Available for new projects · Remote worldwide

I build AI automation that earns its keep.

I'm an AI systems & automation architect for ops-heavy businesses. I fix the broken process and cut your SaaS bloat first, then add AI exactly where it earns its keep, all self-hosted on infrastructure you own, so your data stays private and your stack stays reliable. No tool sprawl, no AI slop, nothing that hallucinates into business-critical work.

50+ processes automated $800K+ in cost saved Self-hosted & private
Book a free 15-min call → See the work

● Free 15-minute call · no obligation · reply within 1 business day

Real result: +30% lead conversion on a capped AI budget
client-intake.workflow · 6 nodes
Triggerwebhook · event
Validatefields · auth
Processtransform · enrich
Deliverwrite · notify
Audit LogPII-masked
Fallbackon error · retry
drag the nodes · click ▶ to run
100%

How it feels

Manual work, all day. One clean system you own.

scroll, watch the busywork resolve into one deterministic pipeline

Proof in numbers

Cost savings delivered
$0

Hard-dollar savings across 50+ automations running in production, manual hours eliminated and redundant SaaS subscriptions cut. A conservative running total, and still climbing.

Automations in production
0
Manual data entry
−70%

on a 3-system lead intake pipeline

Lead conversion
+30%

Deterministic routing + one cheap, capped AI call

Hallucinations in critical paths
Zero

Deterministic logic can't hallucinate. Validation, retries & audit logging from day one.

How I build
  • Fix the broken process before adding any AI
  • You own the infrastructure, your data stays private
  • Deterministic where it's critical, AI only where it pays
  • One system that holds, not five tools that break
The math

What's all that manual work costing you?

Every hour your team spends copy-pasting is money out the door, and the SaaS subscriptions quietly stacking up alongside it. Drag the sliders and watch the manual cost alone add up.

10h
$50
Bleeding out to manual work · yearly
$0
Live · climbing every second, around the clock
Capacity you'd reclaim0x
Lost over 5 years$0
Let's win that back →
● Free 15-min call · no obligation

Not ready to talk yet? Grab a free 1-page roadmap →

Who I help

If you're one of these, we should talk.

You've probably already got AI tools, and the sprawl, the climbing SaaS bills, and the outputs nobody fully trusts that come with them. I build automation that actually holds, on infrastructure you own, and add AI only where it earns its keep.

The Founder

"We bolted AI onto everything, and now it's a tangle of tools and subscriptions I can't keep straight."

Cut the tool sprawl →

The Ops Lead

"Our automation and SaaS bills keep climbing, and things still break every time an integration changes."

Self-hosted & reliable →

The Margin-Watcher

"An AI 'helper' hallucinated a number straight into a client record. Now I don't trust a model near anything critical."

Keep AI safely in its lane →
Who you'd be working with

Hi, I'm Lorenz. I build reliable AI automation that holds up in production.

Lorenz Espinosa, Automation Architect
Available for new projects

I'm an AI systems & automation architect who builds reliable, self-hosted automation for ops-heavy teams. I've automated 50+ real operational workflows across intake, case and record management, billing, and client communication. My approach is pragmatic: fix the broken process and cut the SaaS bloat first, then put AI exactly where it earns its keep, no tool sprawl, no slop. I design for the messy edge cases real businesses actually hit, not the happy path in a demo.

Every build ships with validation, retries, fallback paths, and audit logging from day one, plus documentation so your team owns it. No brittle scripts, no black boxes, nothing that fails quietly, and nothing quietly racking up subscriptions behind your back.

Beyond automation builds, I've run enterprise Microsoft Cloud and multi-tenant IT + marketing automation for Le Cordon Bleu International, so I'm as comfortable in a large multi-tenant environment as I am wiring up a lean ops stack.

n8n Level 2 Certified AWS Cloud Practitioner Azure AZ-900 Make & Zapier Certified GitHub Foundations
Lorenz Leslie EspinosaAutomation Architect · fully remote, worldwide
Self-hosted & private

Runs on your own infrastructure, self-hosted n8n on Docker. No third-party middleman sitting on your data, no per-task SaaS fees, and nothing going down on you when a vendor does.

Logged & PII-masked

Every action is written to an audit log and sensitive fields are masked. You always see what ran and why.

Deterministic, with real error handling

Rules and code do the predictable work, token-free and hallucination-proof. Plus retries with backoff, fallback paths, and dead-letter queues from day one. Open-sourced as reusable patterns.

Selected work

Stuff I've actually shipped.

Not demos or weekend prototypes. Real systems running in production, most of them replacing brittle, sprawling tool stacks with reliable, self-hosted automation, with the numbers to back it up.

Live · v5.0

OpenClaw + Slack: AI agents done right, deterministic at the core

A controller that hands work to specialist AI agents only behind a human approval step, wired to 21 MCP tools inside Slack, while the deterministic work runs as plain code, so no tokens are burned on anything a rule can decide.

21MCP tools · zero tokens on deterministic tasks
OpenClaw · Slack · n8n MCP · Claude
▸ See the before → after → result
Before

Operational tasks were scattered across tools and needed a human to babysit every step. Letting AI touch real systems felt risky, one wrong move with no guardrails.

After

A controller hands work to specialist AI agents only behind a human approval gate, wired to 21 MCP tools inside Slack. Everything a rule can decide runs as plain, deterministic code, no tokens burned, nothing left to a model to guess.

Result

21 MCP tools under one approval-gated controller, with zero tokens spent on deterministic tasks, AI reserved for the few steps that truly need it.

View the stack on GitHub →
Live

n8n + Airtable: deterministic lead intake, no retyping

Catches a new lead, validates and routes it with deterministic rules, writes it to the CRM, and creates the record, with a single cheap AI call reserved only for the messy edge cases. Killed the manual data entry that used to span three separate systems.

−70%manual data entry
n8n · OpenPhone · CRM · Airtable
▸ See the before → after → result
Before

New leads landed by phone and web form, then someone retyped each one into three separate systems by hand. Slow, and records slipped through the cracks.

After

A webhook catches every lead, deterministic validation and routing do the work, and one constrained AI call handles only the genuine edge cases before writing to the CRM and Airtable, with a fallback path if anything fails.

Result

70% less manual data entry, no leads lost between systems, and token spend held to a single capped call per lead.

See the intake templates →
Live

n8n + hybrid AI: lead scoring that lifts conversion +30%, on a budget

Deterministic rules handle the routing and the boilerplate; one constrained model call scores and drafts a reply only where it adds real signal, then hands off to sales. Conversion up, token spend kept lean.

+30%conversion · −70% manual effort
n8n · hybrid routing · CRM
▸ See the before → after → result
Before

Inbound leads were scored and replied to by hand. Slow, inconsistent follow-up meant warm leads went cold and deals slipped through.

After

Deterministic rules route every lead and assemble the boilerplate; a single capped model call scores and drafts a reply only where it moves the needle, with a human in the loop where it counts. Most of the pipeline never touches a token.

Result

+30% conversion and −70% manual effort on lead handling, without an open-ended AI bill behind it.

Live

n8n + Clio: deterministic billing sync, zero missed invoices

Pure deterministic sync between case management and billing, automatic conflict resolution, a single source of truth, and not one token spent. Nothing in the money path is left to a model to guess.

0missed invoices · zero AI in the path
n8n · Clio · Airtable
▸ See the before → after → result
Before

Case management and billing drifted out of sync. Reconciliation was manual, conflicts were caught late, and invoices slipped through the gaps.

After

Bidirectional, fully deterministic sync between case management and billing with automatic conflict resolution and a single source of truth, both systems always agree. No AI in the path; every decision is a rule you can read.

Result

Zero missed invoices, no manual reconciliation, and zero tokens on a business-critical money flow.

See the sync toolkit →

Want results like these on your stack?

Book a free 15-min call →
Open source

The patterns behind the promise.

I open-source the building blocks I ship with, the validate → retry → fallback → audit discipline that keeps the core reliable and AI where it earns its keep, as real, public code you can read before we ever talk. Skim the value line; open the repo only if you want the proof.

openclaw-n8n-stack★ 2

One-command Docker stack, OpenClaw AI agent + n8n + PostgreSQL, with MCP pre-wired for multi-agent ops.

one-command deployShell
n8n-legal-ops-templates★ 2

Production-grade n8n templates for law-firm ops, client intake, missed-call recovery, billing sync, case routing. Zero real client data.

production templatesn8n · templates
n8n-lint★ 1

CLI linter for n8n workflow JSON, catches credential leaks, deprecated nodes, and orphaned connections. 16 rules, zero config, CI-ready.

credential-leak lintingJavaScript
airtable-automation-toolkit★ 1

Python toolkit for Airtable, a staging layer, dead-letter queue, PII-masked audit logging, and bidirectional sync.

PII-masked auditPython
n8n-error-handling-pattern★ 1

The reliability backbone, retry with exponential backoff, a dead-letter queue, fallback paths, and audit logging.

retry · fallback · DLQn8n · patterns
n8n-ai-agent-delegator★ 1

Multi-agent delegation for n8n, an orchestrator routes plain-language commands to specialist agents with confidence scoring and human-in-the-loop gates.

human-in-the-loopn8n · patterns

How every build works

The discipline that keeps your systems from failing quietly, and the repo where each one lives in the open.

  • Input validation & authevery payload is checked before it touches a systemlegal-ops-templates
  • Retry with backoff + DLQtransient failures retry, then park safely instead of vanishingerror-handling-pattern
  • Fallback pathswhen a step fails, there's always a defined next moveerror-handling-pattern
  • PII-masked audit loggingevery action is logged; sensitive fields are maskedairtable-toolkit
  • Credential-leak linting in CIsecrets and bad nodes are caught before they shipn8n-lint
  • Human-in-the-loop gatesAI assists only where it earns its cost; a person approves anything that mattersai-agent-delegator
What I build

Four ways to get your busywork off your plate, done right.

Everything I build is self-hosted, reliable, and shipped with real error handling, logging, and docs, with AI placed only where it earns its keep. Not a brittle stack of AI wrappers that falls over the first time a tool or an API changes.

01

🔍 Automation & Cost Audit

Know what to automate first, and where you're overpaying.

I go through how your team actually works, figure out what's worth automating first, and give you a ranked plan with rough effort and payoff for each one. I also flag the overlapping tools and SaaS subscriptions you're overpaying for, and where self-hosting would cut the bill. The low-risk way to start.

From $750 · fixed fee
Best first step · ~1 week
02

⚙️ Build & Ship

Get one painful manual process off your plate for good.

One project, clear scope, built and tested start to finish on self-hosted infrastructure. You get something your team can lean on day one, private, reliable, and yours to own, plus the docs to go with it.

From $2,500 · fixed scope
Most popular · 2 to 4 weeks
03

🤖 AI Agents & Orchestration

Put AI to work on real operations, without the runaway bills.

AI agents that do real operational work, wired to the right tools, gated by human approval, and kept efficient so they don't sprawl or run up surprise bills. AI placed where it earns its keep, not bolted onto everything.

From $4,500 · per project
For teams putting AI into production
04

🔁 Fractional Automation Engineer

An in-house automation engineer's output, without the hire.

I take your automation backlog off your hands: keep your self-hosted stack healthy and your subscriptions lean, fix what breaks, and ship something new every month.

$1,500/mo · retainer
Ongoing · monthly retainer
How I work

Low-risk to start. Built to last.

A free 15-minute call tells us both whether it's a fit before anyone commits to anything.

Discovery

Free 15-min call

You tell me the goal, the tools, and the timeline. I tell you if it's a fit.

Scope

Proposal

Clear deliverables, a timeline, and exactly what you'll get. No surprises.

Build

Build & test

Built deterministic and self-hosted, with validation, retries, and audit logging from day one.

Launch

Ship + docs

Deployed, documented, and handed off so your team owns it.

Support

Stay ahead

Optional retainer to maintain, monitor, and keep shipping.

"I'm not here to bolt AI onto everything. I fix the process, cut the bloat, and put AI exactly where it earns its keep, so what you ship is reliable, private, and actually worth paying for."

how every build works · fix the process · cut the bloat · AI where it earns its keep · retry + fallback + audit

The stack

Tools I build with.

Self-hosted Orchestration

n8nDocker ComposeTraefikTemporalMake

AI, where it earns its keep

Local modelsClaude (capped)Semantic cacheHybrid routingMCP

Self-hosted Data

SupabasePostgresQdrantAirtablePython
Questions

Before you reach out.

The things people usually want to know first.

How do projects usually start?+
A free 15-minute call. You tell me the goal, the tools involved, and your timeline, and I tell you straight whether it's a fit and the fastest way to ship it. No pressure, no obligation. A lot of people start with a small Automation Audit to test the water before committing to a build.
What does it cost?+
It depends on scope, so I quote per project after the call. As a rough guide: a focused Automation Audit starts around $750, a fixed-scope Build & Ship from $2,500, and an ongoing retainer from $1,500/month. No surprise invoices, and you'll get a clear proposal with deliverables and a timeline before anything starts.
Won't this just be more AI tools and bills?+
No, the opposite. Most teams already have AI sprawl: overlapping tools, climbing SaaS subscriptions, and outputs nobody fully trusts. I consolidate onto self-hosted n8n, which kills per-task SaaS fees and keeps your data private, and add AI only where it earns its keep: gated, efficient, and reliable. You end up with less tooling, lower fixed costs, and AI you can actually trust in business-critical work. (For high-volume teams whose LLM spend is already a real line item, I do token-level cost work too, hybrid routing, semantic caching, self-hosted models.)
Do you work with the tools we already use?+
Almost certainly. I build mostly in n8n, with Make, Zapier, and Power Automate when they fit better, and I connect to whatever you already run: CRMs, billing, phone systems, databases, and AI models over API and MCP. If it has an API or a webhook, it can usually be automated.
What happens when something breaks?+
Things break. That's exactly why every system I build has retries, fallback paths, and logging from day one. When something fails you get alerted, it's logged, and there's a way to recover. And because the core is deterministic and self-hosted, there's no third-party API to rate-limit you or vanish, and no model to hallucinate into a critical field. A retainer keeps me around to catch and fix issues before you notice them.
Are you available, and how do you work?+
Yep, currently taking on new projects. I work fully remote with teams worldwide, across time zones, and usually reply to a new inquiry within one business day.
Will my team be able to maintain it?+
Yes. Everything gets documented and handed off so your team owns it, not locked to me. And if you'd rather never touch it, a retainer means I keep it healthy and keep shipping new pieces on top.
Let's talk

Drowning in tools and subscriptions, or a process that still eats hours? Let's fix it.

Tell me your goal, the tools involved, and your timeline. I'll reply with whether it's a fit and the fastest way to ship it.

Book a free 15-min call → Connect on LinkedIn

Fully remote · working with teams worldwide, across time zones · typical reply within one business day.