Automation · A.I Agents · A.I training for businesses

Give repetitive work to machines. Give work worth doing back to people.

sonnhat builds automations and A.I agents that run the exact tasks your team still does by hand every week, then trains your people to operate and extend them without depending on us.

Start with one taskwe do not sell "digital transformation"
Trial first, pay afterhandover only when it runs
Your team runs ittraining comes with every project
A workflow sonnhat built for a sales teamrunning
Customer messages on Zalo / emailany hour of the day
A.I reads, sorts, drafts a replyin the company's own voice
Written to CRM / Sheetnobody retypes anything
Right salesperson notifiedwith the draft to approve
A person approves in one tap·The machine does the rest
Zalo OAEmailGoogle SheetsCRMn8nMakeOpenAIClaudeGeminiLarkNotionFacebook LeadsAccountingWeekly reportsAfter-sales care

Solutions

Two things sonnhat does well for businesses.

Automation and A.I Agents

01 · Build and hand over

We map what your team does by hand, pick the single most repetitive task, build an automation or an A.I agent for it, run it for two weeks on real data, then hand it over with operating docs.

  • Reply to and sort leads from Zalo, email, forms, ads
  • Data entry, reconciliation and reports from Sheets, CRM, accounting
  • A.I agents that draft content, quotes and contracts from your templates
  • Reminders, task assignment and a morning summary for managers

Automation in detail

A.I training for businesses

02 · Teach the team to do it themselves

In-house classes by department: sales, marketing, customer care, operations, accounting. No A.I theory. Each person brings a real task to class and leaves with that task automated.

  • One day for the whole company, or a four-week track per department
  • Half-day leadership workshop: what to hand to machines, what to keep for people
  • A prompt and process kit for your industry
  • Two weeks of follow-up in a shared group

Training in detail

How we work

Four steps, starting with one task.

30-minute discovery

You describe what the team does by hand. We ask until the task that costs the most time has a name.

Pick one task

One task that repeats a lot, carries low risk and can be measured. We quote approach, time and cost before starting.

Build and trial

Built on your real data, run alongside the old way for two weeks. Your people approve, the machine does the work.

Hand over and train

Accounts, documentation and a session for whoever will run it. Done when your team can fix and extend it alone.

Tasks we have automated

Examples of work machines now do instead of people.

Described by type of task. No client names.

Sales team

A lead messages on Zalo or fills an ad form after hours. A.I reads it, scores it hot or cold, drafts a first reply in the company voice, writes it to the CRM and pings the right salesperson with the draft. Sales taps approve.

Operations

Every morning a flow pulls figures from three Sheets and the accounting software, reconciles differences and writes a five-line summary for the manager. A task that used to take one person a whole morning.

Marketing and content

An A.I agent reads notes from a consulting session, turns them into a post in the approved voice, schedules it and notifies the reviewer. Content comes from real work, not invention.

Why sonnhat

Built by someone who sold and managed sales teams first.

sonnhat was founded by Nguyen Minh Son: nearly eight years in sales and sales management, lecturer at FPT Polytechnic, host of the TikTok channel "Minh Son - A.I Kinh Doanh".

We do not start from technology. We start from the work you are doing yourself, and we only bring in a machine where it does that specific task better than a person. Where the work must stay with people, we say so.

About sonnhat

Learning from leading companies

How the companies that do this best rolled it out.

Read all six lessons

sonnhat did not invent its method. We take what proven companies did and scale it down to fit a Vietnamese business.

Shopify: prove A.I cannot do it before asking for people

CEO Tobi Lütke's 2025 memo: using A.I is the baseline expectation; before asking for more headcount or budget, a team must show why A.I cannot do the job.

Toyota: the machine stops, the person decides

The jidoka principle: machines detect abnormality and stop; humans handle it. Exactly how sonnhat places an approval step in every workflow that carries risk.

Klarna: automated everything, then had to come back

Put A.I in front of most customer chats, then the CEO admitted quality fell and reinvested in people. Lesson: automate tasks, not relationships.

Automation is not about fewer people. It is about people doing less of the machine's work.

sonnhat's principle

Get started

Tell sonnhat about one task your team still does by hand.

The first call is 30 minutes, online, free. By the end you know how far that task can be automated, with what, and how long it takes. If it is not a fit, we stop there. No commitment.