6 Doubao Agent Tricks for Excel That Double Productivity for Office Workers

Handling Excel spreadsheets has long been a frequent pain point for professionals: customer information is scattered across chat histories, resume screening relies on human eyes going through each one, data cleaning is more exhausting than re-entering data, and writing formulas often results in errors... These scenarios are being transformed by AI Agents. A popular video on Bilibili demonstrated six tricks for using Doubao Agent to handle Excel, covering the entire workflow from data organization to analysis reports. Below, based on the video content, we break down the practical logic and key operational points of each trick.

Trick 1: One-Click Organization of Cross-Channel Information into Tables

Anyone in sales or customer operations knows the feeling: customer information is often scattered across chat histories, documents, and even meeting notes. Manually copying and pasting into a follow-up table is not only time-consuming but also error-prone. The video demonstrates opening Doubao on PC, dumping all kinds of materials to it at once, selecting the "Table Skill," and entering an organization prompt. Scattered information is automatically summarized into a clear and complete table, with deduplication completed.

An extended scenario for this trick is mobile work. When your boss sends several customer documents on the weekend or after work asking you to add them to the follow-up table, what do you do without a computer nearby? Doubao's mobile app supports cross-device interconnection, allowing you to directly link to the PC context, find the previous conversation record, send the documents, and say "Add these customer details to the follow-up table." Within minutes, the table is automatically updated. This ability to maintain context across devices is a key differentiator of Agent-type products compared to traditional spreadsheet tools.

Agent (Intelligent Agent) is fundamentally different from traditional chatbots or spreadsheet software. Traditional tools require users to operate step by step (e.g., manual copying, writing formulas), while an Agent has the ability to autonomously plan and execute multi-step tasks. It can understand natural language instructions, invoke built-in "skills" (such as table processing, document parsing, application generation), and automatically complete the entire workflow from information extraction and format conversion to logical calculation. Doubao Agent's "Table Skill" encapsulates spreadsheet operations into callable capability modules. Users don't need to care about the underlying implementation—just describe the goal. This paradigm lowers the barrier to tool usage, enabling non-technical users to efficiently handle complex data tasks.

Trick 2: Batch Extraction of Key Fields from Documents

Extracting specified fields from a large number of documents and structuring them is a typical need for HR, administrative, and operations roles. Take resume screening as an example: going through thousands of resumes one by one is obviously unrealistic. By dumping resumes to Doubao in batches and invoking the table skill, it can batch-extract key information such as name, education, years of work experience, desired position, and contact details, generating a candidate summary table.

The video also offers a practical tip: when generating the online table, upload the resume attachments as well—this makes subsequent screening and maintenance more convenient. This approach is worth adopting: manage "extraction results" and "original files" in the same place to avoid back-and-forth searching when tracing information.

Structured data refers to data organized according to a fixed schema, such as rows and columns in a table, making it easy for computers to store, query, and analyze. In contrast, unstructured data includes free text such as resumes, chat histories, and documents. Extracting key fields from unstructured data and converting them into structured tables is a core step in data preprocessing. Traditional methods rely on regular expressions or custom scripts, which have poor generalization ability and high maintenance costs. In contrast, LLM-based Agents can understand document semantics and automatically identify fields like name and education, accurately extracting even from diverse formats, greatly improving information processing efficiency.

Trick 3: Cleaning Dirty Data According to Rules

Collected data often has messy formats: inconsistent date formats, amounts sometimes with units and sometimes without, mixed duplicate rows and blank rows. Without cleaning, this data can't be used directly, but manual cleaning can be more exhausting than re-entering data.

The operation still involves invoking the table skill and telling Doubao the data cleaning rules in natural language. After entering the prompt, the messy table becomes neat and organized. The key is that "rule descriptions must be specific"—for example, unify date formats to one style, remove units from amounts and convert to numeric values, delete duplicate rows. The clearer the rules, the more controllable the cleaning results.

Trick 4: Complete Data Calculations with Natural Language

Calculating performance bonuses at the end of the month is a nightmare for formula-phobes: different payout amounts correspond to different sales target completion rates. Many people can't write the formula, and even when they do, it often errors due to misplaced parentheses or blank rows.

The video's approach is to give the sales data to Doubao, explain the performance payout rules clearly, and within two minutes you get the calculated completion rates and corresponding bonus amounts, automatically sorted from highest to lowest completion rate. The value of this trick is that it brings the threshold for "writing formulas" down to zero—users only need to explain the business rules clearly, and the Agent handles the calculation logic.

Trick 5: Generate Charts and Interactive Dashboards in One Go

After organizing and calculating data, you can't just present tables when reporting to your boss. The prompt structure given in the video is worth referencing: first describe the data's application scenario, then specify the metrics and chart types to display, and finally add style requirements. With this structure, charts such as sales trends, product sales rankings, sales percentage breakdowns, and regional sales comparisons can be generated in one go and directly placed into presentation slides.

Taking it further is generating an interactive visualization dashboard: give the prompt to Doubao and select the application generation skill. It produces a web page with four large core metric cards at the top, key conclusions attached, and a full set of charts below, supporting filtering and linkage by month and region. The most practical part is that you can generate a link and send it directly to your boss—they can open it on their phone. In the past, building such a dashboard required finding a front-end developer or outsourcing; now it's just one sentence.

Interactive visualization dashboards are dynamic data display interfaces that allow users to explore data through filtering, linkage, and other operations. Their technical implementation typically involves front-end frameworks (such as React, Vue) and chart libraries (such as ECharts, D3.js), which previously required professional developers to write code. Doubao Agent's "application generation skill" can automatically generate web applications containing HTML, CSS, and JavaScript based on natural language descriptions, and deploy them as shareable links. This enables business people to quickly build data dashboards without relying on technical teams, achieving data-driven decision support.

Trick 6: Structured Analysis Reports to Support Decision-Making

Data analysis is the part that best demonstrates professional expertise. Making charts is not the goal—discovering problems and finding reusable experience is. What causes sales fluctuations, what to do next, how conversion rates differ across channels—these are what the boss really cares about.

Operationally, first organize the data that supports the report content, don't miss important data sources, then describe according to the prompt structure: report background and purpose, what analysis content to include, report acceptance criteria, and specify any format or style requirements. Finally, select the long-form writing skill. The output is no longer scattered data but a structured analysis report that can guide business decisions. It's worth noting that AI-generated analysis conclusions still need to be judged in light of actual business conditions—they shouldn't be taken at face value.

Long-form writing skill is an Agent's specialized capability for complex content generation tasks. Unlike simple text completion, long-form writing requires the model to have macro-level structural planning, logical coherence maintenance, and factual consistency assurance. In data analysis reports, the model needs to organize sections, cite data, distill conclusions, and present them in professional language based on the background, purpose, and acceptance criteria provided by the user. This skill typically combines retrieval-augmented generation (RAG) or in-context learning techniques to ensure the report content is closely related to the input data. However, note that generated results still require human verification, especially for parts involving business judgment and causal inference.

The Capability Logic Behind the Six Tricks

Looking at these six tricks together, they correspond to a complete data processing chain: organization, extraction, cleaning, calculation, visualization, and analysis. Doubao Agent's core capabilities can be summarized in three points: first, natural language replaces formulas and scripts, lowering the barrier to tool usage; second, cross-device context interconnection makes continuous tasks possible in mobile scenarios; third, application generation capability turns dashboard building—which previously required development resources—into a one-sentence task.

From a tool selection perspective, this type of Agent is suitable for table tasks with clear rules and high repetitiveness, significantly compressing mechanical operation time. However, data quality judgment and business calibration of analysis conclusions still depend on the user's professional judgment. Treating AI as an execution amplifier rather than a decision-making replacement is the more prudent approach.

According to the video, new users of Doubao PC have a 30-day free trial package, and existing users can earn credits through check-ins and completing tasks. The video author also mentioned that the six types of prompts have been organized and placed in the comment section for readers who need them.

One-line Summary

Six Doubao Agent tricks handle the entire Excel workflow, replacing formulas and scripts with natural language.

Paragraph Summary

This article introduces six practical tricks for using Doubao Agent to handle Excel, covering the complete chain from data organization to analysis reports. The tricks include: one-click organization of cross-channel information into tables, batch extraction of key fields from documents, rule-based cleaning of dirty data, natural language data calculations, one-go generation of charts and interactive dashboards, and structured analysis reports to support decision-making. The article emphasizes that the Agent's core capabilities lie in natural language replacing formulas and scripts, cross-device context interconnection, and application generation, which can significantly lower the barrier to tool usage and compress mechanical operation time. It also reminds that AI-generated analysis conclusions still need to be judged in light of actual business conditions—AI is an execution amplifier, not a decision-making replacement.