How to Scale an 8 Figure Agency 🚀

Get the Fable 5 Set up guide: https://value.8figureagency.co/fable

Most of the agencies you're competing with are already AI-forward. In this solo episode I give you a straight update on where AI actually stands for marketing agencies in summer 2026: what's changed in the last quarter, what's now table stakes, and the new economics it opens up.
The headline shift: you no longer need 125 clients and a bloated team to profit $1.5M a year. With skills, cron jobs, and a lean two-person setup, 40 clients can do it, and your life gets better, not worse. I think we're about to see a wave of lifestyle entrepreneurship unlike anything before it.
In this episode:

Why 50–60% of agencies are already AI-forward, and the real cost of falling behind
Harnesses and loops: agents now running 12–30 hours straight with no human intervention
AI strategist agents that set benchmarks and pace campaigns (the SEO-for-law-firms example)
What a "skill" actually is, and why it went mainstream in Q1 2026
Cron jobs and content briefing: running repeatable client work on a schedule
The new math: $1.5M profit on 40 clients vs. the old $7.5M / 125-client grind
My own story: from a $13M holding company to three lean companies, plus making Lead Whisperer fully autonomous
The domain-expertise goldmine: how a $2M ad agency built 99%-margin software for its own clients
Fractional COO + fractional AIO: how I can help

Takeaways:

Skills and cron jobs are the floor now, not the ceiling. Turn keyword research, briefing, and on-page into files an agent runs on a schedule.
The lifestyle agency is real. One $300K strategist plus one AI-pilled associate can run 40 accounts.
Old way to $1.5M profit: ~$7.5M revenue, ~125 clients, a big team. New way: 40 clients and a lean stack.
Your domain expertise is the edge. The problems only you can see in your niche become high-margin products.
The window is now. AI improves on quarterly cycles, so early movers compound.

Links:

Work with me (fractional COO + fractional AIO): 8figureagency.co
Book a call (I keep very little availability, usually booked ~2 weeks out)
Off-market lead syndication: lead-whisperer.com
Reach me directly: jordan@8figureagency.co (my Hermes agent reads it first, then I do)

TOOLS & RESOURCES From this Episode:
1. Retell AI — The voice agent platform Jeremy chose after testing every option
YC-backed. Open API. Plug into any LLM (GPT 5.2, Claude Sonnet). Build a custom voice in 11 Labs and pipe it right in. White-label to your clients on day one.
👉 Start here: https://dashboard.retellai.com/?via=mmo

2. Versai — The HIPAA-compliant AI SMS tool Jeremy uses for his 200+ clinics
Enterprise-grade. Sub-account architecture so every client lives under your master agency account. Text conversations pipe straight into GHL or HubSpot via n8n. This is what makes the AI feel like it's "magically" working inside the client's CRM.

3. Go AI-Native with 8 Figure Agency
If you're a 7- or 8-figure agency owner and you watched this episode thinking "I need to do this for my book of business" — this is the program. I'll help you install AI-native systems that add $1M in annualized profit to your agency.

👉 Apply here: https://value.8figureagency.co/AINative

📝 Episode Summary
If you have 50, 100, 200, or 300 clients in your book — the easiest way to add seven figures this year isn't selling new customers. It's selling white-label AI to the ones you already have.
Jeremy DuPont did it. 200+ of his 350 physical therapy clinics adopted his AI voice receptionist, AI SMS tool, and chatbot in under 5 months. $2K setup fee per client. Recurring monthly on top.
In this episode he breaks down the exact playbook. The platforms he picked. The 20-person beta. The scarcity script. The team structure. The hiccups. The upsell back into marketing services.
If you run an agency, this is the cheat code.

Chapters:

The white-label AI opportunity hiding in your existing client base
Jeremy's story: PT clinic in 2021 → sold in 2024 → 350-client agency in 3 years
Why he pivoted from "marketing agency" to "AI-first agency" in early 2025
The platforms: Versai for SMS, Retell for voice — and why HIPAA compliance ruled the shortlist
Why YC-backed tools win the AI race (and why he tested everything else first)
The weekend build: from signup to demo-ready in under 24 hours
The exact go-to-market that got 200 clients in 4 months
The script: scarcity + co-creation + demo video to existing clients
The $2K setup fee — waived for the first 20, charged for everyone after
The hiccups: SOPs, CRM piping, the one-off customizations that killed scale
The team: Head of CRM, Head of Ops, Ops Manager, 2 VAs. That's it.
The ratio of customization work per client (and why he doesn't sweat it)
The new go-to-market: lead with AI, upsell marketing
The natural upsell trigger: "Your AI's working. Nobody's calling. Let's fix that."
The SEMrush + AI health score funnel that still prints
Where to find Jeremy

💡 Key Takeaways
1. The opportunity isn't new clients — it's new revenue from old clients.
Jeremy didn't sell 200 new customers. He sold AI to 200 customers who already trusted him. Look at your book. That's your gold mine.
2. Pick platforms with open APIs and multi-LLM support.
Retell won because it lets him swap LLMs and use custom 11 Labs voices. Versai won because of HIPAA + sub-accounts. The tools you pick today need to survive the model upgrades coming tomorrow.
3. Scarcity + co-creation beats every other pitch.
"I'm taking on 10 people to help shape this." That's it. That's the script. The first 20 got the setup fee waived. Everyone after paid $2K and felt lucky to get in.
4. Sell to your highest spenders first.
They have the lead volume to make AI actually work. AI on a clinic doing 16K/mo won't move the needle. AI on a clinic spending $5K/mo on ads will print.
5. Ship the V1, fix it in flight.
Jeremy admits the first 60 days had hiccups — CRM piping, SOPs, custom requests. He didn't slow down. He signed clients and built the SOP in parallel. That's the founder's mindset.
6. Lead with the AI offer. Upsell the marketing.
Hiring an admin is brutal. AI replaces the admin. That's an easier sale than "spend $5K on ads and wait 6 months." Once you've solved the admin problem, the ads/SEO upsell is automatic 60 days later.
7. Team of five can run 200+ AI deployments.
Head of CRM. Head of Ops. Ops Manager. Two VAs. That's the whole machine. Don't over-hire.

🔗 Connect With Jeremy

Agency: patchsystem.com
Instagram: @_jeremydupont
LinkedIn: Jeremy DuPont

🔗 Connect With 8F
8figureagency.co

🛠️ The Full Stack Mentioned

Retell AI — Voice agent
Versai — HIPAA-compliant AI SMS
n8n — Middleware between AI tools and CRM
GoHighLevel / HubSpot — Client CRMs
SEMrush — Google Maps ranking audits
11 Labs — Custom voices for Retell
Claude / GPT 5.2 — LLM brains behind the agents

Start building your own agent with memory and MD files → https://value.8figureagency.co/hermesstart

MetaSwarm (open source) — Dave's agentic harness → https://github.com/dsifry/metaswarm
18 default agents, each with definitions and rubrics, plus defined workflows for software development. Three ways to use it: build agentic systems with it, point it at your own SOPs to analyze them, or study how it does agentic decomposition. (Fun fact from the episode: search "metaswarm" and it ranks right behind Meta itself.)

Dave Sifry has started nine companies. Now he's building a company that builds companies.
In this one, he hands agency owners the framework for the thing almost everyone is getting wrong with AI: memory and context. If your agency runs 50, 100, even 600 clients and information keeps falling through the cracks — stakeholders forgotten, context lost, the same questions re-answered — this is the episode.
No fluff. Dave breaks down what actually deserves to be remembered, the exact MD files that run an agent (soul.md, agents.md, heartbeat), and how to wire a team of agents that report to each other and improve themselves every single day.

What You'll Learn

Not all data is information. A JPEG is a mountain of data and almost no information. One sentence from the CEO saying "go do this" carries more than a thousand video keyframes. Store the signal, not the noise.
Memory is a compression problem. Save the process — who you talked to, how the deal got done, where it went off the rails — not every artifact you produced.
Timeliness is value. Information from 30 days ago usually beats information from 7 years ago. Decide what goes into deep storage and what stays live.
Treat every AI agent like a new employee who always needs onboarding. What's the minimum they need to do the job well? That's your context window.
Three layers of control: policy, guidelines, gates. Policy = "be nice to the client, don't cuss." Guidelines = your stop-word list. Gates = a deterministic wall, like a credit card with exactly $2,000 on it that rejects $2,001.
Run an analysis phase first. Find the 5–6 things a role does 80–90% of the day before you build anything. (An account manager is a farmer, not a hunter.)
The MD-file stack that runs an agent:

soul.md — who am I, why am I different from every other agent, what I do and what I don't do.
agents.md — the bootstrap: read your soul, read your heartbeat, and here are your SOPs as links (loaded only when the situation calls for it, so you never pollute the context window).
heartbeat — the recurring loop. The five tasks every hour, the three tasks every four hours. Keep it light or it eats you alive.

Great agent systems are an org-design problem, not a super-intelligence problem. You don't need one all-knowing brain. You need agents that each know their job and who to talk to. It works like an anthill.
Build a COO agent whose only job is writing and revising SOPs. Feed it a daily retrospective and you've built a self-learning organization

Learn how to use AI to Buy Agencies:
https://value.8figureagency.co/MNA

Systemized Sales (done-for-you sales teams): systemizedsales.com
Revphlo.com (post-call sales attribution — works with GoHighLevel & HubSpot)
Dylan on X: @sdrwhisperer

Most agencies are sitting on a goldmine and treating it like a junk drawer.
That was the exact problem on the table. A giveaway funnel pulling 100+ emails a week. 300 to 500 comments a post on X. Newsletter readers. Ad clickers who never book. Five different lists, none of them talking to each other. A great setter could go to work on that data and print money — if anyone had organized it.
So Jordan brought in Dylan Rich, founder of Systemized Sales. Dylan runs 10 sales teams at once, end to end — name, email, phone number all the way to contract signed and cash collected. In 15 years of sales and seven years working with agencies, this is the most sophisticated GoHighLevel pipeline either of them has seen. They build it live: UTM attribution per post, the "tattoo" rule for tags vs. custom fields, an action-based lead score that pings a setter the second a lead crosses 6 points, a speed-to-lead workflow with 73 million enrollments, and 57 smart lists ranked P1 through P5 so setters never waste a minute guessing who to call.
Then Dylan opens up Revphlow — the post-call attribution tool that kills the EOD form forever and tells you your cash per call by traffic source, by closer, by calendar. If you run organic and paid traffic into a sales team, this is the playbook.
What You'll Learn

The "tattoo" rule — when to use a tag (permanent) vs. a custom field (rewrites itself), and why getting this wrong wrecks your data
One pipeline beats five — why segmented pipelines cap your granularity, and how custom fields + tags + stages stack to do the real work
Lead scoring that runs itself — assign points to actions (email open = 1, application = 3, call booked = 5), and auto-trigger a setter call at a score of 6
Speed-to-lead you can actually measure — a workflow that timestamps every lead and counts only during setter working hours, so a Saturday opt-in doesn't blow up your numbers
57 smart lists, ranked P1–P5 — how setters work the right leads in the right order and never run dry between fresh leads
Ask for the phone number — "if you won't give me your phone number, you won't give me your credit card," and why fewer, higher-quality leads win
Post-call notes (PCN) — auto-dispositioned call notes pulled from the Zoom transcript that update pipeline stages for you, so closers stop skipping the EOD form
Cash per call by source — the metric that tells you which ads, threads, and posts to make more of (and which to kill)

Chapters
— Why scaling SDRs is the hardest (and most profitable) thing an agency can do
— Who is Dylan Rich: 15 years in sales, done-for-you teams at Systemized Sales
— The real problem: a giveaway funnel pulling 100+ emails a week, five siloed lists
— How to organize intent leads so setters book more calls per hour
— UTMs per post, the Zap into GoHighLevel, and tagging the first click
— Live build: one pipeline + custom fields for granularity
— Lead categories that follow a contact all the way down the funnel
— Tags vs. custom fields: the "tattoo" concept explained
— Lead scoring: assigning points to actions and triggering at 6
— The speed-to-lead workflow with 73 million enrollments
— Keeping setters organized so they stop thinking and start dialing
— 57 smart lists ranked P1–P5, built on filters, not vibes
— Why you should ask for a phone number on every opt-in
— "Can we blow your socks off?" — the Revphlow post-call attribution demo
— PCN, Zoom + Fathom + Slack: clean dispositions on every call
— Cash per call by source, by closer, by calendar
— Where to find Revphlow and Systemized Sales

Go AI-Native in the Next 90 Days
This episode is the manual setup. The next level is making it run itself.
8 Figure Agency installs AI-native systems into 7- and 8-figure marketing agencies — the kind that add $1M in annualized profit. Attribution that tags itself. Pipelines that update without a human touching them. Setters fed the highest-intent leads automatically. If you want your agency AI-native in the next 90 days, start here:
→ 8figureagency.co
Work With Dylan
Dylan and the Systemized Sales team run your sales process end to end — they listen to your calls, audit your CRM, review your last 90 days of sales and marketing numbers, find your biggest constraint, and fix it.

Systemized Sales (done-for-you sales teams): systemizedsales.com
Revphlo.com (post-call sales attribution — works with GoHighLevel & HubSpot)
Dylan on X: @sdrwhisperer

Tools & Resources Mentioned

GoHighLevel — the CRM the whole pipeline is built in
RevFlow / revphlo.com — post-call attribution and sales analytics
Typeform — opt-in forms (add a hidden UTM source field)
Beehiiv — newsletter platform, with email opens webhooked back to GHL for lead scoring
HYROS — marketing attribution into the call
Zoom + Fathom — call recording and transcripts feeding the PCN
Slack — where post-call notes drop for closer confirmation
Vidalytics — VSL view tracking
UTM Builder (utmbuilder.net) — Jordan's no-affiliation pick for building UTMs

FREE DOWNLOAD: How to Set Up the Hermes Agent → https://value.8figureagency.co/hermes

Ready to become an AI-native agency? Book a call at 8figureagency.co

Guest: Ben Fisher, founder of Skinny and Bald. Hampton member. Been coding since fifth grade. Has been CEO twice, CTO three times. Describes himself as 60% product, 40% engineer.

The pull: Ben is one of the sharpest guys in the Hampton AI channel. This is his second time on the pod. Jordan came with real questions about funnels, databases, and skills — Ben answered live, then pulled up his actual meeting-processing skill on screen share.

What we covered

The build vs. buy question. Jordan’s giveaway funnel pulls comments → emails → form fills → booked calls across LinkedIn and X, six campaigns a week. His partner said use GoHighLevel. Ben’s framing: build custom with Claude when you control the maintenance, use a tool when it solves 80%+ without workarounds. Texting is the exception — Twilio’s regulatory rabbit hole can eat days even with Claude Code.

Databases and custom funnels. Jordan wants the funnel experience to mirror what the user clicked — landing page copy, follow-up sequence, everything. Ben’s example: he still uses Kit.com for his newsletter, but layered custom API logic on top. He didn’t rebuild Kit. He enriched it.

The thing that separates real builders from vibe coders. “What distinguishes really effective builders really comes down to workflow.” Same models. Same Claude. Different results because of how people work. Ben’s non-negotiable: test-driven development. Plan first. Write the tests. Then build. Otherwise Claude tells you it shipped something that doesn’t exist.

The “your friend Ben is absolutely correct” story. Hampton buddy building a chief of staff agent in Slack and WhatsApp. Asked Claude if it was secure. Claude said yes. Ben listed four gaps. Buddy pasted Ben’s message into Claude. Claude wrote back: “Your friend Ben is absolutely correct. We don’t do this, this, this, and this.” Lesson: you cannot ask the AI to verify the AI.

Claude Skills, real talk. Skills are mostly plain English text files the AI reads. They get highly personal fast — Ben said his public repo of skills is becoming less useful to others because the nuances are his. Best move for most people: use someone else’s skill as a reference, have Claude analyze how it works, then build your own flavor.

Ben’s content-from-meetings workflow (live demo). Fireflies records every call → transcripts get stored as markdown files in a local folder → a Claude skill called process meeting notes runs on demand, pulls the last 3 days, formats each meeting in EOS Level 10 format (clear accountability, agreements, action items), and routes to-dos to the right project repo. Why local files instead of remote Fireflies calls? Speed. His second brain reads disk faster than it makes API calls across nine months of transcripts.

Writing in your voice with AI. Ben studied journalism and advertising. He uses Claude as a sparring partner first, last-mile editor second. Reference for anyone serious about this: every.to publishes their full editorial AI process, including how to build an anti-AI style guide. Ben also actively removes em-dashes from his AI output now because they’ve become the tell.

Markdown files as the convention. .md is what the AI world runs on. Pound signs for headers, asterisks for bold. Doesn’t really matter if you use .txt or .docx — but markdown gives the AI hierarchy it can parse.

Tools and references mentioned

Fireflies, Claude Code, N8N, Zapier, Kit.com, GoHighLevel, Twilio, Obsidian, every.to, Superpowers (Claude skill harness), Hampton, EOS Level 10 format, Ruben’s “How AI” Substack.

Where to find Ben

skinnyandbald.com — consulting offers
dearben.ai — Ben’s AI podcast where execs submit questions and he answers live with screen share, plus his newsletter

Ready to build an AI-native agency that runs on systems, not scrambling?

8 Figure Agency helps seven-figure agency owners install the agents, automations, and AI workflows that turn your team into a 10x operation. Done-for-you implementation starting at $2K/month.

Book your call: 8figureagency.co
value.8figureagency.co
hermes
This playbook shows you the exact stack. Install instructions, the 30-day roadmap, the five daily prompts that turn your agent into a second brain, and eight use cases pulled straight from agencies doing it right now.

8figureagency.co
AI Solutions for Marketing Agencies | 8figure agency
Optimize your marketing agency with our AI solutions. Join 1,000+ agencies and scale your revenue today!

every.to
Every
Every — The only subscription you need to stay at the edge of AI. Ideas, apps, and training from practitioners who build with AI daily.
http://every.to/

skinnyandbald.com
Ben Fisher
I help companies figure out where AI fits — and then build it.