Qwen Office + AI Spreadsheets: Building a CRM for the AI Era

Build Your Own Custom CRM AI Assistant in Just Half an Hour

A Practical Guide to QwenWork

After spending a fortune on a CRM system, it often ends up becoming just another cookie-cutter tool—

Three months after launch, despite round after round of training, the only people consistently using it are the IT admins maintaining the system, the sales reps entering leads, and the managers who check the data at the end of the month.

Sales reps say, "After running around all day, I still have to update follow-up records at night. I'm not just writing logs; I'm writing my memoirs."

Managers say, "Half the fields in the forms are empty. Are we supposed to make decisions based on this, or just practice our imagination?"

IT says, "Your requests are queued for next quarter. Please make do with what you have for now."

The problem isn't a lack of effort; it's that we got the direction wrong from the very beginning: we are forcing people to serve the tools.

Shift Your Mindset: Let the Tools Serve You

The issue with traditional CRMs is that the tools are passive: if you don't push them, they don't move.

In the AI era, what we need isn't just a better CRM, but an AI employee that can proactively drive work forward—

Proactive at work: Every day at 9 AM, the morning report is automatically sent to DingTalk, prioritizing which clients to follow up with, which deals to push, and which payments to urge, all sorted by urgency without needing further instructions.

Proactive at organizing: After a call, you just need to say, "The client said they'll pay after the holiday," and it will expand this into a complete follow-up record and archive it; contracts and proposals are stored together as well. The next time you write a new proposal, it will automatically review past documents and communication records with that client for reference.

Proactive at analyzing: Tables are converted into dashboards with one click; quarterly reports are no longer just a list of data, but clearly explain the current situation, identify problems, and provide solutions.

Today, we'll show you how to use "QwenWork + DingTalk AI Tables + DingTalk" to personally build an AI employee that automatically drives your sales forward, ensuring your CRM no longer gathers dust—

First, Let's See It in Action

Seeing is believing. Let's first take a look at how this AI employee works in practice:

Interaction Layer - DingTalk: View Data Anytime, Anywhere

‹‹ Swipe left to view

Data Layer - DingTalk AI Tables: Data-Driven, Accessed by Permission

‹‹ Swipe left to view

Execution Layer - QwenWork: You Command, It Executes, Visualized

‹‹ Swipe left to view

Each of the three plays to its strengths, and together they form a complete AI employee.

Build It Yourself in Four Steps

01 Step 1: Define the Role

When hiring an employee, the first thing to do is provide context: what business you are in and what tasks you want it to handle.

Only with clear context will it know what to focus on.

Open QwenWork and send it this message directly:

Prompt (copy directly): Hello, I am the Head of Sales at a medical device sales company. I want you to be my sales operations assistant, helping me follow up with clients, advance team projects, and remind me of to-do items daily. You are goal-oriented and need to coordinate well with me. We primarily sell XX products, with a strong focus on client payments and renewals.

There are two key points here:

Always specify the industry. Client logic varies completely across different industries. In healthcare, it's about payments and renewals; in government, it's about process nodes; in e-commerce, it's about repurchases. Being clear about this tells it where to direct its efforts.

Don't worry about providing too much background information. What products you sell, how many people are on the team, what metrics to focus on—feed it all the information you have. The more information it has, the closer its judgments will align with your business.

This content will be stored in QwenWork's long-term memory (you can view and modify it anytime in "Settings" -> "Memory"). In every future conversation, it will carry this context with it.

02 Step 2: Set Up the Data

You don't need to build the tables yourself; just let the AI do it with a single sentence:

Prompt (copy directly): Please create a DingTalk knowledge base named "AICRM" for me. Inside it, create an AI table called "Customer Management" with four sheets: Customer, Team, Follow-up Records, and Payments. Also, create a folder to store documents like contracts, proposals, and quotes.

If you haven't connected DingTalk yet, go to the "Connectors" in QwenWork to enable the DingTalk connection first.

Once the tables are created, have it read through its own "database" first:

Prompt (copy directly): Read through the tables in the AICRM knowledge base and tell me what fields each table has.

Here are two tips:

One follow-up record table is enough; no need for manual categorization, as that's the AI's job. Just tell it directly, "Log a follow-up for me," and it will handle writing to the table, linking the client, and updating the status on its own.

Contracts, proposals, quotes, and meeting minutes transcribed by AI listening—all of these can be stored in the knowledge base. When you need to write a new proposal for a client later, it can pull up past proposals and communication records for reference. Once the proposal is generated, you can download it as a Word or PDF file with one click and send it directly to the client.

03 Step 3: Set Up the Workflow

This is the crucial step from "passive" to "proactive."

Let's do a test run first:

Prompt (copy directly): Check which clients need to be followed up with today and which payments are overdue. Make a list, and include suggested talking points for each item.

Once you're satisfied with the results, you can even upgrade it to a scheduled task:

Prompt (copy directly): Set the workflow we just ran as a scheduled task to execute every day at 9 AM, and send the results to me. If any item corresponds to a specific person in charge, send the relevant content to their DingTalk.

From then on, every morning you will automatically receive a sales morning report in your DingTalk: high-priority clients, today's to-dos, and payments to chase, each with suggested talking points ready to use directly.

Daily follow-ups also become much easier. After meeting a client, just throw a simple, plain-language sentence to your AI employee:

Just sent a payment reminder to client XX. They said they'll transfer the money tomorrow. So happy!

It will automatically write the follow-up record, link the client, update the payment status, and even casually remind you: verbal promises don't count, remember to confirm the receipt of funds tomorrow.

04 Step 4: Optimize the Experience

The first version of the AI employee you build will most likely be a "data mover." For example, if you ask it to prepare a quarterly report, it might just spit out a pile of dashboard screenshots, page after page of listed numbers.

Don't settle for that. Just have it revise it directly:

Prompt (copy directly): The current content is too much like a running account and lacks insights; it doesn't count as a report. Summarize the current situation in one sentence, put the numbers later as evidence, and conclude with next steps and recommendations.

The revised version will be completely different: it starts with one sentence summarizing the situation—"Clients are growing, but cash flow isn't keeping up"—followed by 837,000 in pipeline opportunities, a 22% collection rate, and which client types have concentrated overdue payments, breaking it down layer by layer, and finally concluding with what to do next.

For AI-generated results, 80% of the quality comes from guidance, and 100% comes from iteration. If you're not satisfied with something, express it in plain language and have it revise it. Each round of revision makes it better.

Advanced Playbook

Once the basic version is up and running, you can continue to expand:

Create a sales dashboard: Have QwenWork convert the data in the AI tables into a dashboard with one click—client distribution, sales funnel, payment progress, and team workload are all displayed on the dashboard. The dashboard can be cast to the department's large screen for daily viewing and morning meeting discussions.

Automated follow-up on group messages: If you have dedicated DingTalk groups with clients, you can have QwenWork periodically organize key information from client communications and extract follow-up items to reduce manual data entry. QwenWork can also read chat records from Feishu and WeCom through connectors.

Integrate with existing CRMs: If you already have a CRM system, have your IT colleagues turn the system's data interface into an MCP to connect with QwenWork, allowing QwenWork to work on the original system.

Set data permissions by role: When multiple people share the same data, you can configure access scopes through the advanced permissions in AI Tables. For example, allow sales reps to view only the clients they are responsible for, and let managers view the data of the teams they manage.

Conclusion

AI Tables are like universal building blocks; you can build any lightweight business system with them.

But "being able to build" is one thing, and "knowing how to build" is another—many users report a steep learning curve and find it hard to get started.

Now, tasks like building tables, fetching data, and executing workflows can all be handed over to QwenWork; you just need to instruct it in plain language.

DingTalk manages accounts and reach, AI Tables manage data and permissions, and QwenWork handles the execution.

With this iron triangle combined, you can build most business systems yourself.

What would you like to see us practice in the next article? Leave a comment in the comments section~

· · ··· · ·

An in-depth column to make QwenWork even more useful, turning everyone into an Agent expert

We dedicated to serving clients with professional DingTalk solutions. If you'd like to learn more about DingTalk platform applications, feel free to contact our online customer service or email at This email address is being protected from spambots. You need JavaScript enabled to view it.. With a skilled development and operations team and extensive market experience, we’re ready to deliver expert DingTalk services and solutions tailored to your needs!