OUTBOUND / WORKFLOWS

How I build outbound workflows for a marketing agency

How I build outbound workflows for a marketing agency

Scarlett Vu

October 23, 2026

Here’s how I build outbound workflows for a marketing agency using Codex, Clay, AI Ark, Apollo, GetLeads, Prospeo, MillionVerifier, and Smartlead, connected through APIs. I’ll walk through how I learn the business, find and qualify prospects, write the emails, and turn replies into conversations.

The work starts well before the first email goes out. We need the right companies, the right people, accurate contact details, and a reason for someone to respond.

Here’s how I put those pieces together at Kyder.

The tools each have a job. Codex helps me research and connect the workflow. Clay organizes the research and qualification. The data providers fill specific gaps, and Smartlead runs the email sequences. I still own the targeting decisions, the writing, and the conversations.

THE STACK, AT A GLANCE

Each tool has a specific job.

Tool

Where I use it

Codex

Research, API workflows, and tightening my writing

Clay

Company research, qualification, organization, and scoring

AI Ark

Company lookalikes and contact discovery

Apollo

Contact discovery and missing email lookups

GetLeads

Remaining contacts and phone numbers for priority prospects

Prospeo

Missing email lookups

MillionVerifier

Email verification

Smartlead

Email sequences and campaign execution

1 - Learn the agency’s business inside and out before sending anything.

The first two weeks are our inbox warm-up period. While the inboxes warm up, I’m learning the agency’s business.

I take their past sales calls, customer stories, past wins, common objections, and onboarding answers, and feed that context into AI to help me work through it.

I’m looking for the details that should shape our outreach:

  • Who buys from them, and what usually prompts that decision?

  • How do customers describe the problem in their own words?

  • What have they already tried?

  • What objections come up, and how does the agency address them?

  • Why do they win deals, and why do they lose them?

  • What results can we actually back up?

I use Codex to help identify patterns across that material, then work through any gaps with the agency. Hearing the same concern come up in several sales calls tells me more than guessing from their homepage.

Those insights become the foundation for our targeting, offers, email sequences, and reply handling. We also agree on exactly what makes a meeting qualified.

Before any email gets sent, I want to understand their business inside and out: what they sell, who they help, why people buy, and where they’re strongest.

We’re going to be talking with potential customers on their behalf. We need to sound like an extension of their team, and that starts with doing the work to understand the business.

The output of those first two weeks should be a usable campaign brief: who we want to reach, why they might care, which results we can reference, and what makes a meeting qualified. That gives every later step something concrete to work from.

2 - Find similar companies with AI Ark.

I use AI Ark’s lookalike search to find businesses similar to the agency’s ideal customers.

Then I narrow the results by location, company size, industry, and the client’s requirements.

I’m selective about the companies I use as examples. An agency might have a customer that pays well but sits outside its usual expertise. That doesn’t necessarily make it the best starting point for an entire campaign.

I want examples the agency has a clear reason to win more of.

3 - Use Codex and APIs to connect the workflow.

I use Codex to help build workflows that connect the tools through APIs.

Each step checks what we already know and determines what needs to happen next:

  • A company fits our criteria → find the decision-maker.

  • The right contact is missing an email → try another provider.

  • An email is found → verify it.

  • A prospect scores highly → look for a phone number.

  • A contact is ready for outreach → route them into the appropriate campaign.

I also need a path for incomplete records. If a lookup returns nothing or the research is unclear, that record should stay flagged until we resolve it.

Otherwise, missing information quietly becomes bad targeting.

For example, finding a founder at a qualified company shouldn’t trigger every lookup we have. If we already have their email, the next step is verification. If the email is missing, we try another source. If they’re a strong fit, we also look for a phone number. The next action depends on the record.

4 - Research and qualify the companies in Clay.

I use Clay to organize the company list, enrich the records, and check whether each business fits.

For a marketing agency, that can include what the company sells, who its customers are, and whether it matches the agency’s service focus.

An agency selling ecommerce marketing needs different qualification rules from one selling video production to B2B companies.

Clear bad fits get removed. Unclear records get flagged for review.

Doing this before contact enrichment helps us avoid spending money finding people at companies we shouldn’t be contacting.

5 - Find the decision-makers with AI Ark and Apollo.

Once the companies qualify, I use AI Ark and Apollo to find the relevant people.

At a smaller business, that might be the founder or owner. At a larger company, it might be the head of marketing, VP of marketing, or someone responsible for a specific function.

The role has to match what the agency sells.

Someone being senior doesn’t automatically make them the right buyer. I want the person who owns the problem, can evaluate the service, or can bring the right people into the conversation.

6 - Use GetLeads to fill the remaining contact gaps.

The first searches won’t find everyone.

I use GetLeads to pick up remaining leads at companies where we still need relevant contacts.

As the data comes together, I check the titles and deduplicate the records. The workflow keeps track of which companies have usable contacts and which still need work.

That gives us a more complete list without treating every result from every provider as a new prospect.

7 - Give missing emails their own workflow.

Sometimes we find the right person but don’t have their email.

Those contacts move into another lookup step, using Apollo and Prospeo to help fill the gaps.

The workflow checks the existing record first and only requests what’s missing. Once an email is found, it moves to verification.

I don’t need to send every contact through every provider. I need each provider to do a specific job where the previous step left a gap.

8 - Verify the emails with MillionVerifier.

Finding an email and verifying it are separate steps.

I run emails through MillionVerifier before deciding how to route them.

Valid emails move forward. Invalid emails get excluded. Catch-all and uncertain results stay separate for additional checks.

Verification helps reduce avoidable bounces. The domains, inbox setup, sending volume, and messaging still need attention too.

9 - Score prospects against the agency’s criteria.

I use Clay to score prospects based on the fit we can support with research.

That might include company size, business model, relevance to the agency’s services, and whether we’ve identified the right decision-maker.

The score helps me prioritize where to spend more time.

I also want the reasoning behind it. “High score” doesn’t tell me much when I’m about to pick up the phone. Knowing why the company fits gives me something to work with.

And fit isn’t the same as buying intent. A company can be a strong match without actively looking for an agency.

10 - Get phone numbers for higher-scoring prospects.

For higher-scoring prospects, I use GetLeads to find phone numbers so I can call them.

By this point, I have the company research, the person’s role, and a reason we think the agency could help.

I use that context to prepare for the conversation. What might matter to this person? What can we credibly offer? What do I still need to ask?

The score tells me who to prioritize. The conversation tells me whether there’s an opportunity.

11 - Segment the list around the conversation we want to have.

I group prospects by something that changes what we should say.

An agency offering both website development and video production shouldn’t send everyone the same pitch just because they all fit the same company-size filter.

Each segment needs a relevant offer and a clear reason to consider it.

The research helps us choose that angle. If we can’t confirm a detail, we don’t turn it into a confident statement in the email.

We can ask a relevant question without pretending we know everything about their business.

12 - Write every email myself, word by word.

I draft every email myself, word by word. It doesn’t need to be perfect, but it needs a human touch. I want it to sound like something I’d actually type and send to another person.

Before writing, I work through what the buyer is dealing with. What’s taking up their time? What are they trying to improve? What would make hiring this agency worth the effort and money?

I jot those pains down and use them to build the sequence.

Each email needs a reason to exist. One might lead with a problem the buyer recognizes. Another might explain a relevant customer result. Another might offer a specific first step that’s easy to consider.

I don’t want three emails repeating the same pitch with “just following up” added to the top.

Each email in the sequence has four variations. That gives us room to test different ways into the conversation: the opener, the problem we lead with, the offer, or the question at the end.

Once I’ve got the draft down, I throw it into Codex to help tighten and revise it. Cut a few words, make something clearer, but keep my voice. I also use it to help with research and flag anything that needs checking.

Then I go back through it myself.

Would I actually say this? Is there a clear reason to respond? Am I making an assumption we can’t support? Does this sound like a person talking with another person?

My thinking is simple: people are more likely to reply when an email feels like someone put thought into talking with them. That starts with understanding what matters to the person reading it.

Here’s an example of how I might structure four variations for one segment. These are testing directions, not claims about a specific client’s campaign.

I want to know what I’m testing before launching. If I change the audience, offer, opener, and CTA at the same time, it becomes harder to tell what made a difference. I keep a note of the angle behind each variation and compare the quality of the replies, not just how many came in.

13 - Route the contacts into Smartlead.

Once the contacts are verified and the copy is ready, I move them into the appropriate Smartlead campaigns.

Before launching, I check the field mapping, sequence, sending schedule, inbox limits, and that tracking is turned off.

I also review sample emails with the actual contact data filled in.

A template can look fine until a long company name, missing field, or awkward phrase lands in it. I want to catch those problems before sending.

14 - Handle replies and use them to improve the workflow.

Once someone responds, we work the conversation.

We answer questions, check fit and timing against the agency’s criteria, and handle scheduling, confirmations, follow-ups, and rescheduling.

Those replies also tell us where the workflow needs work.

If interested replies keep coming from companies that don’t qualify, we revisit the targeting. If we’re reaching people who keep redirecting us, we review the contact criteria. If prospects misunderstand the service, we tighten the messaging.

I pay attention to what happens after the booking too. Did the person attend? Did the conversation meet the agreed criteria? What did the agency learn on the call?

The workflow is only doing its job if it leads to conversations that are useful to the business.

That’s why I connect the research, outreach, and feedback. Every campaign gives us information we can use to make the next round better.

The agency should be able to see more than a meeting count. I want to know how many calls happened, how many met the agreed criteria, and what those conversations turned into. That feedback tells us which prospects and offers deserve more attention.

If you want a workflow like this built for your business, we can talk through your market, your offer, and what a qualified conversation looks like for you. At Kyder, we handle the work up to the sales call.