Skip to content

00 / KAREL HERNANDEZ RODRIGUEZ · AGENTIC SYSTEMS · BUILT ON 20 YEARS OF OPERATIONS

Your next department isn't a hire.
It's a system.

I design, build, and run production AI agent systems for mid-market companies — fixed scope, measured outcomes, every action logged with a receipt. Twenty years of keeping business infrastructure alive taught me to build things that keep running. Not a consultancy that ships decks.

FIG. 1 — ANATOMY OF A PRODUCTION AGENT SYSTEM

01 INTAKE02 PROCESS03 GOVERN04 RECORDEMAILCALLSDOCSTRIAGE AGENTDRAFTSCHEDULERBILLINGHUMAN APPROVALYOUR CRM✓ RECEIPT LOGGED — EVERY ACTION, EVERY TIME

SYSTEM LOG — STDOUT

SIMULATED FEED

SHEET 2 OF 2 · REV 2026.10 · DRAWN: KAREL · CHECKED: HUMAN

2 PRODUCTION AGENT PLATFORMS2,500+ COMMITS SHIPPED284 DATABASE MIGRATIONS1,400+ TEST FILES10 NAMED PRODUCTION AGENTS20+ YEARS RUNNING BUSINESS IT286-PAGE BOOK ON RAG AGENTS0 DECKS SHIPPED

01 / THE PROBLEM

Mid-market AI is stuck between decks and demos.

Big firms sell strategy and hand the build to juniors. Automation freelancers ship flows that die in production. Platform vendors sell tools without governance. The gap between them is where your margin is going — and where I work.

WHAT YOU'VE BEEN OFFERED

A-01A 90-page strategy deck, then an invoice
A-02Juniors do the build — if a build ever starts
A-03Hourly billing with no ceiling
A-04A pilot that never leaves the sandbox
A-05Gone the week after go-live

HOW I WORK

B-01Working software, measured against your baseline
B-02I build it myself — with an agent fleet behind me
B-03A price with a ceiling, agreed before I start
B-04Production in 30 days, receipts included
B-05I stay and run it — that's the business I come from

The consultants who can strategize can't build. The builders who can build have never run a business. I've done both for twenty years.

02 / WHAT I BUILD

Four ways in. Each one de-risks the next.

Fixed scope, fixed price, fixed timebox — never hourly. Start with the audit; every step after it is a decision you make with evidence in hand.

S-01 · 2–3 WEEKS

The Opportunity Audit

Where agents pay for themselves — with the math.

$9,500fixed · credits toward the pilot

  • Workflow mapping across your operations — industrial-engineering style, not vibes
  • Scored opportunity map: impact × feasibility × data readiness
  • Build / buy / wait verdict on every candidate
  • 90-day roadmap you can execute with or without me
Book the audit

S-02 · 30 DAYS TO PROD · 45 ALL-IN

THE FLAGSHIP

The Pilot Build

One agent, in production, in 30 days.

from $24,000fixed scope

  • The audit's top opportunity becomes a working agent
  • Wired to your real systems — email, phones, documents, CRM, practice management
  • Human approval gates and receipts on every action
  • Measured against your baseline: hours, response time, cost
Start with the audit

S-03 · ONE–TWO 90-DAY CYCLES

The Systems Build

The agent system that runs a department.

from $60,000milestone-billed

  • Multi-agent orchestration: intake, triage, drafting, follow-up, reporting
  • Governance built in: permission ladders, audit trails, live monitoring
  • Change management as a workstream — training and adoption, not hope
  • A trained internal owner before I step back
Discuss a build

S-04 · ONGOING

The Operating Partner

Your fractional AI officer — plus the crew that keeps agents sharp.

from $6,500/moquarterly terms · 90-day decision points

  • Roadmap ownership, vendor evaluation, governance, board reporting
  • Managed agents: monitoring, accuracy testing, model upgrades, incident response
  • The MSP discipline — SLAs and uptime culture — applied to your AI
  • Scale, continue, or stop at every quarter. Your call.
Talk retainers

THE AUDIT MATH — YOUR INPUTS · YOUR NUMBERS

$/ HOUR
HRS / WK

That's today's cost of the work — not a savings promise. The audit's job is to find where, and whether, agents can carry it. If it can't find 3× its fee in annualized savings, you don't pay.

$0

WHAT THAT WORK COSTS YOU PER YEAR

THE OPPORTUNITY AUDIT$9,500 FIXED

GUARANTEE THRESHOLD — 3× THE FEE$28,500

3× ITS FEE OR IT'S FREEBook the audit

03 / EVIDENCE

Don't take my word. Take the receipts.

Everything on this receipt is countable in a repository or a public channel. No client logos I can't back, no invented outcomes — engineering facts, printed.

Before I sold any of this, I built two full agentic platforms of my own. They are the lab: every pattern I deploy for you — permission ladders, approval gates, agent memory, receipts — is already running in production there. You buy patterns that exist, not experiments on your dime.

KAREL.BUILD — ENGINEERING RECEIPT

VERIFIABLE FACTS ONLY · REV 2026.10

PRODUCTION AGENT PLATFORMS×2
COMMITS SHIPPED2,500+
DATABASE MIGRATIONS284
TEST FILES1,400+
NAMED PRODUCTION AGENTS10
SYSTEM INTEGRATIONS15+
YEARS RUNNING BUSINESS IT20+
PAGES IN THE BOOK286
DECKS SHIPPED0
✓ VERIFIED IN PRODUCTION

E-01 · AGENTIC WORK-CREW PLATFORM

Cuadrila

Ten named agents — intake, reception, drafting, sales, review — operating under a default-deny permission ladder. Every action needs an explicit grant; every output is saved as an immutable receipt. Built for regulated work: legal, insurance, claims.

  • 10-verb permission ladder
  • Immutable audit receipts
  • Scope isolation, fail-closed
  • Email · phones · DocuSign · Drive · Slack

E-02 · COMMUNITY PLATFORM + AGENT RUNTIME

clases.community

A revenue-bearing learning platform with agents woven through it: multi-provider model routing, two-tier persistent memory with cited sources, plan-approval gates on every risky action, and Stripe Connect money flows underneath.

  • 5 LLM providers, gateway-routed
  • Persistent memory w/ citations
  • Plan triage + approval workflow
  • Payments · video · 21 scheduled jobs
VERIFY — CLASES.COMMUNITY ↗

E-03 · THE OPERATING BASE — SINCE 2004

Zoltis Technologies

The managed-services firm that has kept Los Angeles law firms and mid-market companies running for two decades. Uptime culture, security posture, and the domain knowledge of how work actually moves through your systems.

  • 20+ years of operations
  • Law-firm specialization
  • Security & continuity discipline
  • The clients agents must serve
VERIFY — ZOLTIS.COM ↗

E-04 · THE TEACHING CHANNEL

Cuban Hacker

Where I teach AI and automation, in English y en español. Change management is teaching — and teaching is a practiced muscle.

  • YouTube channel
  • Newsletter
  • AI + automation lessons
  • Bilingual delivery
VERIFY — CUBANHACKER.COM ↗

04 / ENGAGEMENT SEQUENCE

Ninety days. Then you decide.

Every engagement runs as a contained 90-day cycle with a hard decision point at the end: scale, continue, or stop. Here's the first cycle, day by day.

DAYS 01–14

01Map

Workflow mapping, systems inventory, data readiness. The industrial-engineering pass: where hours go, what they cost, what an agent could carry.

DAYS 15–21

02Decide

Scored opportunity map on the table. Build, buy, or wait — each verdict with the math attached. You choose the pilot with evidence, not enthusiasm.

DAYS 22–52

03Build

One agent into production: wired to your systems, gated by human approval, logging receipts. Your team trained while it's built, not after.

DAYS 53–83

04Measure

Agent runs against the baseline I captured on day one. Hours saved, response times, cost per unit of work — in a report you can hand to the board.

DAY 90

05Decide again

Scale it, continue it, or stop. Every quarter ends with that decision in your hands. That's the whole deal — no lock-in, just receipts.

05 / THE OPERATOR

One builder. Twenty years of showing up.

I'm Karel Hernandez Rodriguez — industrial engineer, founder of Zoltis Technologies, author, and builder of two agentic platforms. When you hire me, I'm who shows up. The bench behind me is silicon.

“I spent twenty years keeping other people's systems alive. I'm not going to hand you something I wouldn't operate myself.”

— KAREL HERNANDEZ RODRIGUEZ

2004

Founds Zoltis Technologies. Managed IT for Los Angeles law firms and mid-market companies — uptime, security, and the discipline of being the one they call at 2 a.m.

—

Industrial Engineering + Business training. The habit of mapping a process before touching it, and pricing the waste in it.

2023

Goes deep on AI engineering — and starts teaching it as Cuban Hacker: AI and automation, in English y en español.

2024

Publishes LangChain for RAG Beginners — a 286-page, hands-on guide to building AI agents grounded in your own data with retrieval-augmented generation. The pattern behind the production builds that followed.

2025

Ships clases.community — a revenue-bearing community platform with a full agent runtime: multi-provider routing, persistent memory, plan approvals.

2025

Builds a production agentic-AI platform for a California law firm — case-document analysis with vector retrieval, a firm-wide Teams assistant, and settlement forecasting. Still running in production today.

2026

Ships Cuadrila — governed AI work crews with permission ladders and audit receipts, templated for legal and insurance. Opens karel.build.

06 / OBJECTIONS, HANDLED

Ask the hard ones.

These are the questions operators actually ask me — answered the way I'd answer across a table.

That fear is why my systems are built the way they are. Agents operate under default-deny permissions: they can't send, publish, or modify anything without an explicit grant, and risky actions park at a human approval gate before anything leaves the building. Every action is logged with an immutable receipt, so 'what did the AI do?' is answered in seconds, not days.

07 / START THE SEQUENCE

Book the audit. Get the math.

Two to three weeks. $9,500, credited toward the pilot. If I can't find three times the fee in annualized savings, it's free — and either way you keep the roadmap.

RESPONSE TIME≤ 1 BUSINESS DAY

AUDIT SLOTS / MONTH2

DIRECT LINEHELLO@KAREL.BUILD

DIRECT LINE

hello@karel.build

Tell me where the work piles up — intake emails, client follow-ups, document processing. Your name, your company, and one paragraph is plenty.

ANSWERED BY ME · WITHIN ONE BUSINESS DAY · NO NEWSLETTER AMBUSH