Practical AI / human controlled

I turn messy business work into AI systems people can control.

I design practical systems that collect information, prepare the next action, and keep a person in control of every important decision.

Discuss a project Explore my featured build

Working full-stack demo Human approval by design Clear business purpose

Featured build / LeadPilot live demo

Lead check

Good-fit opportunity

Needs review

86fit score

This client has a clear business problem, a realistic budget, and a practical timeline. A proposal is ready for review.

BudgetGood fit

Timeline2–4 weeks

Next stepSend proposal

Waiting for your approval

Featured case study

Human-controlled automation

2working interactive demos

4controlled LeadPilot stages

56 secreal-action walkthrough

0autonomous client sends

What I build

Automation that works beyond the AI prompt.

End-to-end systems

From the first form submission to the owner dashboard, approval, and delivery.

Human control

AI prepares the work while important client-facing actions wait for a person.

Production thinking

Validation, safe fallbacks, private credentials, audit history, and responsive interfaces.

Featured case study

LeadPilot turns a new inquiry into a reviewed, ready-to-send proposal.

Built as a complete AI automation portfolio project—not a chatbot mockup. It handles intake, qualification, proposal drafting, owner approval, delivery, and audit history.

Try the live demo Open the owner dashboard

Business problem: Teams repeatedly read, sort, and answer similar inquiries by hand.
Automation: Structured AI analysis and proposal preparation with a human approval gate.
Business value: Faster preparation, consistent decisions, and complete control over client communication.

Try it in under a minute

See the automation make a real, reviewable decision.

1
Open LeadPilot and choose “Use sample data.”
2
Analyze the inquiry to create a qualification and proposal draft.
3
Review the unsent output and confirm where human approval stops the automation.

Start guided demo

56-second walkthrough

Watch every real action from intake to sending.

Follow one inquiry through form filling, analysis, proposal review, approval, and delivery. Press play when you are ready—nothing starts automatically.

LeadPilot preview showing a business inquiry ready for review

Human-style narrated guide English captions included56 seconds

How LeadPilot works

From new inquiry to approved reply in four simple steps.

LeadPilot handles the preparation. The business owner keeps the final say over who receives a response and what it says.

A new inquiry arrives

A potential client shares what they need, their budget, and when they want to start.

AI checks the opportunity

LeadPilot summarizes the request, checks whether it is a good fit, and suggests the next step.

You review the reply

A personalized proposal is prepared for you. Edit it, approve it, or decide not to proceed.

The approved reply is sent

Only the version you approve is emailed, and the full activity history stays available.

Human-first design

The AI writes the first draft. A person makes the decision.

LeadPilot never contacts a potential client by itself. The owner can review the lead, change the message, approve it, or reject it from the dashboard.

Try a sample inquiry

Nothing sends automatically

Every client-facing email waits for your approval.

Clear reasons

You can see why a lead received its score and recommended next step.

Private credentials

Service keys stay on the server and never appear in a visitor's browser.

Complete history

Lead updates, approvals, and sends are recorded in one place.

The technology

Working software, not a clickable mockup.

The demo has a real data layer, AI analysis, approval workflow, and email delivery path. It runs safely without paid services in portfolio mode.

Next.jsInterface + API

SupabaseAuth + Postgres

GeminiStructured analysis

ResendApproved delivery

Next portfolio builds

More business workflows I can turn into controlled automations.

One is a working interactive demo; the other is openly labeled as a concept. Neither is presented as a client result.

Working demo

Support Triage Assistant

Classify incoming support requests, prepare a suggested answer, and route urgent cases to a person.

Faster first response without autonomous replies

Next.jsGeminiSupabase

Try the working demo

Concept - next build

Operations Report Assistant

Turn structured weekly data into a concise management summary with exceptions that need attention.

Less time assembling repetitive updates

Next.jsGeminiResend

How I approach a build

Start with the business process, then choose the AI.

A useful automation needs more than a model call. It needs clear ownership, recovery paths, and an interface people can trust.

Understand

Map the manual work, the people involved, and the business result that matters.

Design controls

Decide where AI helps, where a human approves, and what happens when a service fails.

Build the workflow

Connect the interface, data, AI, and delivery into one usable system.

Verify and hand off

Test the happy path, errors, mobile layout, accessibility, and deployment setup.

John Francis M. Venancio Manila, Philippines / Remote

About John

I care about automation people can understand and control.

I build full-stack AI workflows that connect a clear business need to a usable interface, secure server logic, structured AI output, and human decision points. My customer-facing sales technician background helps me design for the people who use these systems every day.

Available for freelance AI automation and web app projects.

Next.jsTypeScriptSupabaseGeminiResendVercel

GitHub LinkedIn View resume

Start a conversation

Tell me which repetitive process you want to improve.

Share the current manual steps, the tools involved, and what a better result would look like. I’ll use that context to understand whether an automation is a good fit.

No credentials or private customer data Clear business context is enough Your message is never sent to an AI model