Doc JC-01AI process automationUS · Canada · UK · TaiwanCBAP® · PMP® · n8n Verified CreatorRev 2026.08

I automate the office work your team repeats every week.

I'm Jason Chuang, an AI process automation consultant for small businesses, including US firms of about 5–50 staff. I map how the work actually runs, then automate the repeat steps in your own accounts. I offer a free diagnostic call, a $1,500 audit, or a $3,000–5,000 build. I take on clients in the US, Canada, the UK, and Taiwan.

US company? Fixed prices in USD, working remotely from Taiwan →

How it works

  1. 01

    Free call

    Bring one process and the material behind it. You find out whether it is worth automating.

    Your time: one call

    Free

  2. 02

    Audit

    A written answer: what is worth automating, which tools fit, and what should stay with a person. Sometimes the answer is not to build.

    Your time: working sessions with your team

    $1,500 fixed

  3. 03

    Build

    I build what the call or the audit points to. You see progress as it happens.

    $3,000 – 5,000, covers steps 3 to 5

  4. 04

    Test

    I test it before your team relies on it.

    In the build price

  5. 05

    Go live and hand over

    It runs in your own accounts, with documentation your team can follow without me.

    Your time: a short training session

    In the build price

  6. 06 · optional

    Keep it running

    Monitoring, fixes, and the next automation when you are ready. Only worth it once something is live.

    $750 – 1,500 / mo

Most automation fails before anyone opens a workflow editor. The tooling is not the hard part. Deciding what the process should be — and which steps a machine should never touch — is the hard part. That decision is the work I sell. The build follows from it.

1.1

Map the process as it actually runs

Not as the org chart says it runs. I sit with the people doing the work, time the steps, and find where the hours and the money are actually going.

1.2

Decide what stays human

Then I mark the steps that should not be automated, and say why. A shorter system you trust beats a longer one you have to check behind.

Work I take on

  • Inbox triage

    Intake, triage, and routing of inbound inquiries, including customer email. A live demo on this site lets you paste a customer email. A workflow sorts it, says who should handle it, and drafts the reply. No account needed. This is a live demo, not a client case.
  • Lead triage

    A web-form inquiry is scored and routed down one of four paths: book, redirect, ask for missing info, or escalate. This was built and tested end to end as a prototype, and was not adopted by the end client.
  • Reporting

    Reporting and reconciliation that gets re-typed every week. An in-house daily executive briefing collects incoming reports, drafts a summary, then checks the figures against the source. Adding the second check raised accuracy on the figures in the summary from about 8 in 10 to about 19 in 20. This is internal work, not a client engagement.
  • Knowledge

    Knowledge that only lives in one person's head, turned into something the team can ask, with each answer showing the document it came from. A brand-consultancy assistant of this kind is being built. It is not a finished client result.
  • Document review

    Document-heavy review cycles with a fixed checklist. I don't have a published client case for this yet.

Work I turn down

  • The judgement call at the centre of your process
  • A chatbot dropped onto a website with nothing behind it
  • Replacing a system that already works
  • Work where nobody will let me measure the before

The right-hand column is not modesty. Every item on it is something I have been asked for and said no to, because automating it would have made the business worse.

WhatForDescriptionStatus
Daily executive briefingEnterprise · internalCollects the morning's incoming reports, drafts a bilingual summary, then runs a second pass that checks the draft against the source and patches it before anything is sent. Adding the second check raised accuracy on the figures in the summary from about 8 in 10 to about 19 in 20.Live
Internal knowledge assistantBrand consultancy · TaiwanTwenty years of method, made askable. Answers from the firm's own files, and shows which document each answer came from.In build
AI-visibility (GEO) auditCatering · via UK partnerSeven pages on why AI assistants were describing the business wrongly: stale schema, missing org fields, a misdated founding year.Delivered
Inbox triage demo →This sitePaste a customer email and an n8n workflow sorts it, says who should handle it, and drafts the reply. No account needed.Live

Tell me the process that costs you the most

I will tell you whether it is worth automating. If it is not, I will say so, and that answer is free. Every new client gets one diagnostic call at no charge: bring one process and the material behind it, and we look at it together. I do not build anything on that call.