Practical AI help for small businesses  ·  English

Make AI useful
in real work,
without the mess.

PuenteWorks helps small teams pick one workflow, build a working version, and put clear approvals and docs in place so people can trust what ships. Honest scope. No fake results.

How I work
6 weeks
Six-week pilot
One workflow, clear scope, a baseline, and a written plan for what to do next
Rules first
Clear permissions
Who can do what, where a human must approve, and how to stay safe day to day
Train+doc
Training + docs
Walkthroughs and simple playbooks so your team can run the system without you as a bottleneck
S
Simon Gonzalez De Cruz Founder  ·  Long Beach, CA
Past employers
Capital Group SoCal Edison PIMCO American Red Cross

The AI Ops brief (email)

A short occasional newsletter—starting soon—with plain-language notes on what to try, what to skip, and how to avoid expensive mistakes.

  • Simple planning templates
  • Checklists for approvals and safety
  • Honest lessons from what I ship
What we do

What I offer
for small businesses.

Three ways to work together: a six-week pilot, help setting up approvals and safety, and training for your team.

01

6-Week Pilot

Pick one workflow that eats up time. You get a clear scope, a working prototype, and a short written plan for next steps—including what not to automate yet.

Discuss a project  →
02

Rules and approvals

Light policy notes and approval checkpoints built into the workflow so AI can be used safely in real operations—not as a slide deck.

Discuss a project  →
03

Team training

Walkthroughs and simple playbooks so your team can run and improve the system after implementation.

Discuss a project  →
Our approach

Plan · Build ·
Test. Run.

Same discipline every time: clear scope, realistic builds, and honest review—not a demo that falls apart later.

1

Name the problem clearly

We break the work into small steps with clear inputs and outputs. If the problem is fuzzy, the AI will be fuzzy too.

2

Build with approvals where it matters

Anything risky or hard to undo gets a human checkpoint. We plan for mistakes, because they will happen.

3

Test before you rely on it

Example: mcp-video—public on GitHub—has 276 automated tests and a standard release process. Your build gets the same seriousness.

4

Operate and improve

We document who owns what, how to roll back, and how to review results so the system keeps working after launch.

I do not publish made-up testimonials. Proof on this site is public work you can inspect: repositories, shipped tools, and clear scope.
S
Simon Gonzalez De Cruz Founder, PuenteWorks
Transparent positioning · No fake client stories
Engagements

Pick your
starting point.

Three ways to work together. Details depend on your tools and constraints—we scope honestly before writing code.

Start small

6-Week Pilot

Pick one workflow that costs real time. You get a clear scope, a working prototype, and a short memo on what to do next.

  • Map the workflow and constraints
  • Build a working prototype
  • Short written plan for next steps
  • Handoff notes and a walkthrough for your team
Start a pilot
Bigger build

Custom Agent System

A multi-step system tailored to your tools and data, with clear checkpoints and a path to recover when things go wrong.

  • Architecture and written specification
  • Build with automated tests where it matters
  • Integrate with your existing tools and data
  • Monitoring and iteration support (as agreed)
Discuss your project
Figure it out

AI Strategy & Enablement

Not sure where to start? We map workflows, compare realistic options, and build a simple roadmap. English first; Spanish translation comes after the English site is final.

  • Workflow review and opportunity mapping
  • Tool options with honest tradeoffs
  • Team walkthroughs and documentation
  • Safety and approvals checklist
Book a strategy call
Public proof

Built and shipped.
Not slides.

19 tools
Video tools for assistants
Public project that exposes video-editing actions an AI assistant can call in a controlled way.
GitHub · mcp-video
276 tests
Automated tests
Checks run before release so changes are less likely to break behavior quietly.
GitHub · mcp-video test suite
16 tools
Dialect tools shipped
Public Spanish dialect tools published to npm with regional variants.
npm + GitHub · DialectOS
About the founder

Simon
Gonzalez De Cruz

I build AI systems and help teams adopt them responsibly. My background is operations-heavy: about 11 years across Capital Group, Southern California Edison, and PIMCO—then I started building the systems myself.

Public work includes mcp-video (video tools for AI assistants via MCP—think “a standard plug-in layer so assistants can use tools safely”), DialectOS on npm, and other experiments on GitHub.

I'm based in Long Beach, California. This site is English-first; a Spanish version will follow once the English copy is locked.

PuenteWorks means "bridge works" — building bridges between where you are and where AI can take you, without the hype and without the chaos.

Selected projects
mcp-video Video tools for AI assistants (MCP) · 19 tools · 276 tests · Python
GitHub →
DialectOS Spanish dialect tools (MCP) · 16 tools · 20 regional variants · npm
GitHub →
EF-VA Executive assistant experiment · Python · MCP
Liminal Creative coding agent experiment · TypeScript
Cerafica Ceramics studio · live e-commerce · self-built
Visit →
Get in touch

Let's figure out if AI is the right tool
for your problem.

First conversation is free. No pitch deck. We look at what you do, what costs time or money, and I tell you honestly whether AI is a fit.

Start a project inquiry
Short Google Form—under a minute. If you only want the email newsletter, use the AI Ops brief signup instead.
Open project inquiry form

Newsletter signup is separate from project inquiries. Prefer email? Write simon@puenteworks.com.

Privacy: Form responses go through Google Workspace. Signing up for the project form does not add you to the newsletter automatically.