[Scenario]SME Owner: AI as Your All-in-One Assistant

Summary: Tight budget? Let AI cover half your team — copy, support, dashboards in one click.

Preface

Module E is the "Scenario" track — it picks tools based on who you are. E5 is written for SME owners, sole proprietors, family-run shops, and early-stage teams (3–10-person studios). These readers share one profile: small team, tight budget, scattered workload.

This article gives you a "save-headcount" recipe for the three most frequent scenarios: AI marketing copy, AI customer support, and AI data dashboards. Every case prioritizes free or low-cost SaaS — leaning on the built-in capabilities of 微信, 抖音, 飞书, and 钉钉 first, so you don't waste money on things you don't need.

Pain Point: The Owner Is the Whole Company

Small business owners almost all live by the same daily schedule:

  • Morning: reply to customers, coordinate with suppliers, watch the store dashboards.
  • Afternoon: write posts for 朋友圈 / 小红书 / 抖音, approve staff schedules, handle after-sales.
  • Evening: check the numbers, plan tomorrow's promotion, scroll peer accounts for inspiration.

The result: you run flat-out, yet the results aren't great — 朋友圈 posts feel flat, slow replies trigger bad reviews, store data never gets looked at, let alone turned into decisions.

And the harshest part is the bill: a copywriter costs 4–6k a month, a support agent 3–5k, a data analyst 5–8k. Three people add up to 15k — a small shop can't carry that; hire just one and the workload still won't fit.

Good news: the biggest shift in the 2024–2025 wave of AI tools is that jobs "you used to pay a specialist to do" have been compressed into "things the owner finishes with a few clicks." Below are three highest-frequency tasks, plus a playbook you can actually deploy on a tight budget.

Goal: AI Covers Half the Team

After reading this article, you should at minimum be able to let AI cover these three blocks:

Scenario Tool Combo Hours the Owner Saves per Week
Marketing copy 豆包 / 千问 / Kimi + 小红书 built-in AI 4–6 hours
24/7 support Coze（扣子）/ Dify + WeChat customer service 2–3 hours per day
Data dashboard 飞书多维表格 / 钉钉智能表格 3–5 hours

Combined, that's two to three workdays a week. Time the owner can spend on what only humans should do: walk with customers, watch trends, shape strategy.

Let's get practical — three cases, one by one.

Case Studies

Case ①: Auto-Write Copy for 朋友圈 / 小红书 / 抖音 with AI

Pain point, unpacked: every day the owner has to publish content to WeChat 朋友圈, 小红书, and 抖音. A single well-written 小红书 post takes 30–60 minutes — one person can't keep up. And the style differs by platform — 朋友圈 feels like chatting with friends, 小红书 needs a "planting the seed" vibe, 抖音 needs a hook in the first 3 seconds.

Tool picks (all free or low-cost):

  • 豆包 (ByteDance ecosystem, free, native to the 抖音 world).
  • 千问 (Alibaba ecosystem, free).
  • Kimi (Moonshot AI, the free tier is enough).
  • 小红书 built-in AI assistant (first choice for writing 小红书 posts).

Steps:

Step 1: Lock in one core message. One sentence that says what today's post is about. For example, "A new batch of osmanthus oolong just arrived — buy one, get one free today."

Step 2: Use this prompt template (copy-paste into any general AI):

You are my private-domain copywriting assistant, fluent in the styles of WeChat Moments (朋友圈), Xiaohongshu (小红书), and Douyin (抖音).
Based on the product info below, write 3 versions of the copy:
1. WeChat Moments version: conversational, like chatting with a friend, under 80 characters, ending with exactly 1 emoji.
2. Xiaohongshu version: catchy title, 3–5 hashtags, body broken into bullet points, with a strong "I need this" vibe, under 200 characters.
3. Douyin voiceover version: a hook in the first 3 seconds, fast-paced, under 60 characters.

Product info: <fill in your product and offer>
My shop is called <shop name>; its vibe is <young / artsy / budget-friendly…>

Step 3: Let AI output all three at once, then pick the version you like best. If one version falls flat, just tell it: "Make the 小红书 version more lively, drop in the meme '打工人续命神器'" — and let it revise. This is iterative prompting (i.e., the craft of refining what you say to the AI, round by round, until it lands).

Step 4: Tweak 30 seconds by hand, then post. AI copy can come out a little "too clever" or with off details (prices, times) — double-check the key numbers yourself before posting.

Before / after:

  • DIY: 30–60 minutes per post, with no guarantee of internet flavor.
  • AI draft + tweak: 3 versions in 5 minutes, pick one, polish it, post.

💡 Pro tip: Save the prompt above in your phone's "Shortcuts" or WeChat "Favorites." Every day before posting, open it, copy-paste, and swap out only the "Product info:" line.

Case ②: AI Customer Support — Build a 24/7 Q&A Bot with Coze / Dify

Pain point, unpacked: the biggest headache for a small shop is "someone messages at midnight." The owner sleeps, the customer leaves; the owner replies, the customer has already bought elsewhere. Hire someone for round-the-clock coverage? That salary alone runs into the thousands — a small shop can't carry it.

Tool picks:

  • Coze（扣子）: made by ByteDance; the domestic version is fully free, and building a bot feels like snapping together Lego. Best for beginners.
  • Dify: open-source, with fuller enterprise features, and supports local deployment.

What's RAG? First, a term — RAG (Retrieval-Augmented Generation). In plain words: give the AI a "product manual," so before answering a customer it flips through the manual first, instead of guessing from "imagination." That manual can be your shop FAQ, product specs, return policy, or price list.

Steps:

Step 1: Sign up for Coze (www.coze.cn) and click "Create Bot."

Step 2: Give the bot a role. In the "Persona & Reply Logic" box, write:

You are the customer-support assistant for XX Tea House. You mainly answer customer questions about products, prices, delivery, and promotions.
- Tone: warm and friendly, like chatting with a friend — nothing stiff or corporate.
- If you don't know something: say so honestly ("Great question — let me pass that one to the owner for you"), and never make things up.
- Questions you must always be able to answer: the menu, delivery range, opening hours, and current promotions.

Step 3: Attach a knowledge base — this step is the key; it's where RAG lands.

  • Click "Knowledge Base" → "Create Knowledge Base" → upload files (PDF / Word / 飞书 Doc all work).
  • Upload your menu PDF, return policy, and FAQ docx.
  • Coze auto-chunks and indexes them, so when a customer asks "do you have any cold-brew teas," the bot flips through your menu before answering — it doesn't answer from "impression."

Step 4: Publish to WeChat / your website.

  • Coze supports one-click publishing to "微信公众号 Support," "WeChat Mini Program," and "web floating widget."
  • Pick the channel you need, follow the prompts to bind — done in 10 minutes.

Step 5: Test before going live.

  • Ask 5–10 questions in the support window yourself; check whether the answers are accurate.
  • Have a colleague or friend play a "tough" customer, e.g., "what if this tea is past its expiration date?" — to test the bot's fallback.

Result: a customer asks at midnight, "Do you have cold-brew tea?" — the bot replies within 30 seconds: "Yes! We have 3 cold-brew options: osmanthus oolong / white peach oolong / jasmine green tea — best flavor after 8 hours of cold brewing~" — and the owner sleeps soundly.

Case ③: AI Data Dashboard — Store / Website / WeChat Group Data on One Screen

Pain point, unpacked: a small owner's "data" lives in seven or eight places — WeChat group order chains, 有赞 / 微店 orders, 美团 / 抖音 store dashboards, 小红书 backend, website analytics… every day you open five or six dashboards and your eyes glaze over, and you still can't tell "how much we sold today, which product is best, which group is most active."

Tool picks (by budget):

  • 飞书多维表格: the free tier is enough for a small team, with built-in AI fields — strongly recommended.
  • 钉钉智能表格: for small teams already in the DingTalk ecosystem; ties in with attendance and approvals.
  • Excel + Power Query: Power Query is Excel's "data mover" — it can pull from multiple sources and merge and clean them automatically. A good fit for the solo owner who lives in Office.

Steps (飞书多维表格 edition, zero code):

Step 1: Create a Bitable with a few sub-tables:

  • "Daily Orders" (fields: date, platform, order ID, amount, product name).
  • "Customer Sources" (fields: date, source channel, new customers).
  • "Group Activity" (fields: date, group name, active speakers, new members).

Step 2: Bring the data in.

  • Most platform dashboards have an "Export Excel" option; export daily or weekly, paste into the right table.
  • Advanced: 飞书多维表格 supports webhooks (auto-callback interfaces); you can have AI write a small script to sync automatically — but when you're starting out, just import manually. Don't over-engineer on day one.

Step 3: Use AI fields to generate summaries.

  • Create a new "Dashboard" (the Kanban view inside 飞书).
  • Add an AI field with the prompt: "Based on this month's order data, summarize the top 3 best-selling products, the 3 highest-sales days this month, and the growth rate vs. last week."
  • 飞书 AI auto-runs a written summary — the owner doesn't stare at a pile of numbers; trends at a glance.

Step 4: Set up an auto-pushed "Daily Dashboard".

  • 飞书多维表格 supports "Scheduled Send" — set a time (9:00 a.m. daily) and the day's dashboard auto-pushes to the owner's 飞书 / 微信.
  • Every morning at 9, the owner opens the phone and sees not a wall of numbers but AI-written text: "Yesterday's sales: X yuan, +Y% vs. last week; best-seller: osmanthus oolong."

Power Query alternative (if your daily driver is Excel):

  • Open Excel → Data → Get Data → From Folder.
  • Drop the weekly exports into one folder; Power Query merges them automatically.
  • Then build a PivotTable (Excel's tool that "auto-rolls up a pile of detail rows into any dimension you want") as your dashboard.
  • Advanced: have AI write the M-formula Power Query script for you — you don't need to know M; just tell AI what you want and it'll write it.

Principle Recap: The Same Logic Behind All Three Cases

The three cases look different, but they're the same playbook — the standard workflow for small businesses in the AI era:

  1. Write copy: turn "the scattered ideas in the owner's head" into "a prompt the AI understands" → AI produces → human tweaks.
  2. Run support: turn "the shop's product knowledge" into "a knowledge base the AI can query" (RAG) → AI auto-answers → fall back to human.
  3. Read data: turn "data scattered across platforms" into "one roll-up table" → AI auto-analyzes → owner sees it on one screen.

Lock this logic in: whatever new scenario you meet, first ask yourself — is this "producing content," "answering questions," or "reading data"? Then pick the matching tool. What other owners pay a specialist for, you may finish in five minutes by handing it to AI.

Pitfall Guide: The Five Traps Small Owners Most Often Fall Into

  1. Buying expensive SaaS on day one. Many "AI marketing platforms" or "AI customer service systems" run hundreds to thousands a month. Start by getting the work done with free built-in capabilities in the WeChat ecosystem, 飞书, and 钉钉, then think about upgrades.
  2. Pushing the support bot live without testing. The worst thing an AI agent can do is "make stuff up" — a customer asks "do you have a sugar-free version," it's not on your menu, and AI invents "we have a zero-calorie version" — instant crash. Run a small test; ask every question you can think of first.
  3. Feeding sensitive data straight into AI. Customer lists, unreleased financials, core recipes — best to scrub them (strip names, phone numbers, exact amounts) before uploading. Or use a locally-deployed Dify / Ollama to keep data on your own machine.
  4. Posting AI copy as-is, no eyeballs on it. About 80% of what AI writes is good, but key numbers, names, prices, and times must be checked by you — especially promo copy; a wrong price costs real money.
  5. "AI can do everything". AI shines at "repetitive, templated, information-dense" work. Judgment, decision-making, talking to customers, shaping strategy — these still belong to the owner. Spending the saved time where it counts — that's the real value of AI.

Advanced Extensions: From "Using Tools" to "Orchestrating Workflows"

The three cases are the on-ramp. For owners who want to take one more step, here are four paths — pick by ROI.

Path A: Multi-Agent Orchestration — Chain AI into a Pipeline

What is multi-agent orchestration? In plain words: let multiple AIs each do their own job, then chain them together — one AI writes copy, one AI generates images, one AI auto-publishes. Coze's "Workflow" feature does exactly this:

  • Node 1: the owner enters "today's product" → triggers Node 2.
  • Node 2: AI writes 朋友圈 / 小红书 / 抖音 versions → triggers Node 3.
  • Node 3: AI generates the 小红书 cover image (using Coze's built-in image generation) → triggers Node 4.
  • Node 4: copy + image auto-save to "Drafts"; the owner confirms and publishes with one click.

ROI trade-off: don't build complex workflows on day one; get a single step running smoothly first, then think about chaining. Multi-agent is an efficiency multiplier, not an on-ramp requirement — master 1 bot before chaining 10.

Path B: Private Domain (your owned customer pool, e.g., WeChat groups, contact lists) + Full AI Stack

String cases ①②③ together into one line:

  • 朋友圈 / 小红书 to attract traffic → add on WeChat → join the group → AI support answers → data dashboard tracks conversion.

Tool combo: WeChat ecosystem + 飞书多维表格 + Coze (all basically free).

Path C: Advanced Players — Use VBA Macros / Power Query to Automate Daily Work

If your daily driver is Office, you can have AI write VBA macros for you (VBA = Visual Basic for Applications, the scripting language that lives inside Office and automates repetitive work in Excel / Word):

  • Auto-export orders from 有赞 / 微店 → write them into a local Excel → trigger PivotTable refresh → email the summary to the owner.
  • Auto-group daily orders by customer → use mail merge (Word's tool that "applies one template to hundreds of recipients at once") to batch-generate thank-you notes.

VBA comes with Office for free; you only need to tell AI in plain English "I want Excel to automatically do X," and it'll write the script.

Path D: Build a CRM (Customer Relationship Management) Prototype

Many small owners still manage customers in Excel, but 飞书多维表格 + AI fields can already build a lightweight CRM:

  • Customer table: name / source / first-order date / lifetime spend / preferred category.
  • AI fields auto-calculate "customer value tier" (high / medium / low) and auto-remind "time to reach out to this loyal customer."

Much tastier than SaaS CRMs that cost thousands a year — free, customizable, and tightly connected to 飞书 / 钉钉 workflows.

Wrap-up + Next Article

The core mindset for SMEs using AI: split your work into "produce content / answer questions / read data," and for each one get a single point running with a free or low-cost tool first; once it's smooth, think about chaining into workflows or moving to pricier platforms.

That wraps up the industry-scenario articles. Next we move to the community track: F1, From User to Author — turn your "team of one" know-how into a tutorial that helps others. See you in the next article.

This article is part of vba.net's "AI-Powered Office" series, Scenario track E5. Author: 贺迅.