Cold Outreach Sequence: Research → Emails → Follow-Ups

What this chain helps you accomplish

Turn an individual prospect and a handful of research inputs into a coherent, human-reviewed four-touch B2B email sequence. Instead of asking AI to invent a trigger or write disconnected templates, you build an evidence register, select a defensible messaging angle, draft an initial email, add differentiated follow-ups, and prepare a low-pressure close-out.

The deliverable is a sequence pack with source-mapped personalization, email copy, conditional touch timing, explicit stop rules and a send-readiness checklist. It is designed to help SDRs and AEs write more relevant outreach without turning guesses into claims about a prospect.

Promptifi provides the prompts and instructions; the sequence is drafted in your own AI assistant. Nothing is sent or entered into a CRM automatically.

When to use it

• Starting one-to-one outbound to a named account where you want a coordinated sequence rather than four unrelated email drafts.
• Preparing an SDR calling/emailing block, a targeted AE expansion of prospecting, or a small account-based outreach list with available business context.
• After a relevant public business development or role-specific signal has been checked—but also when no reliable trigger exists and you must use a cautious role-level angle.
• Rebuilding underperforming generic emails with tighter evidence labels and more useful follow-up angles.
• Before moving an approved draft to an email-sequencing system operated under your employer's policies.

Do not use the chain to contact suppressed or opted-out recipients, mass-generate prospect personal data, bypass sending rules, or replace legal/compliance review. This is not an automation and does not monitor inbox replies, suppression updates, or whether messages were actually delivered. If the person responds, the sequence should stop and the seller should handle the response individually.

Before you begin

Tools and access

Required: a text-based AI chat service authorized by your employer, such as ChatGPT, Claude, Gemini or Microsoft Copilot; your approved contact/context source; and a way to copy drafts into your normal email system manually. Paid AI features, web browsing, CRM access, mail integrations and automated sequencing are not required. Research links must be opened using genuinely available, permitted browsing tools before a model may call their contents verified.

Optional: permission-controlled public research, an approved CRM to check last touches and suppression, and an existing email sequencer to load the finished copy later with human authorization. None is silently assumed.

Starting context inputs

• {{ACCOUNT_NAME}} — Required: company/legal business name from your approved list or CRM. Example: Cedarline Components (fictional). If uncertain, verify before targeting.
• {{PROSPECT_NAME}} — Required to send; optional for drafting: person listed in your approved contact source, e.g. Dana Ruiz (fictional). If not verified, draft with a generic role and HOLD sending.
• {{PROSPECT_ROLE}} — Required: known business title/function from an authorized record, e.g. VP of Operations. If unclear, label unconfirmed and do not invent responsibilities.
• {{INDUSTRY}} — Optional: verified/known category such as industrial distribution. Use unknown if uncertain.
• {{SOLUTION_SUMMARY}} — Required: accurate 1–3-sentence product/service description from approved collateral, including who it helps and what it actually does. If missing, pause drafting.
• {{ICP_CRITERIA}} — Required: your own fit criteria, such as industry, region, customer size, use case and exclusions; use only criteria you know. Example: US distributors with distributed warehouse operations; exclusions = direct competitors.
• {{APPROVED_PROOF_AND_ASSETS}} — Optional: permissioned case studies, links, proof points and their exact scope/source. If absent, avoid named customers, improvement percentages or implied results.
• {{RESEARCH_EVIDENCE}} — Optional: permissible public news, job postings, official site extracts or approved CRM notes; include exact claim, source/title, URL or document ID, published date and checked date. If none, provide 'none'; the sequence must use role-relevant questions.
• {{RELATIONSHIP_AND_PRIOR_TOUCHES}} — Required: actual known relationship and previous outreach from CRM/inbox, including dates and response/opt-out if available; 'none confirmed' is valid but does not prove no earlier touches.
• {{DESIRED_NEXT_STEP}} — Required: one realistic low-friction step, e.g. 'ask if comparing renewal processes is relevant'; not an assumed meeting.
• {{SENDER_IDENTITY}} — Required: accurate sender name, organization and approved signature identity. Do not impersonate a colleague or invent credentials.
• {{OUTREACH_JURISDICTION}} — Required before sending: countries/regions relevant to sender and recipient and applicable company review; e.g. 'US only, company compliance approval required'. If uncertain, HOLD and ask compliance.
• {{SUPPRESSION_STATUS}} — Required before sending: checked against approved suppression and do-not-contact records; use CLEARED / BLOCKED / NOT VERIFIED with check date. NOT VERIFIED means HOLD.
• {{SENDING_POLICY_AND_FOOTER}} — Required before sending: employer-approved email policy, identity/advertisement notices where applicable, valid sending address, functional opt-out mechanism, approved company postal address and signature. Missing values require HOLD, not guessed details.
• {{CADENCE_CONSTRAINTS}} — Optional: touch caps, time windows, send days, prohibited claims, allowed channels. Sample four-touch plan is illustrative, not an optimum. If none, use suggested timing for drafting only.
• {{DATA_SHARING_RULES}} — Required: which AI tool/data is approved, source restrictions and whether browsing is permitted. Example: 'public corporate pages only; do not paste private phone numbers or unpublished CRM notes'.

Research and safety preparation

1. Gather 1–3 relevant, permitted sources when possible and record a specific claim, its source and date. Prefer first-party corporate disclosures, role-specific postings, or explicitly approved materials; do not claim a URL was read just because it was pasted.
2. Check the actual prospect record for relationship, past touches, recent replies, hard bounces and any do-not-contact or opt-out requirement. Unknown suppression is a hold, not permission.
3. Prepare an honest seller/product statement and any authorized proof. Remove confidential customer or personal data that your AI tool is not approved to receive.
4. Identify the recipient jurisdiction and use your approved sending policy and compliance footer. In the US, commercial B2B email is within CAN-SPAM's scope; review official FTC guidance rather than treating AI output as compliance approval.
5. Confirm outbound technical prerequisites with the sending system owner, including email authentication and deliverability requirements. A well-written draft does not establish domain readiness.

Reference materials for policy review: US Federal Trade Commission, CAN-SPAM compliance guide (ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business); Google, Email sender guidelines (support.google.com/mail/answer/81126). These are reference points, not a substitute for your organization's policies or advice for other jurisdictions.

How to run the chain

1. Start a new chat in a company-approved AI assistant and paste the completed STARTING CONTEXT that appears in Chain Steps. Do not include a live email address or regulated data unless permitted.
2. Run Prompt 1. Review every E# research source, whether it was actually accessed, fit exclusions and suppression flags. Edit missing or inaccurate input, then explicitly approve RESEARCH_BRIEF_v1.
3. Run Prompt 2 in the same chat with the approved RESEARCH_BRIEF. Inspect the angle and the suggested Day 0/5/10/16 sequence as a draft timing illustration, not a sending instruction. Approve ANGLE_PLAN_v1.
4. Run Prompt 3 using both approved prior artifacts. Select the primary email or role-based fallback only after checking company-specific wording. Approve INITIAL_EMAIL_v1. A marked compliance HOLD can still allow an editorial draft but not sending.
5. Run Prompt 4 to produce two genuinely different conditional follow-ups, or an explicit SKIP if credible new value is missing. Review the claims and word limits; approve FOLLOW_UPS_v1.
6. Run Prompt 5 with the four approved artifacts. Inspect the close-out, compiled sequence, evidence map, stop rules and send-readiness checklist. If any mandatory check is not verified, leave the pack on HOLD.
7. If using a new chat for a step, paste the STARTING CONTEXT and each required approved artifact verbatim. AI conversation memory is not a substitute for handoff. Use the included context template if necessary.
8. After completing the drafts, check the actual contact and suppression records again before each planned send. Manually approve and send each touch, or separately authorize an approved system to schedule it. Stop the cadence on any response or disqualifying event; the chain itself does not monitor changes.

Artifact approval and change control

Approve a version by replying, for example: APPROVED — RESEARCH_BRIEF — v1. If corrected, ask for RESEARCH_BRIEF_v2 and reply APPROVED — RESEARCH_BRIEF — v2. Replace old versions in downstream contexts. The AI may not approve its own results.

If the base research or selected angle changes, regenerate affected emails. If an approved email changes substantially, revisit dependent follow-ups and recompile the final sequence. A compliance HOLD remains active until an actual human review clears it.

What successful completion looks like

A complete, coherent and source-mapped four-touch draft pack with truthful email copy or clearly marked skipped touches, a verified stop-conditions checklist and explicit human next actions. It does not require that any email has been sent.