How to Build an AI Sales Outreach Squad: Personalized Cold Emails at Scale

Cold outreach is broken. Sales reps spend 40% of their day writing emails that sound like templates, while the "personalized" tools produce messages that feel robotic and generic. The math is brutal: a good SDR can research and write maybe 25 personalized emails per day. At scale, most teams default to mail merge templates with a first name swap and hope for the best.

But what if you could deploy a team of AI agents -- one that researches each prospect in depth, one that writes a genuinely personalized email, and one that plans a full follow-up sequence -- all working together in a coordinated pipeline? With Ivern AI, you can build exactly that. This guide walks you through setting up a multi-agent sales outreach squad that produces 50+ highly personalized cold emails for $1-2 in API costs using your own model keys.

Why Multi-Agent AI Beats Traditional Sales Automation

Sales automation platforms like Instantly, Apollo, and Outreach.io excel at sending volume. They warm up domains, rotate sending accounts, and track opens. But their personalization is surface-level: first name, company name, maybe a recent news mention if you pay for the data add-on.

Multi-agent AI workflows are fundamentally different. Instead of filling in template variables, each agent specializes in one task and hands off its output to the next agent. The result is outreach that reads like a human spent 20 minutes on each prospect -- because the AI actually did spend computational effort researching, reasoning, and writing.

Here is where the approach diverges from traditional tools:

Capability	Sales Automation Tools	Multi-Agent AI Workflow
Personalization depth	Variable swaps (name, company)	Full contextual research and custom messaging
Prospect research	Manual or basic enrichment	Automated deep research per prospect
Email writing	Templates with merge fields	Unique emails per prospect
Follow-up sequences	Generic drip campaigns	Context-aware follow-ups referencing specific details
Cost per personalized email	$0.50 - $3.00 (platform + data costs)	$0.02 - $0.05 (API costs only)
Setup time	1-2 weeks	30-60 minutes

The key advantage is not cost -- it is quality. When a prospect reads an email that references their specific product launch, quotes their LinkedIn post, and connects it to a real use case, they respond. Generic templates get deleted.

The 3-Agent Sales Outreach Squad

Your outreach squad consists of three specialized agents, each with a clear role. You can run all three in a single Ivern AI workflow, passing outputs between agents automatically.

Agent 1: Prospect Researcher

Role: Gathers detailed information about each prospect and their company to feed into the email writer.

Recommended model: GPT-4o-mini or Claude 3.5 Haiku (fast, cheap, good at structured data extraction)

System prompt:

You are a sales research analyst. For each prospect provided, produce a structured research brief with the following sections:

1. COMPANY_SNAPSHOT: Company name, industry, size (employees and revenue estimate), product/service summary, target market.
2. RECENT_ACTIVITY: Any recent news, product launches, funding rounds, hires, or partnerships. Note the source and date if available.
3. PROSPECT_PROFILE: Role, likely responsibilities, potential pain points based on their position and industry.
4. ICE_BREAKERS: 2-3 specific, verifiable details that could be referenced in a cold email to show genuine research (e.g., recent LinkedIn post topic, conference speaking engagement, product feature they launched).
5. ANGLE: A one-sentence hypothesis about why your product (Ivern AI - multi-agent AI workflow platform) would be relevant to this prospect specifically.

Format your output as clean, structured text. Be factual. Do not fabricate information. If you cannot verify a detail, note it as [UNVERIFIED].

Agent 2: Email Writer

Role: Takes the research brief and writes a personalized cold email that feels human and specific.

Recommended model: GPT-4o or Claude 3.5 Sonnet (strong writing quality, follows tone instructions well)

System prompt:

You are an expert B2B cold email copywriter. Using the prospect research brief provided, write a personalized cold email with the following constraints:

- Subject line: 3-7 words, curiosity-driven, references something specific to the prospect
- Email body: 80-150 words maximum
- Opening line: Reference a specific, researched detail about the prospect or their company (never a generic compliment)
- Value proposition: One clear, concrete benefit tied to their specific situation
- CTA: A low-friction ask (e.g., "Worth a 10-minute chat?" or "Mind if I send over a quick example?")
- Tone: Conversational, confident, not salesy. Write like a peer, not a vendor.
- No exclamation marks. No buzzwords ("revolutionary", "game-changing", "synergy"). No "I hope this email finds you well."

The email should feel like it was written by someone who spent 15 minutes researching this specific person. Because it was -- by the research agent.

Output format:
SUBJECT: [subject line]
BODY:
[email body]

Agent 3: Follow-Up Planner

Role: Designs a 3-5 email follow-up sequence, each building on the original email with new angles and value.

Recommended model: GPT-4o-mini or Claude 3.5 Haiku (structured planning task, does not need the strongest model)

System prompt:

You are a sales follow-up sequence designer. Given the original cold email and prospect research brief, create a 4-email follow-up sequence.

For each follow-up email, provide:
- TIMING: When to send (e.g., "3 days after initial email", "1 week after follow-up 2")
- ANGLE: The specific approach or value prop for this email (each should differ -- e.g., case study, data point, question, resource share)
- SUBJECT: Subject line (3-7 words, can reference previous email)
- BODY: Email body (60-100 words, shorter than the original)

Rules:
- Follow-up 1: Add a relevant case study or data point
- Follow-up 2: Ask a specific question related to their role/challenges
- Follow-up 3: Share a useful resource (no ask, pure value)
- Follow-up 4: Break-up email -- short, respectful, leave the door open
- Never repeat the same angle or subject line
- Each email must be able to stand alone (prospects may not have read previous emails)

Setup Instructions: Building the Workflow in Ivern AI

Setting up your sales outreach squad takes about 30 minutes. Here is the step-by-step process:

Step 1: Create Your Ivern AI Account

Sign up at Ivern AI. Ivern AI uses a BYOK (Bring Your Own Key) model, which means you connect your own OpenAI, Anthropic, or Google API keys. You pay only for the tokens you use -- no per-seat fees, no per-email charges, no markup on API costs.

Step 2: Add Your API Keys

Navigate to Settings and add your API keys. For the sales outreach squad, you will want at least an OpenAI key (for GPT-4o and GPT-4o-mini) or an Anthropic key (for Claude 3.5 Sonnet and Haiku). Having both gives you flexibility to assign the best model to each agent.

Step 3: Create the Three Agents

Create three separate agents in your Ivern AI workspace, one for each role described above. Paste the system prompts exactly as provided, adjusting the ANGLE section in the Prospect Researcher to reference your actual product instead of Ivern AI.

Step 4: Build the Workflow

Create a new workflow and connect the three agents in sequence:

1. Input node: Accept a list of prospects (name, title, company, LinkedIn URL, or website)
2. Prospect Researcher agent: Processes each prospect
3. Email Writer agent: Receives the research brief as input
4. Follow-Up Planner agent: Receives both the research brief and the written email
5. Output node: Compiles everything into a structured document (one entry per prospect)

Step 5: Prepare Your Prospect List

Create a spreadsheet or CSV with your prospect data. Minimum fields:

• Full name
• Job title
• Company name
• Company website
• LinkedIn profile URL (optional but improves research quality)

Step 6: Run the Workflow

Upload your prospect list and run the workflow. For 10 prospects, the entire pipeline takes roughly 3-5 minutes. For 50 prospects, expect 15-25 minutes depending on the models you selected.

Real Workflow Example: 10 SaaS Prospects

Let us walk through a concrete example. Imagine you are selling Ivern AI to VP-level prospects at mid-market SaaS companies.

Your prospect list:

Name	Title	Company
Sarah Chen	VP of Sales	CloudMetrics (Series B, 200 employees)
Marcus Johnson	Head of Revenue Ops	StackSync (Series A, 80 employees)
Elena Rodriguez	Director of Sales Development	Funnelwise (Seed, 35 employees)
James Park	VP of Growth	RelayCRM (Series C, 450 employees)
Aisha Patel	Head of Outbound	PipelineIQ (Series B, 150 employees)
Tom Wheeler	SDR Manager	Basecamp Analytics (Series A, 60 employees)
Priya Sharma	VP of Marketing & Sales	NexaFlow (Series B, 180 employees)
David Kim	Director of Business Development	OrbitData (Series A, 90 employees)
Rachel Foster	Head of Demand Gen	SalesBridge (Series B, 120 employees)
Carlos Mendez	Chief Revenue Officer	UpsellPro (Series C, 350 employees)

Running the workflow:

The Prospect Researcher agent processes each person, producing research briefs. For Sarah Chen at CloudMetrics, it identifies that CloudMetrics recently launched a new analytics dashboard, raised a $25M Series B, and her LinkedIn posts frequently discuss outbound scaling challenges.

The Email Writer takes that brief and produces:

Subject: Scaling outbound without scaling headcount?

Sarah -- saw CloudMetrics just shipped the real-time dashboard. Congrats on the launch.

Quick question: with the Series B behind you and the team growing, how are you handling outbound personalization at scale? Most VP Sales I talk to at this stage are stuck between hiring more SDRs or settling for template blasts.

We built a way to run AI agent squads that research and write personalized outreach -- each email reads like a human spent 15 minutes on it. One of our users generated 200 personalized emails last week for under $4 in API costs.

Worth a 10-minute chat?

The Follow-Up Planner then creates four follow-ups: one with a relevant case study, one asking about her specific scaling challenge, one sharing a resource on AI-assisted outbound, and a break-up email.

Cost Breakdown

Here is the approximate cost to run the full 3-agent pipeline using BYOK pricing through your own API keys:

Component	Model	Tokens per Prospect	Cost per Prospect
Prospect Researcher	GPT-4o-mini	~2,000 input + 800 output	$0.003
Email Writer	GPT-4o	~1,500 input + 400 output	$0.015
Follow-Up Planner	GPT-4o-mini	~1,800 input + 1,200 output	$0.004
Total per prospect			$0.022

Cost scenarios:

Volume	Total API Cost	Cost per Email (including follow-ups)
10 prospects (50 emails total)	$0.22	$0.004
50 prospects (250 emails total)	$1.10	$0.004
200 prospects (1,000 emails total)	$4.40	$0.004
500 prospects (2,500 emails total)	$11.00	$0.004

These costs assume GPT-4o for the writer and GPT-4o-mini for the other agents. Using Claude 3.5 Sonnet for the writer increases the cost by roughly 2x but some teams prefer its writing style. With Ivern AI's BYOK approach, you control the model choice and cost for every agent.

How This Compares to Sales Tools

Feature	Ivern AI (BYOK)	Instantly	Apollo	Outreach.io
Monthly cost	API costs only ($5-20 typical)	$37-100/mo	$49-119/mo	$100+/mo per user
Per-email cost (personalized)	$0.02-0.05	$0.50-1.50	$0.30-1.00	$1.00-3.00
Personalization depth	Full research + custom writing	Variable swaps	Variable swaps + basic AI	Templates + snippets
Prospect research	Built-in agent	Separate tool needed	Built-in database	Separate tool needed
Follow-up planning	Context-aware AI agent	Basic sequences	Basic sequences	Rule-based sequences
Email sending	Not included (use your own tool)	Included	Included	Included
Customization	Full control over prompts and models	Limited	Limited	Moderate

The tradeoff is clear: Ivern AI handles the intelligence layer (research, writing, planning) at a fraction of the cost, while you use your existing email sending infrastructure (Gmail, Instantly, or any SMTP tool) to actually deliver the emails.

Tips for Better Sales Outreach Output

1. Feed better inputs to the Prospect Researcher. The more context you provide (LinkedIn URLs, recent news links, company website), the more specific and accurate the research will be. A prospect with a LinkedIn URL gets 3x better ice-breakers than one with just a name and company.

2. Adjust the Email Writer's tone for your ICP. The default prompt targets VP-level buyers. If you are selling to developers, change the tone instructions to "technical, direct, no fluff." If selling to C-suite, shift to "strategic, concise, bottom-line focused."

3. Add negative examples to the prompt. Include 2-3 examples of bad cold emails in your system prompt and explicitly tell the agent to avoid those patterns. This dramatically improves output quality.

4. Human-review a sample before bulk sending. Run the workflow on 5 prospects first. Review the output, identify patterns you want to adjust, and refine the prompts. Then run the full batch.

5. Rotate models for variety. If you are sending 200+ emails, run half through GPT-4o and half through Claude 3.5 Sonnet. This produces natural variation in writing style that prevents your outreach from having a recognizable "AI tone."

6. Use the Follow-Up Planner for A/B testing. Generate two different follow-up sequences and test them against different prospect segments. The data from your CRM will tell you which angles convert best.

FAQ

Can I use this for LinkedIn outreach instead of email?

Yes. Modify the Email Writer system prompt to specify LinkedIn DM format (shorter, more casual, 300 character limit for connection requests). The research and follow-up agents work the same way.

How do I actually send the emails?

Ivern AI generates the email content. For sending, use your existing infrastructure -- Gmail with GMass, Instantly, Lemlist, or any email sending tool. Export the generated emails from Ivern AI and import them into your sender. This separation keeps costs low and gives you full control over deliverability.

What if the Prospect Researcher hallucinates information?

The system prompt explicitly instructs the agent to flag unverifiable information as [UNVERIFIED]. Always review research briefs before sending emails. For high-stakes outreach (enterprise accounts, C-suite), manually verify ice-breaker details. The agent gets it right 85-90% of the time, but that 10-15% error rate matters when you are emailing a Fortune 500 CRO.

Can I add more agents to the squad?

Absolutely. Common additions include a Compliance Checker agent that reviews emails for CAN-SPAM/GDPR compliance, a Subject Line Scorer that ranks subject lines by predicted open rate, or a CRM Updater agent that logs all activity to your Salesforce or HubSpot instance. Ivern AI supports unlimited agents in a single workflow.

Is BYOK really cheaper than subscription tools?

For personalized outreach, yes -- dramatically so. A team sending 500 personalized emails per month with follow-ups (2,500 total emails) pays roughly $11 in API costs with Ivern AI BYOK. The same volume on Outreach.io costs $100-300 per seat per month. The savings scale linearly: more emails means more savings.

Get Started

Ready to build your AI sales outreach squad? Sign up for Ivern AI, connect your API keys, and deploy your first multi-agent workflow in under 30 minutes. With BYOK pricing, you only pay for the tokens you use -- no per-seat fees, no hidden charges, no markup.

Related Posts

• AI Agent Sales Outreach Workflow
• AI for Sales Teams: Complete Guide
• How to Automate Workflows with AI Agents: Complete Guide
• AI Agent Workflows: 10 Examples
• Free AI Agent Tools
• Compare AI Agent Platforms