001 — Infrastructure / Spring 2026
New essay AI as Infrastructure

Leverage AI inside your company. Outpace the ones that won't.

Embedded AI operator. Production systems in thirty days, not roadmaps in twelve months.

Currently Embedded · growth-stage B2B
Available June 2026 onward
Range Assess · Train · Deploy · Embed
Live · operations evan.knox / live-systems
3 Active engagements 01 Train 02 Deploy 03 Embed
Spirituscoaching co-pilotrunning
Bowmanbooking + voicerunning
Embed (B2B)demand-gen v23running
Operating today, in production —
Spiritus Bella Cottage Bowman Fly Fishing FundEasy Horizon Hygiene Homegrown + confidential B2B
002 — The Thesis

Software used to be a product you bought. A line item. A logo on a slide. That era is closing.

The companies I work with no longer ask which SaaS to license. They ask which operator to build. The answer is rarely a product. It is an operator who builds infrastructure inside the company — production systems that run underneath the org chart the way electricity runs underneath a building. The shift is not subtle. The market has not finished pricing it in.

For most of the last decade, the AI conversation was about products: which tool to subscribe to, which vendor to evaluate, which stack to integrate. That conversation is fading. The new question — the one that decides which companies compound and which calcify — is not which AI, but whose AI. Whose hands are on the wires inside your company.

The companies that win the next decade will not hire vendors. They will hire operators who build the wiring. I build the wiring.

003 — The framework, in motion

The Embedded Loop.

Hover a ring to see what each motion delivers. The team takes over. The loop continues.

3 Active engagements 01 Train 02 Deploy 03 Embed
Train teaches. Deploy ships. Embed runs. The team takes over. The loop continues.
004 — Three motions, three engagements
i.
Train · 60 min → multi-day
The team gets fluent.
A leadership-team session that transfers operating capacity. Plain-language foundations on what LLMs actually are. Hands-on with Claude. Walkthroughs of production systems your team can adopt Monday.
14:32Sessionmodule-15 → live200
14:18Hands-on5 prompts → operator-gradeok
ii.
Deploy · 30–60 days
The system goes to production.
A working AI system shipped inside your business. Voice, content, outreach, dashboards — whatever the operation needs. The team owns the keys when I leave.
d-01buildoperator.deploy → main200
d-30handoffteam owns the keys
iii.
Embed · 90-day minimum
Inside the org chart.
The deepest tier. I operate from inside, not across a vendor relationship. AI demand-gen architecture, ICP, GTM. Shipped where the work happens. Monthly cadence — for as long as the wiring keeps compounding.
wk-01embedorg-chart entry200
wk-13compoundweekly digest liveok
005 — Time to production

Thirty days to shipped.
Not twelve months to strategy.

The traditional consultant ramp peaks at month six. The embedded operator peaks at day thirty — when the system is already running in production and the team is already trained on the keys.

Traditional consultant
Embedded operator
0% 25 50 75 100% d-0 d-30 d-90 d-180 d-30 · in production d-180 · just ramping
— What the first 90 days look like

Embed, week by week.

Week 01

Org-chart entry.

Two-day immersion with the founders and the team. Ship the first revenue-bearing change before the week ends. No discovery phase.

Week 02 – 04

Wiring goes in.

Production AI systems shipped where the operation needs them. Outbound, voice, content, dashboards. The team watches the wiring get installed.

Week 05 – 09

The team takes over.

Operators inside the team start running the systems independently. I document what compounds. Standups become weekly.

Week 10 – 13

The wiring compounds.

The team is running production systems. I am running the next one. Knowledge stays as code. The retainer stays for as long as the work compounds — and most do.

006 — Engagement structure

Four ways to engage.

Most engagements start with the $1,500 AI Assessment — a one-week look at how the company uses AI today and the three highest-leverage moves to make next. From there, Train, Deploy, and Embed are how the work gets done. Embed is monthly, ongoing, billed for as long as the wiring keeps compounding.

Engagement
Assess
7 days
$1,500 flat
Train
60 min – multi-day
From $6,500
Deploy
30 – 60 days
Scoped per engagement
Embed
Monthly · ongoing
From $5,500/mo
Outcome
A clear-eyed audit of how AI is being used today and where the leverage is hiding
Team fluent in AI, ready to ship Monday
Production system, team owns the keys
Operator inside the org chart, work compounds month over month
Format
A one-week deep look at your stack, prompts, workflows, and team usage. Working session at the end.
Live session · in-person or remote · executive briefings, team workshops
Project · scoped, shipped, handed off · async + weekly cadence
Embedded · weekly cadence inside your team · standups, builds, deploys
Deliverable
An AI Assessment report — current state, gaps, the three highest-leverage moves, and a 90-day plan
A team that knows how to build with AI on its own
A working system in production with documentation
Continuous infrastructure built inside the company
Best fit
Founders & ops leaders who suspect AI is leverage they have not unlocked
Leadership teams, faculty, ICs · 5 – 50 people
Founders & operators · clear problem, no AI engineer yet
Growth-stage and at-scale companies · billed monthly, no exit clock
Capacity
Available now
Available
Available · June 2026 onward
1 slot, late 2026
005 — The operator stack

The infrastructure I run on.

Production tooling I have shipped in real engagements. Claude Code as the build layer. MCP for tool access. Modal and Vercel for compute. Retell for voice. Webflow for content surfaces. The same stack I deploy for partners.

buildclaude code · mcp · cursorlive
runtimeclaude api · openai · anthropic-economic-indexlive
computemodal · vercel · cloudflarelive
voiceretell ai · elevenlabs · twiliolive
dataairtable · clickup · webflow cmslive
outboundinstantly · apollo · slack · gmaillive
surfacewebflow · vercel · framerlive
testplaywright · puppeteer · githublive
"AI is no longer the product. It is the infrastructure."
005 — Evidence, in two parts

Bella Cottage.

A direct-to-consumer home goods brand. Healthy product. Broken funnel. The team did not have a strategy gap. They had a deploy gap.

$770K → $3.5M.In twelve months. 4.5× revenue.
Before$770K
After · 12 months$3.5M

I rebuilt the operating stack end-to-end. Paid acquisition architecture. Lifecycle email. Post-purchase. Inventory-aware merchandising. A creative engine the team could run without me. No two-hundred-slide deck. No discovery phase. The first revenue-bearing change shipped in week one.

"The deliverable was never a recommendation. It was a company that no longer needed the recommendation."

Seven hundred seventy thousand dollars to three and a half million in twelve months. Four-and-a-half times. The team kept running it after I stepped out — which is the only result that compounds.

SituationDTC home goods · healthy product · broken funnel
MoveRebuilt operating stack end-to-end · acquisition + lifecycle + ops + creative engine
Result$770K → $3.5M in 12 months · 4.5× revenue · team owns the keys
PrincipleStrategy is free. Deploy is the moat.

Bowman Fly Fishing.

An outfitter that ran on word-of-mouth and a phone. Beautiful trips, no infrastructure. The constraint was not demand. It was capture.

4× revenue.In a single season. Word-of-mouth → digital infrastructure.
Before · word-of-mouth
After · 1 season

I built the operating stack from zero. Booking and capture, voice agent for missed calls, retention sequences, ops dashboards, content engine. The brand stayed the same. The wiring underneath did not exist before, then it did.

"Growth-stage companies do not have a strategy gap. They have a deploy gap."

Four times revenue in one season. The trips kept selling themselves. The systems kept running while the founders were on the river. That is what compounding looks like when the operator stays embedded.

SituationOutfitter · word-of-mouth · phone bookings · no infrastructure
MoveBuilt stack from zero · booking + voice agent + retention + ops dashboards + content engine
Result4× revenue in a single season · word-of-mouth → digital infrastructure
PrincipleThe constraint was not demand. It was capture.

Two cases. Different verticals. Same shape: situation, move, number, principle. The framework is portable. The wiring is not.

006 — Operating today, in production

Seven companies. Different verticals. Same wiring.

A look at what the infrastructure actually shipped — booking flows, coaching co-pilots, demand-gen dashboards, vendor onboarding.

Illustrative previews · representative of systems shipped · live numbers vary

spiritus.coach / co-pilot
Coach AI · 2 min ago
Three sessions reviewed. Two clients showing the avoidance pattern from last cohort — recommended next-session prompts attached.
Reframe pattern Embodied practice +3 more
Spiritus
Coaching firm · 21 years
AI coaching co-pilotShipped Mon
bella-cottage.com / ops
Revenue · trailing 12 months
$3.5M+354%
JAN '25DEC '25
Bella Cottage
DTC · home goods
$770K → $3.5M4.5× / 12 mo
bowmanflyfishing.com / book
Available trips · May
M
T
W
T
F
S
S
Half-day · Chattahoochee$485
Bowman Fly Fishing
Outfitter · GA
Booking + voice + retention4× rev / season
fundeasy.com / pipeline
Demand-gen funnel · Q1
Visits48,210
MQL3,840
SQL912
Closed+58%
FundEasy
Fintech · SaaS
Demand-gen rebuild+58% growth
horizonhygiene.com / dispatch
Today's route · 7 stops
1Buckhead · pressure wash8:30a
2Sandy Springs · driveway10:15a
3Roswell · holiday lights est.12:00p
Horizon Hygiene
Service · Atlanta
Booking + dispatch + voiceRev share
findhomegrown.com / vendor
Vendor onboarding · live
Hohosh Bakehouse
Decatur, GA · sourdough
Trial
7 days
Then
$10/mo
Homegrown
Marketplace SaaS · founder
0 → 50+ vendorsBootstrapped
Confidential · embedded
Currently embedded · vii.
Growth-stage B2B SaaS · $125M Inc 5000.

AI demand-gen architecture, ICP rebuild, GTM motion — shipped from inside the org chart, not across a vendor relationship. The work compounds because I'm there when it breaks.

Embed · ongoing monthlyActive · 2026
— Operating today, in production

Live infrastructure. Real numbers. Running tonight.

Trailing 12 months 27,728 Emails sent autonomously by deployed systems — outbound that runs while the team sleeps. Most of these never touched my hands after deploy week.
528+ Production systems shipped this year
5 Exits across SaaS, eCommerce, services
99.94% Uptime, last thirty days, all live systems
007 — What this is not

A few clarifications.

Not a strategy consultancy.

Strategy is the cheapest deliverable in the building. You have one already.

Not an AI agency.

Agencies sell hours. I sell infrastructure that outlives the engagement.

Not a fractional CMO.

Fractional roles optimize the org chart you have. I rebuild the wiring underneath it so the next CMO inherits leverage, not maintenance.

008 — Operating principles

A working memory.

Seven sentences. Pinned above the desk. They survive every framework, every cycle, every cold-start of a new engagement.

  1. i.

    Knowledge is code. Code persists. People rotate.

  2. ii.

    Ship in week one or do not ship.

  3. iii.

    The deck is not the deliverable.

  4. iv.

    Build infrastructure your team can run.

  5. v.

    Strategy is free. Deploy is the moat.

  6. vi.

    Operators compound. Vendors expire.

  7. vii.

    Build the wiring. Leave the keys.

010 — Field notes

Where the operator is, this week.

Currently shipping A leadership-team Claude session for a 21-year coaching firm — module 15, stewardship and legacy.
Currently building The cottage moat for a marketplace SaaS I co-founded. Vendor onboarding, drops engine, signup velocity.
Currently embedded A growth-stage B2B partner. AI demand-gen architecture from inside the org chart. Confidential.
Currently writing AI as Infrastructure — long form. The piece you are reading is the short version.
Currently reading Dario Amodei, Machines of Loving Grace. Anthropic's quarterly Economic Index.
012 — Common questions

Before you ask.

Why "embedded" and not "fractional"?+
A fractional CMO optimizes the org chart you have. An embedded operator rebuilds the wiring underneath it so the next CMO inherits leverage, not maintenance. Different unit of work. Different durability.
Do you sign NDAs and work confidentially?+
Yes. I currently operate inside a growth-stage B2B partner I cannot publicly name. Most embedded engagements end up here — under NDA — until the team is ready to talk about the work themselves.
What does "production system" mean, specifically?+
A working AI system that runs daily inside your business. Voice agents that answer calls. Outbound systems that send 1,500 personalized emails per day. Content engines that publish. Dashboards your team uses. Documented, handed off, owned by your team. Not a notebook, not a prompt library, not a deck.
Why a thirty-day Deploy and not a six-month roadmap?+
The traditional consultant ramp peaks at month six. The embedded operator peaks at day thirty. Strategy is the cheapest part — every founder has one already. Deploy is the moat.
What stack do you ship on?+
Claude Code as the build layer. MCP for tool access. Modal and Vercel for compute. Retell for voice. ElevenLabs for synthesis. Twilio for telephony. Webflow for content surfaces. Playwright for browser automation. Whichever combination the system needs. The same stack I run for my own businesses.
How do you decide whether to take an engagement?+
Three filters. The team has product-market fit. The work compounds inside the org chart, not just for me. And there is something the team will own permanently when I leave. If any of those is missing, I pass.
Evan Knox
011 — About the operator

A short note.

Five exits across SaaS, eCommerce, and services. I have been the buyer, not just sold to one. Most of what I do now is shipped from inside the org chart — embedded with founders and growth-stage teams who have product-market fit and need someone to build the wiring underneath.

I work with a small number of partners at any given time. The work compounds because the team takes over. I leave the keys.

Atlanta, Georgia. Father of two. Christ-follower. Currently building Homegrown — an online farmers' market for cottage food vendors — alongside the consulting practice.

5Exits
7Active partners
12+Years operating
— What happens next

Three steps. No intake form.

i.

Send a paragraph.

Email me three sentences about your business and the operating problem you are looking at. No NDA needed yet. No discovery call to schedule a discovery call.

ii.

I respond in twenty-four hours.

I will tell you whether the problem is a Train, Deploy, or Embed shape — or whether you should hire someone else. I do not take engagements I cannot ship.

iii.

We talk for thirty minutes.

If it looks like a fit, we get on a call. By the end you have a written scope, a price, and a start date. Or you walk with a clearer read on the problem and no obligation.

— The quarterly

Get the essay when it ships.

One long-form piece per quarter — when there is something worth saying. AI as Infrastructure drops next. Operator-grade thinking, not newsletter cadence noise.

  • i.4,200-word essays · drafts in public
  • ii.One per quarter · skip the rest
  • iii.No third-party tracking · no upsells
Issue 01 · Spring 2026 AI as
Infrastructure.
Why software stopped being the product.
No spam · Unsubscribe any time · 4,200 words
"Knowledge is code. Code persists. People rotate."

If you are at-scale and AI-absent, that is the conversation.