[Scenario]Sales & Marketing Scenarios: AI for Data Analysis & Proposals
Summary: Let AI help you with customer profiling, sales analysis, and customized proposals.
1. The Problem: Sales Folks' "Overtime Triple Threat"
If you work in sales or marketing, these scenes will hit home:
- Customer data is a total mess: Your CRM (whether it's 销售易, 纷享销客, or any customer management system) holds hundreds or thousands of customer records. But if you want to slice them by tags like "30-40, female, new mom, monthly income over 10K," the CRM only gives you raw fields — you're stuck filtering records one by one.
- Reports that wreck your week: When the boss asks "Which SKU (i.e., product code) sold best in East China last month?" you have to open Excel, pull the data, write formulas, and build a PivotTable (a dynamic table in Excel that auto-summarizes data). Half an hour vanishes before you've answered a single question.
- Proposal decks that make you want to cry: Every new customer means rewriting the proposal from scratch. 80% of the content is the same, but you can't copy-paste — because it has to look tailor-made for that specific buyer.
Add these three chores up, and they'll swallow 10-15 hours of your week. Today, we're going to hand them off to AI so you can focus on what actually needs a human touch — talking to customers.
2. Goals: Three AI Weapons for Today
In this article we'll nail down three things. Once you're done, your sales and marketing workflow will look like this:
| Weapon | What it does | Input | Output |
|---|---|---|---|
| AI Customer Profiling | Tag and segment customers | CRM data + public data | Customer tag table + segmentation suggestions |
| AI Sales Analysis | Ask data in plain English; AI auto-generates charts | Natural-language questions | Charts + one-line conclusions |
| AI Proposal PPT | Generate customized proposals in one click | Customer requirements doc | PPT outline + first draft |
Tool stack: 销售易 / 纷享销客 (CRM data source) + Power BI / Excel (data analysis) + general AI (Kimi, Qwen, Wenxin Yiyan, ChatGPT — any chat-capable AI).
3. Hands-On Cases: A Step-by-Step Walkthrough
Case 1: AI Customer Profiling — Give Every Customer a "Persona"
Step 1: Pull the data together
Export a customer table from your CRM with at least these columns: customer name, industry, role, last contact date, deal size, notes. Then prep a public-data sheet — say, company size, registered capital, and industry — scraped from Tianyancha or Qichacha.
Step 2: Write the prompt for AI
Paste both sheets to the AI and use this prompt (feel free to copy it verbatim):
You are a sales and marketing analyst. Below are two customer datasets:
1. Internal CRM data (table)
2. Public business-registration data (table)
Please complete two tasks:
① For each customer, generate 5-8 persona tags (reference dimensions: industry, size, decision role, purchasing power, potential needs, follow-up stage).
② Segment customers into three groups: high-value follow-up, nurture-potential, low-priority. For each group, write one sentence of follow-up advice.
Output format: Markdown table.Step 3: Push the results back into CRM
What you get back is a tagged customer table. Ask IT to write a VBA macro (VBA is Excel's built-in scripting language — think of it as a tiny program that tells Excel to run tasks automatically), or use Power Query (Excel's dedicated tool for cleaning and merging data) to bulk-write the tags back into CRM. Done.
Time saved: from 3 hours down to 30 minutes.
Case 2: AI Sales Analysis — Ask Data in Plain English
Step 1: Pipe the data into Power BI / Excel
If you're already on Power BI, just import the sales-detail table (order ID, date, region, SKU, amount, salesperson). If you only have Excel, reshape the data into a flat/one-dimensional table — one row per order, one column per field.
Step 2: Ask questions in natural language
You don't need to learn DAX (DAX is the formula language used for calculations in Power BI and PivotTable) or any functions. Just open a natural-language-capable AI (Power BI's built-in Q&A, or general AIs like Kimi / Qwen) and type:
"Which SKU sold best in East China last month? List the top 5 and calculate their share of total revenue."
Step 3: Paste in the data and AI drafts an analysis + a one-line takeaway
When you paste the data in as well, the AI can produce an analysis like: "Last month's East China sales champion was SKU-A123, taking 23.5% of that region's total revenue." Note: general chat AIs cannot see the data inside your Power BI, and they will not execute anything on your behalf; any numbers or charts must be based on the real data you paste in, verified by you before use (to "ask questions directly inside Power BI", you need Power BI Copilot — see B21).
Need deeper analysis — say, "Compare to last year and find the three regions with the steepest decline, plus the reasons"? Ask in natural language all the same. The AI can help you translate the question into a DAX approach — but running the query is up to you inside Power BI (a general AI cannot execute it for you).
Time saved: each question drops from 15 minutes to 30 seconds.
Case 3: AI Proposal PPT — Turn Customer Requirements into a Proposal in One Click
Step 1: Hand the customer's requirements doc to AI
The customer has sent you an RFP (Request For Proposal) or a requirements-meeting summary, PDF or Word — either works. Copy the key sections (project background, pain points, goals, budget, timeline) and paste them to the AI.
Step 2: Write the prompt to generate the PPT outline
You are a senior sales rep. Based on the customer requirements document below, write a customized proposal PPT outline.
Requirements:
① Clear structure: cover, project understanding, pain-point analysis, solution, value proposition, implementation plan, pricing, case studies, team, contact us.
② Each section should list 3-5 core points.
③ In "Pain-point Analysis" and "Value Proposition," quote the customer's original words directly so they feel "you really get me."
④ Use the customer's industry terminology throughout — no generic template language.
Output in Markdown format.Step 3: Use an AI tool to generate the PPT first draft
Paste the outline into a PPT-capable AI tool (recommended: Gamma, iFlytek Zhiwen, Tongyi Zhiwen — note that "Tongyi Tingwu" is an audio/video transcription tool; for making PPTs use "Zhiwen"). Within 10 minutes you'll have a 15-20 page draft. Then you only need to do three things: swap in the customer's logo, adjust the color palette to match their VI (Visual Identity — the customer's color scheme + fonts + logo style), and drop in screenshots of your past success stories.
Time saved: from 6 hours down to 1.5 hours, including revisions.
4. How It Works: What AI Is Doing Behind the Scenes
In short, the three cases sit on three different AI capabilities:
- Customer profiling: at its core, it's "text classification + clustering." AI tags and segments customers based on the dimensions you provide. The engine under the hood is the large language model's grasp of "industry context."
- Sales analysis: at its core, it's "natural language → code." What you say gets translated by AI into DAX / SQL queries that run on Power BI or the database, then visualized. Power BI's built-in Q&A works exactly this way.
- Proposal PPT: at its core, it's "templated generation + personalized injection." AI applies a standard proposal structure but slots in the customer's verbatim language in key chapters (pain points, value), giving you both speed and customization.
Remember this: AI doesn't replace your professional judgment — it pulls you out of repetitive grunt work so you can spend time on the parts that need a human touch.
5. Pitfall Guide: 5 Traps in Sales & Marketing AI Work
| Trap | How to avoid it |
|---|---|
| Trap 1: Feeding customer data straight to AI and leaking privacy | Anonymize before upload — replace customer names, phone numbers, and emails with "Customer A," "Contact 1." Keep sensitive data on local machines; don't push it to public AI services. |
| Trap 2: AI tags are inaccurate, but you apply them anyway | AI output is "suggestion," not "verdict." First-time use: validate against historical tags. If accuracy is below 80%, don't deploy — tune the prompt first. |
| Trap 3: Sales analysis uses hallucinated data | Every number in an AI-generated chart must be checked against the source table. Make it a habit: look up each number in Excel. |
| Trap 4: Proposal PPT is pure AI and reads like a machine wrote it | At minimum, edit two spots: ① swap in the customer's verbatim quotes; ② add a short paragraph like "How we understand the real pain points of your industry." |
| Trap 5: You only run it once and never iterate | Save the prompts that work. Build a "Sales AI Prompt Library" and reuse them. Prompts become more valuable as you accumulate them. |
6. Advanced Extensions: Wire This Playbook into Your Existing Systems
If you want to take the impact up another notch, try these directions:
- Push to WeCom / DingTalk: Have AI auto-push daily sales-analysis results into the group chat. By 9 a.m. your boss sees yesterday's war report — no need to ask you.
- Batch proposals: Batch-feed customer requirements to AI and generate first-draft proposals for 10 customers in one go — done in under an hour. Spend the remaining time fine-tuning each one.
- API integration with CRM: Both 销售易 and 纷享销客 expose APIs. Ask IT to write a script so AI reads CRM data and writes tags back automatically — fully unattended end-to-end.
- Pair with VBA / Power Query for landing: If your customer base is large, learn a little VBA (even just record-a-macro and tweak it) or Power Query. They bridge AI output straight into your Excel workflow.
Quick wrap-up: In sales and marketing, what AI really helps with isn't "doing the selling for you" — it's "doing the dirty, tedious, repetitive work for you": tidying customer records, running the numbers, drafting proposals. Take that saved time and, honestly, go have coffee with a customer, or just chat. That human touch is something AI can't learn.