpot updated

Human in the loop: roles

90-day rollout

KPIs and auto-pause rules

Estimated costing analysis

What it costs to build the Email Outreach Engine in 20 days, run it at 1,008 emails a day, and what each lead, reply and meeting costs. All figures are USD estimates from 2026 public pricing and planning assumptions; confirm vendor quotes before purchase.

1. One-time setup costs (20-day build)

2. Monthly tool and infrastructure costs

Where the base monthly tool budget goes

3. Team time (internal cost)

Hours are from the operating model in section 17. The hourly rate is a placeholder assumption for a fully loaded team cost; replace it with Ability's actual figures.

4. Total cost by scenario

5. Unit economics at full volume (base case)

Volume: 1,008 emails a day for 22 business days, about 8,500 new leads a month. Reply and meeting rates are the planning ranges from section 1, not guarantees. All-in means tools plus team time ($3,651 a month).

6. Break-even view

Ability's engagements (AI teams, cyber services, banking operations, OrviQ) are high-value, so a single win typically covers the engine for months. Months of base all-in running cost ($3,651) covered by one closed deal:

7. Scaling to 1,500 emails a day

Compliance by market

This is a practical summary, not legal advice. Have counsel confirm before launching in each market.

Risks and mitigations

Suggested service providers

One recommended vendor for each part of the engine, with alternatives and indicative 2026 prices. Your chosen tools (Apollo, Claude + Apify, MillionVerifier, NeverBounce, Mailscale, PlusVibe, mail-tester, Claude) stay the first pick. Prices move often, so confirm with each vendor before buying.

Master checklist