Project direction · Lean businesses

AI systems

Small team. More capacity to build.

A direction for developing AI systems for student founders, solo entrepreneurs and small businesses. Start with a specific task, connect tools and workflows, and reduce repetitive work to make more time for customers and create practical value.

02 / Applications

Start with one task

These possible applications are a starting point for discussion. A suitable workflow depends on each project’s needs, data and way of working.

Research & content

Summarise materials you have permission to use, organise ideas and prepare draft content. An accountable person checks sources, verifies information and approves publication.

Customer conversations

Sort questions, identify missing information and suggest replies. Pricing, service terms and commitments require approval from an authorised person before sending.

Coordination & follow-up

Turn confirmed requests into task lists, reminders and progress summaries. An accountable person sets priorities, assigns work and handles exceptions.

03 / The approach

From a need to a controlled trial

  1. Understand the workflow

    Choose one problem, define the inputs and intended output, and record time spent, common errors and data permissions as a baseline for comparison.

  2. Build a prototype

    Use synthetic or permitted data, test each step, identify human approval points and define how to return to manual work when something fails.

  3. Run a limited pilot

    Agree on scope and responsibility before a trial. Compare time, quality, cost and ease of handover to decide whether to adjust or expand.

04 / An illustration

An example: from a question to a draft reply

A static illustration using synthetic data. This page does not run AI, connect to external systems or send messages. Open each step to explore the workflow.

01 / Receive a sample request

A student developing project A wants to organise incoming questions.

Sample brief: project A, a one-person team, a need to sort questions. Question types, volume and current tools are not yet known. No real personal data or customer records are used.

02 / Prepare a draft

The illustrative draft asks for more information to understand the need.

“Thank you for sharing project A. What kinds of questions do you receive, which tools do you use to manage them, and which step would you like to improve first? An anonymised example would help.”

03 / Human review

Check the content, make changes and decide on the next conversation.

An accountable person checks suitability, data permissions and unconfirmed details. A draft is sent through a real channel only after approval; this page’s example does not send anything.

05 / Working principles

People remain responsible for decisions

Data you have permission to use

Use only the data needed, with defined access and retention. Begin with an anonymised example; initial enquiries do not need account credentials or customer records.

Approval before important actions

Public content, quotes, commitments and actions involving money or data need checks and approval appropriate to their risk.

Measurement & human handover

Monitor errors and output quality, assign someone to handle exceptions and keep a way to stop. Choose the degree of automation for each task using trial results.

06 / Connect

Start with your workflow

Contact Le Kim Uyen to discuss your needs and a suitable direction for a trial. Send a short description by email or get in touch by phone.

The implementation scope, provider, costs and responsibilities will be defined in a separate agreement before work begins.