Running an agency means you’re always doing two jobs at once: delivering for the clients you have, and finding the clients you’ll need next. Most agency founders know how to do both. The problem is that you can’t do both at the same time.
When a project starts, prospecting stops. When it ends, you scramble. That’s not a motivation problem or a skill problem. It’s a structural one, and the tools most people recommend don’t fix the structure.
Why Prospecting Is Harder for Agencies Than for Anyone Else
A solo founder building a SaaS product can prospect in the morning and code in the afternoon. A sales rep at a 50-person company prospects full-time because that’s their job. An agency founder is neither.
You’re simultaneously the person who does the work, the person who manages the client relationship, the person responsible for quality, and the person who’s supposed to be building the next pipeline. In practice, the last one gets dropped every time. Not because it matters less, but because it has no deadline and nobody’s holding you accountable to it today.
The result is a pattern that most agency founders recognize immediately: full pipeline, everyone’s busy, prospecting stops. Projects end, pipeline is empty, panic sets in, start prospecting again. Six weeks of catch-up, one or two new clients signed, back to full capacity, prospecting stops. Repeat.
This cycle isn’t broken by trying harder. It’s broken by removing the requirement for consistent human availability from the prospecting process.
The Stack Problem That Costs You More Than It Saves
The standard advice for agency new business is to build a stack: Apollo or Sales Navigator for lists, an enrichment tool for contact data, a sequencer for emails, LinkedIn for direct outreach, a spreadsheet to track everything. Maybe a CRM if you’re organized.
Each of these tools works. Individually. What they don’t do is work together without an operator.
Someone has to pull the list, push it to the enrichment tool, upload it to the sequencer, set up the sequence, handle bounces, check LinkedIn separately, reconcile what happened where, and update the spreadsheet. That someone is you. And when you’re delivering for clients, you stop doing it.
The stack creates the illusion of a prospecting system. What it actually creates is a prospecting obligation: a set of tools that do nothing unless you operate them daily.
What agencies need isn’t another tool. It’s something that runs the full loop on its own and surfaces only what requires a human decision.
What Changes When You Replace the Stack With an Agent
An AI prospecting agent doesn’t split the task across tools. It owns the full loop: find prospects that match your target, score them for fit, enrich their contact data, determine the best next action, write and send the message, follow up, detect and classify replies.
For an agency founder, the practical difference is this: you define your target once, configure how aggressive the outreach should be, and the pipeline builds itself in the background while you’re on a client call or finishing a deliverable.
When a prospect replies, you get notified. That’s the moment that actually needs a human. Everything before it doesn’t.
This is what separates an agent from a sequencer. A sequencer sends messages you wrote to a list you built. An agent builds the list, qualifies it, writes the messages, sends them, and keeps going.
For the full picture of how this loop works in practice, this walkthrough of what an AI agent does at each step of B2B lead generation explains the mechanics clearly.
How to Set Up LEO for Agency New Business
The first thing to understand is the context separation. Your agency has its own prospecting environment, completely separate from any client work. Concretely, this means your agency’s business context, offers, personas, and prospect list live in one workspace; if you also use LEO to prospect on behalf of clients, each client gets its own workspace with its own isolated environment. Nothing bleeds across: a prospect in your agency’s pipeline cannot appear in a client workspace, and the targeting criteria you’ve defined for winning new business don’t mix with what you’ve configured for a client campaign. If that separation isn’t in place, the business context I use to generate messages becomes ambiguous, the scoring criteria calibrate on the wrong target, and your agency’s prospect list gradually fills with profiles that match a client’s audience rather than yours. Setting this up correctly from the start is the difference between a useful system and a confusing mess.
When you sign up and go through onboarding, I ask how the product will be used: solo, as a team, or as an agency. If you select agency, you then clarify whether the first workspace represents the agency or one of its clients. For your own new business development, the workspace represents your agency.
Step 1: Define your agency’s offer clearly
Your offer is not what you do for your clients. It’s what your agency sells, to whom, and why it matters to them.
This is where most agency founders get stuck, and for a reason: you’re used to positioning your clients’ products. Positioning your own services with the same clarity takes deliberate effort.
What problem do you solve? For what kind of company? What’s the outcome they care about? What makes you different from the twelve other agencies pitching the same thing?
I ask you to describe your business context and then generate an offer from it, including key features, the primary promise, problems solved, and benefits. You review and edit until it accurately reflects what you sell. This isn’t busywork. Every message I generate later draws from this foundation. A vague offer produces generic messages. A precise offer produces messages that land.
Agencies that serve fundamentally different client types should create separate offers: one for each distinct service line or target segment. If you run a content agency that serves both B2B SaaS companies and professional services firms, those are two different offers with two different personas and two parallel prospecting flows. The B2B SaaS offer targets one set of decision-makers with one set of pain points; the professional services offer targets a different role, different industry, different message. Both run in the same workspace, each with its own persona and its own stream of prospects.
For a detailed walkthrough of how the offer and persona setup work in the broader prospecting loop, AI prospecting for founders covers the same four-step mechanism with more depth for anyone who wants the complete picture.
Step 2: Build a persona for each type of client you want
A persona defines the exact profile of the decision-maker you’re trying to reach and how you want to approach them.
For a performance marketing agency, the persona might be the CMO or Head of Growth at a mid-size e-commerce company with a team already doing paid acquisition but struggling with ROAS. For a recruitment process outsourcing firm, it might be the HR Director at a company going through a hiring surge.
Targeting criteria include roles, industries, company sizes, geography, and the specific challenges that make your services relevant. I use these to score prospects from 1 to 5 stars as I find them. Prospects scoring 3 stars or more make it into your pipeline; below that, they’re set aside.
The persona also defines how outreach should work: what channel to use first, tone, message length, and the commercial objective (book a meeting, start a conversation, drive the prospect to complete a specific signup). These preferences apply across all first contacts and follow-ups.
Step 3: Let me find and qualify prospects
Once your offer and persona are in place, I start finding prospects that match. I search, score each profile against your criteria, retain those above your minimum score, enrich their data (role, company details, LinkedIn status, challenges, recent signals), and queue them for the next action.
You can review each prospect, validate or reject them, and adjust the score threshold if you’re getting too many weak matches. Prospects you reject get archived so they don’t resurface.
This is also where I search for email addresses when they’re not available from the LinkedIn profile. Email search costs a credit only when a usable address is found. An unsuccessful search costs nothing.
Step 4: Choose your level of involvement
This is the decision that determines whether the system actually works for an agency.
Three modes are available. The default mode requires you to validate each action before it’s executed. I recommend what to do next, prepare the message, and wait for your approval. You stay in control of everything that goes out, which is useful if you have time and want to review each contact.
Semi-Auto handles discovery, scoring, and enrichment automatically. Your prospect database fills itself. You then contact the queued prospects yourself, with me preparing each message and action. This works when you have two or three sessions per week to dedicate to outreach.
Auto is what actually solves the intermittency problem. In Auto, I find prospects, enrich them, send LinkedIn invitations, send messages, follow up across LinkedIn and email within your configured limits, and detect replies. When a prospect replies, automated outreach stops immediately for that prospect and control returns to you. You only step in when the conversation is live.
For an agency founder who’s deep in a project for three weeks, Auto means the pipeline keeps moving. No daily attention required. No batch of 50 contacts waiting to be processed when you come up for air.
Both Semi-Auto and Auto are available across all plans. You configure the limits before activation: how many new prospects per day, the channels to use, the maximum number of messages per prospect, delays between messages, and the days and hours when outreach should happen. I prepare a recommended configuration. You review it, modify what you need, and activate.
The right level depends on your current situation. When you have bandwidth, the default mode or Semi-Auto keeps you close to every contact. When you’re at capacity, Auto keeps the machine running. The mode can be changed at any point, and any prospect already in the pipeline continues from where it left off.
For more detail on how each mode works and when to use which, this guide on how to prospect with AI covers the mechanics step by step.
If you want to start building your agency’s pipeline, a 14-day trial gets you through the full setup in your first session.
What Makes It Harder Specifically for Agencies
The structural problem is real, but there are a few friction points specific to agencies that are worth naming.
Positioning your own services is uncomfortable. You help clients communicate their value every day. Doing it for yourself feels either obvious (“everyone knows what we do”) or presumptuous (“we can’t claim to be better than the next agency”). The discomfort is real. I ask you structured questions about your services and generate a first draft you can react to, which is easier than starting from a blank page.
The pipeline is discontinuous by design. Project-based work means revenues come in waves. That’s not fixable. What is fixable is decoupling the pipeline from your personal availability. When Auto is running, it doesn’t know you’re in a client workshop. It keeps going.
Pitch creep. Agency founders often adapt their pitch to whatever the last conversation required. After six months of inbound referrals from the tech sector, suddenly everything is positioned for tech. When you start prospecting outside that zone, the pitch doesn’t fit and reply rates drop. Defining your offer and persona explicitly forces the clarity that prevents this.
The agency founders who get the most out of AI prospecting are not the ones who use it to send more volume. They’re the ones who use it to stay consistent. Across 2,000+ accounts, the pattern is clear: agencies and founders who run prospecting continuously at moderate volume outperform those who run intense bursts followed by silence. The data behind this, including what cadence thresholds actually move pipeline, is in this analysis of prospecting results.
Once the system is running, the output is not more time on prospecting. It’s a pipeline that doesn’t stop when you do.
