AI workflows for small businesses: answers, translation and sector tools
I designed client-facing AI workflows for small businesses (answers from the firm's own documents over web chat and WhatsApp, English and Chinese document translation with the firm's own rules, and tools for their sector), and they cut turnaround time by around 70%.
- Sector
- Small service businesses, including consumer services firms and businesses that work in English and Chinese
- Stack
- C#, ASP.NET Core, SQL Server, n8n, OpenAI, WhatsApp
- Status
- Built for client businesses
The problem
Small teams answer the same customer questions every day by digging through their own files. Documents that need to go between English and Chinese wait for whoever can translate them.
Sector tasks, such as matching a customer to the right product, take a specialist's time even when the answer follows a pattern. Customers wait, and the team's best people spend their day on repeat work.
What I did
I built a reusable client portal and set it up for each business with its own site, members and settings. Inside it:
- An assistant answers customers on the website or WhatsApp from the business's own knowledge: uploaded files are analysed and indexed, and the business keeps its own list of approved FAQs.
- Document translation between English and Traditional or Simplified Chinese runs on prompts and rules the business sets itself, with results shown page by page so a person can review them before use.
- Sector-specific tools go on top where they fit, for example product questions and matching for a consumer services business, or support tickets with chat for another.
The workflow logic runs in n8n, connected by webhooks, so I can change a workflow for one client without rebuilding the portal.
Tools and stack, in plain words
ASP.NET Core in C# with SQL Server for the portals, n8n for the AI workflows, OpenAI models, a vector store for each business's documents, WhatsApp messaging, and speech-to-text and text-to-speech where voice helps.
A separate Node.js service gives chat assistants a consistent persona and memory, for use over WhatsApp and in n8n.
Outcome
Across the client-facing AI workflows I designed, turnaround time came down by around 70%. Routine questions get answered from the firm's own material straight away, and translations arrive as a draft to review rather than a job to start.
What it shows about how I work
I build one solid base and fit it to each business, rather than starting from zero every time. And the business stays in charge of its own knowledge, rules and final answers.
FAQ
Can a small business use AI to answer customers from its own documents?
Yes. Him Yip sets up assistants that answer on the website or WhatsApp from a firm's own files and approved FAQs, and across his client AI workflows turnaround time fell by around 70%.
Can AI translate business documents between English and Chinese reliably?
As a first draft, yes. In Him Yip's workflows each business sets its own translation rules, and results come back page by page for a person to review.