Using TRAE Work for Industry Research and Batch Office Tasks: A Software Blogger's Real Workflow Transformation Record

This article documents the entire process of how I used TRAE Work's Work mode to complete two daily tasks: an SSD market research project and a brand collaboration proposal. The focus is not on hyping the tool, but on sharing reusable commands and lessons learned from pitfalls.

Background: Who I Am and Why I Use TRAE Work

I run a public account that popularizes software tools, focusing on computer hardware and software utilities. My daily work can roughly be divided into two parts: content creation (topic selection, research, writing) and operational chores (organizing material libraries, managing publishing records, replying to comments).

These two types of work share a common characteristic—they don't involve coding, but require processing large amounts of text and data, and are highly repetitive.

My previous workflow looked like this:

• Industry Research: Open 10 browser tabs, visit manufacturer websites, e-commerce pages, and review articles one by one, copy parameters into Excel, and manually categorize and compare. An SSD research project would take most of a day.

• Material Organization: Export the material list from the public account backend, manually copy and paste it into a spreadsheet, categorize by topic, and tag. Organizing dozens of articles once would make my eyes blur.

Later, I tried using general AI conversation tools to help, but found problems: the context couldn't be remembered; when you asked it to organize a second batch of data, it had already forgotten the format of the first batch; the generated tables still had to be manually copied back into Excel, and the formatting often got messed up in between.

What attracted me to TRAE Work is its Work mode. Unlike pure conversational AI, it can directly manipulate files, execute code, and maintain workspace context. In other words, it doesn't just "help you write a paragraph of text," but can continuously advance tasks within a workspace.

Below are two real-world cases.

Case 1: SSD Market Research—From Information Gathering to a Structured Report

Real Pain Points

Doing hardware popularization, research is routine. Recently, NVMe SSD market prices have been fluctuating quite a bit. I wanted to write a "Mid-2026 SSD Buying Guide" and needed to first understand the parameters and price ranges of mainstream models from various brands.

The old process was roughly like this:

1. Open JD.com and Tmall, search for "NVMe SSD," sort by sales volume, and click on product pages one by one to view parameters.
2. Copy the brand, model, capacity, interface, read/write speed, TBW, price, and warranty period into an Excel sheet.
3. Search for a few review articles to supplement actual test data and word-of-mouth information.
4. Manually organize into a comparison table and write a summary analysis.

The problems with this process are obvious:

• Scattered Information: Parameters are spread across different pages, making manual copy-pasting extremely inefficient and prone to omissions.
• Format Inconsistency: Price labeling methods differ across platforms; some show launch price, some show promotional price, some show the original price with a strikethrough.
• Easily Outdated: By the time you finish organizing, prices may have changed again.
• Hard to Reuse: Next time you do similar research, you have to start from scratch again.

The entire research process usually took 4-5 hours, of which truly valuable information analysis might only account for 1 hour, with the rest being mechanical information transfer.

Practical Process

Step 1: Describe Research Requirements in Work Mode

Open TRAE Work, switch to Work mode, and enter in the conversation:

I want to do a market research on NVMe SSDs for mid-2026, targeting readers from computer novices to advanced users.
Please help me gather the following information:
1. The flagship and mid-range NVMe SSD models currently on the market from mainstream brands (Samsung, Western Digital, Zhitai, Kioxia, Hikvision, etc.)
2. Key parameters for each model: capacity, PCIe generation, NAND flash type, controller, sequential read/write speed, random read/write IOPS, TBW, DRAM presence
3. Approximate price ranges on current e-commerce platforms (JD.com/Tmall public pages)
4. Key word-of-mouth points for each model (common evaluations based on public reviews and user feedback)

Output format: First, a structured comparison table, then a market trend analysis of about 300 words.

A key point here: the requirement description needs to be structured. Don't say "help me research SSDs"; provide a specific brand scope, parameter fields, and output format. TRAE Work's understanding and execution of structured instructions are significantly better than vague ones.

Step 2: Information Gathering and Initial Organization

After receiving the instructions, TRAE Work uses its search capabilities to gather public information. My observations during this process:

• It searches in batches, first by brand, then supplements parameters for specific models.
• The searched information is organized into the workspace in real-time, not all at once after searching everything.
• It proactively confirms midway: "Zhitai TiPlus7100 and TiPro7000 are both on sale, should both be included in the comparison?"—this kind of fork-point confirmation is very important to avoid going off track.

A pitfall encountered here: the first time I asked it to search, it included some discontinued models. I corrected it once: "Only look at models currently on sale, ignore discontinued ones." After that, it was accurate.

Step 3: Generate a Structured Comparison Table

After information gathering was complete, TRAE Work directly generated a Markdown table in the workspace. I asked it to export as a CSV for subsequent processing:

Organize the above comparison table into CSV format and save it to the workspace, filename ssd_comparison.csv,
Field order: Brand, Model, Capacity, PCIe Generation, NAND Type, Controller, Sequential Read, Sequential Write, Random Read IOPS, Random Write IOPS, TBW, DRAM, Price Range, Warranty

The key to this step is specifying the filename and field order. Without specifying, it uses its own default format, which might not match the subsequent processing flow.

Step 4: Generate Trend Analysis

After getting the comparison table, continue in the same conversation round to ask it to analyze:

Based on the above comparison table, please help me analyze the following three points:
1. Main technology trends in the NVMe SSD market for mid-2026 (PCIe 5.0 adoption status, changes in NAND flash solutions, etc.)
2. Recommended choices for each price segment (below 500 yuan, 500-800 yuan, above 800 yuan)
3. The most common pitfalls for average users (e.g., installing a PCIe 4.0 drive on a PCIe 3.0 motherboard halves the speed)

The analysis in this step is based on the previously gathered data, not generated out of thin air. TRAE Work maintains context within the same workspace, so it "knows" which models and parameters were in the previous table.

Final Deliverables

Final deliverables:

1. SSD Comparison Table (CSV format, 15 models, 14 parameter fields)—directly usable for citation when writing articles.
2. Market Trend Analysis (about 800 words, covering three dimensions: technology trends, price segment recommendations, and pitfall avoidance tips)—serves as the research base for the article's first draft.

Efficiency comparison:

Step	Old Process	TRAE Work	Difference
Information Gathering	~2 hours	~20 minutes (search + organize)	-83%
Data Tabulation	~1.5 hours	~5 minutes (auto-generated)	-94%
Trend Analysis Writing	~1 hour	~15 minutes (data-based generation + manual correction)	-75%
Total	~4.5 hours	~40 minutes	-85%

It should be noted that the analysis generated by TRAE Work is not a final draft ready for publication. Price data needs secondary verification (e-commerce prices change frequently), and some technical judgments require correction based on my professional knowledge. But the completion level of the draft is about 70-80%, and the remaining refinement is much faster than starting from scratch.

Case 2: Brand Content Collaboration Proposal—From a Brief to a Complete Plan

Real Pain Points

Once a public account reaches a certain scale, brand collaborations become unavoidable. The most common scenario is this: a hardware or software brand finds me through business channels and sends a collaboration brief—usually a product introduction + collaboration intent, which might be just two or three paragraphs.

For example, a company that makes system monitoring tools contacted me some time ago. The brief was roughly: "We have launched a new hardware temperature monitoring software that supports real-time CPU/GPU/SSD temperature monitoring and alerts. We want to do a round of content promotion on tech public accounts. Please help plan it."

Just these two sentences, and I had to turn it into a content collaboration proposal that the brand could approve. The old process was like this:

1. First, understand the product—visit the official website to see features, download and try it, search for user reviews.
2. Think of a few topic angles; which entry point gives readers a sense of gain without being a hard sell.
3. Outline the article structure: how to start the hook, what to write in the middle, how to end.
4. Define evaluation dimensions: what metrics to test, what competitors to compare against, what testing methods to use.
5. Schedule the publishing rhythm: how many articles, when to publish, what the different focus of each article is.
6. Organize into a Word document, format it, and send the email.

This whole process, if it went smoothly, took 3-4 hours. If it was an unfamiliar product, I had to spend time researching first, which took even longer.

The problems were:

• Starting from scratch every time: Different brands, different products, the proposal structure had to be re-thought, with no stable template to apply.
• Easy to miss dimensions: When manually listing an outline, I often found halfway through writing that I had missed an evaluation dimension and had to go back and add it.
• Inconsistent structure: Detailed when in a good mood, rough when in a hurry; the proposal quality depended on my state of mind.
• Repeated revisions: After the brand reviewed it and proposed changes, revising one version took another half hour.

Practical Process

Step 1: Throw the Brand Brief and Product Materials into the Workspace

Put the brand's requirement brief text, the product's official website link, and a few notes I jotted down while trying the product all into TRAE Work's workspace. Then enter:

There is a brand collaboration requirement brief and product materials in the workspace:
- Requirement Brief: The brand wants to do a round of public account content promotion for a hardware temperature monitoring software product.
- Product Official Website Link: [https://www.librehardwaremonitor.cn/]
- My Trial Notes: Recorded a few usage impressions (attached in the workspace notes.txt)

Based on these materials, please help me organize a complete content collaboration proposal, including the following modules:
1. Product Overview: Core features, target users, differentiating selling points (no more than 3)
2. Topic Angles: 3 different entry angles, each explaining which readers it suits and why this angle is attractive.
3. Article Structure Outline: The skeleton of each article (opening hook → core content → closing guidance), with 2-3 sentences describing what to write for each part.
4. Evaluation Dimensions: Suggest which metrics to test, which competitors to compare against, and how to set up the testing environment.
5. Publishing Rhythm: Total number of articles, interval between each, and how to allocate the focus of each article.
6. Expected Results: Target audience and expected readership range for each article.

Output in Markdown format, with each module separated by a second-level heading.

The key point of this step is to define the module structure of the proposal in advance. You tell it what parts the proposal should contain, and it will fill in block by block without missing anything. If you vaguely say "help me write a collaboration proposal," it might only give you a paragraph of overview text, which is not directly usable.

Step 2: Review the First Draft and Supplement Product Understanding

TRAE Work generated the first draft in about 3 minutes. I quickly scanned it and found that the topic angles were written too generically—the three angles were "Feature Introduction," "Usage Tutorial," and "Comparison Review," which were too conventional and not in my style.

My public account's tone is "hardcore but accessible," and my readers lean towards advanced users, so I added supplementary instructions:

The topic angles need adjustment. My public account style is "hardcore but accessible," and readers are more advanced users.
Change the three angles to better fit this tone:
- Angle 1: Start from the pain point of "Why is your computer always overheating?" and discuss the necessity of temperature monitoring.
- Angle 2: Deeply dissect the monitoring principle of this software (which sensor data it reads, the difference from HWMonitor).
- Angle 3: Practical scenario—use this software to troubleshoot a real computer overheating problem.

For each angle, supplement: which type of reader it suits, the estimated read completion rate, and the synergistic relationship with existing content.

This step is crucial: the first draft gives you the framework, and you inject style and judgment. TRAE Work doesn't know your public account's tone unless you tell it. But the framework it provides saves you the struggle of "where to start writing."

Step 3: Refine Evaluation Dimensions

For the evaluation dimensions part of the proposal, the first draft listed three directions: "Feature Completeness," "Ease of Use," and "Performance Overhead." I asked it to refine them:

Further refine the evaluation dimensions, providing specific testing methods for each dimension:
1. Feature Completeness: List the core feature checklist that similar tools all have, and mark item by item whether this product supports them.
2. Sensor Data Accuracy: Do a horizontal comparison with readings from HWMonitor and Core Temp (same machine, same moment).
3. Resource Usage: CPU/Memory usage at idle and under full load, recorded with Task Manager screenshots.
4. Alert Function: Set thresholds and simulate an over-temperature scenario to test if the alert is timely and what methods are available (pop-up/sound/email).

The testing method for each dimension should be specific to "what tool to use for testing, how to record data, and what the comparison benchmark is."

After refinement, the evaluation dimensions changed from three vague directions to four executable testing plans. When a brand sees a proposal at this level of detail, they basically know you are serious about content creation, not just casually writing a sponsored post.

Step 4: Export the Proposal Document

After confirming the proposal content, ask it to export as a structured document:

Organize the complete proposal above into a document and save it to the workspace, filename brand_collab_plan_monitor_tool.md.
Format requirements: Each module uses a second-level heading, topic angles and evaluation dimensions are presented in tables, and the publishing rhythm uses a timeline list.
Add a line at the end: "Proposal Author: [Blogger Name] | Date: 2026-08-13"

Final Deliverables

Final deliverables:

1. Content Collaboration Proposal (Markdown document, 6 modules, ~2500 words)—includes product overview, 3 topic angles (with tone matching explanations), 4 article structure outlines, a 4-dimensional evaluation plan (with specific testing methods), publishing schedule, and expected results.

1. Topic Angle Comparison Table (3 angles × 3 dimensions: suitable readers/expected completion rate/content synergy)—convenient for quick comparison and selection when discussing with the brand.

1. Evaluation Dimension Execution Checklist (4 dimensions × 3 fields: testing method/tool/comparison benchmark)—directly usable as a checklist during evaluation execution.

Efficiency comparison:

Step	Old Process	TRAE Work	Difference
Product Understanding + Material Organization	~1 hour	~15 minutes (search + organize into workspace)	-75%
Proposal Framework Setup	~1 hour	~3 minutes (auto-generate first draft)	-95%
Topic Angle + Outline Refinement	~1 hour	~20 minutes (supplementary instructions + manual adjustment)	-67%
Evaluation Plan Refinement	~30 minutes	~10 minutes (instruction refinement + confirmation)	-67%
Document Organization and Formatting	~30 minutes	~2 minutes (auto-export)	-93%
Total	~4 hours	~50 minutes	-79%

Key experience difference: The most painful part of the old process was "staring at a blank document not knowing where to start writing." TRAE Work directly gives you a framework that is 70% complete, allowing you to spend your energy on injecting tone and refining details, rather than building the framework from scratch.

Reusable Experience

This part is what I find most valuable. The following commands and methods have all been verified by me in practice and can be directly applied when encountering similar problems.

1. General Command Template for Industry Research

Replace the [Keyword] in the following text with the topic you want to research; the structure is basically universal:

I want to do a market research on [Keyword], targeting readers who are [Target Audience].
Please help me gather the following information:
1. Mainstream brands/product models on the market
2. Key parameters for each model (list the fields you care about)
3. Approximate price ranges on current e-commerce platforms
4. Key word-of-mouth points for each model (common evaluations based on public reviews and user feedback)

Output format: First, a structured comparison table, then a market trend analysis of about 300 words.

Key Points:

• Give a brand scope, otherwise it might search a bunch of obscure brands to make up the numbers.
• Clearly specify parameter fields so it knows what to fill in the table.
• For "word-of-mouth points," emphasize "common evaluations based on public reviews and user feedback," otherwise it might fabricate user comments.

2. General Command Template for Proposal Organization

When you receive a requirement brief or collaboration intent and need to turn it into a structured proposal:

The workspace contains the following materials: [List material files/content]
Based on these materials, please help me organize a complete [Proposal Type] proposal, including the following modules:
1. [Module A Name]: [What Module A should contain]
2. [Module B Name]: [What Module B should contain]
3. [Module C Name]: [What Module C should contain]
...

Output in Markdown format, with each module separated by a second-level heading.

Key Points:

• The module structure must be defined by you. This is the most important rule—you tell it what sections the proposal should have, and it fills them in block by block, without omissions or repetitions. If you don't define the modules, it might give you a long, rambling paragraph.
• Throw all relevant materials into the workspace first (brief text, product links, your own notes). It generates based on the workspace context, which is much more accurate than describing from memory.
• The first thing to do after output is to review the framework logic. If it's wrong, overturn it and start over; don't patch and mend on a wrong framework.

3. Command for Proposal Refinement and Iteration

After the first draft of the proposal is out, to deeply refine a specific module:

The [Module Name] part of the proposal needs further refinement:
1. Expand each point to an executable level—specific to "what tool/method/comparison benchmark to use."
2. Supplement each point with a one-sentence reason for "why do it this way."
3. If there are industry common standards or benchmarks, note them.

Keep the original structure unchanged, only expand the depth of content.

Key Points:

• Add "Keep the original structure unchanged"—otherwise it might reorganize the entire proposal, disrupting the parts you've already confirmed.
• "Specific to an executable level" is the keyword, forcing it to give steps that can be followed, rather than staying at the level of nonsense like "recommend a comprehensive evaluation."
• Only refine one module at a time; execute multiple modules in separate sessions.

4. Pitfall Avoidance Checklist

Pitfalls I've actually stepped into during use, listed here to save you from stepping into them again:

Pitfall 1: Vague Commands Lead to Divergent Results

• Negative example: "Help me look at the SSD market" → It might give you a popular science article instead of a parameter comparison table.
• Positive example: Give brand scope, parameter fields, output format.

Pitfall 2: Not Specifying Output Format

• TRAE Work defaults to Markdown output. If you need subsequent processing in Excel, explicitly say "Output as a CSV file and save to the workspace."
• If you need to use it directly, say "Output as an Excel file," and it will use code to generate an .xlsx file.

Pitfall 3: Asking It to Do Too Much at Once

• Throwing "research + write article + format" at it all at once results in each task being half-done.
• Break it down into steps: first research the data, confirm the data is correct, then ask it to write an analysis based on the data, and finally polish it.
• The context retention capability of Work mode is an advantage, but when the task is too complex, the context can also become "diluted."

Pitfall 4: Not Verifying Data Accuracy

• The prices and parameters searched by TRAE Work may be outdated or wrong.
• Key data (especially prices) must be double-checked.
• The content it generates is a "high-completion draft," not an "inspection-free final draft."

Pitfall 5: Casual Naming of Workspace Files

• Give the files it generates meaningful names, otherwise the workspace will be full of indistinguishable files like output.csv and result.xlsx.
• Naming suggestion: [Topic]_[Type]_[Date], e.g., ssd_comparison_202608.csv.

5. A Universal Tip

The essence of TRAE Work's Work mode is an AI workspace that can manipulate files and execute code. The core of using it well is not about writing exquisite prompts, but:

Break down large tasks into verifiable small steps, and confirm the results of each step before moving forward.

This is the same logic as writing code—you wouldn't write the logic of an entire system in one function, nor would you deploy without testing. Collaborating with AI is the same: break down tasks, verify outputs, then advance.

Final Words

The two scenarios recorded in this article are essentially "information transfer and format organization" type tasks. The characteristic of this type of work is: it doesn't require creativity, but requires accuracy and consistency. This is precisely the area where AI excels and can save you the most time.

TRAE Work's Work mode has several tangible advantages when handling such tasks:

• Persistent Workspace: Unlike conversational AI that forgets after chatting, it can continuously advance within one workspace.
• File Manipulation: Directly generates CSV, Excel, no need for manual copy-pasting.
• Code Execution: Operations like data cleaning and batch replacement are orders of magnitude faster when run with code than manual changes.

But it also has boundaries. It won't judge for you whether "the angle of this SSD buying guide is right," nor will it decide for you "what topic to choose for the next public account issue." Strategic decisions are your job; execution and transfer are its job. Only by separating these two things can efficiency truly be improved.

The above is my real record of transforming my daily workflow with TRAE Work. The commands and pitfall avoidance experiences can all be directly taken and used. If you have better ways to play, welcome to communicate in the comments section.