Analyze Excel and CSV with Tukun.ai: Upload a file and talk to your data

Many data analysis tasks do not start in a database. They start with an Excel or CSV file.

Sales exports, order records, campaign reports, inventory lists, customer files, finance ledgers, survey results: these files move through chat apps, email attachments, and local folders every day. The data is already there, but making sense of it still takes work. Open the spreadsheet, inspect the columns, filter rows, build pivot tables, write formulas, create charts, then turn the findings into a report.

Tukun.ai is designed to make that workflow shorter.

In the Tukun.ai workbench, you can upload an Excel or CSV file, let the system understand its tables and fields, and then ask questions in plain language. The file becomes part of the current conversation context, so analysis can continue across follow-up questions instead of restarting from scratch.

Start from the + menu next to the input

The upload entry point sits beside the workbench input.

Click the + menu, choose “Upload file”, and select an Excel or CSV file from your computer. Once the upload finishes, the file is added to the current conversation, and later questions can use it as the analysis source.

You can ask:

Analyze the overall performance in this sales file.

Break down revenue, order count, and average order value by month.

Find unusual orders and explain why they look unusual.

Turn the findings into a short weekly business update.

You do not need to create a table first. You do not need to write SQL. You do not need to build pivot tables before asking the first question.

That is the difference between Tukun.ai and a simple file Q&A flow. The uploaded file is not just an attachment to read once. It becomes an analysis-ready data source inside the workbench.

After upload, AI understands tables and fields

Real business spreadsheets rarely have perfect column names.

You may see fields like:

dt
uid
cid
amt
gmv
is_refund
pay_time
channel

Someone close to the source system may know what those names mean. Everyone else needs time to decode them.

After upload, Tukun.ai reads the file structure and infers field types and field descriptions. It tries to identify:

• which fields are time fields
• which fields are amounts, counts, or ratios
• which fields are useful dimensions for grouping
• which fields look like order IDs, user IDs, or product IDs
• which fields may represent channels, regions, statuses, or categories
• what business questions the file is likely able to answer

This makes the next step more natural. You do not always need to name the exact column first. You can start with the business question.

For example:

Show me the revenue trend.

Instead of starting with:

Group pay_amount by pay_time.

If you know the business meaning of a field, you can still add it:

pay_amount is paid revenue and refund_amount is refunded revenue. Analyze net revenue over time.

This table and field understanding is the foundation of Tukun.ai as a Data Agent. It does not only translate questions into queries. It first turns a file into data context that can support ongoing analysis.

Not just one query, but a chain of analysis

Traditional ChatBI is often used to ask for one number:

What was revenue this month?

Count orders by channel.

That is useful, but many real business questions do not end after one query.

Suppose you upload a CSV export of marketing spend. What you really need may be:

Which channels spent the most?

Which channels had the lowest ROI?

For high-spend, low-ROI channels, is the issue conversion rate or order value?

Which channels should we keep investing in, and which should we reduce?

Write this as an operations review.

Tukun.ai is built for this kind of follow-up analysis. The file stays in the current conversation context, so you can continue from the previous result without uploading the file again or repeating the background.

That is one practical difference between a Data Agent and conventional ChatBI. ChatBI is often about querying metrics. A Data Agent is closer to completing an analysis task.

Example: analyzing a sales Excel file

Imagine you upload 2026_Q1_sales.xlsx, a spreadsheet with first-quarter sales orders.

You can start with:

Analyze this Q1 sales file and give me an overview first.

Tukun.ai will inspect the tables and fields, such as date, order, product, region, and amount fields, then return a first overview.

Then you can ask:

Break down revenue, order count, and average order value by month.

Then:

Find the products with the largest revenue decline and explain possible reasons.

Then:

In which regions did those products decline the most?

Finally:

Turn the analysis into a business review with key findings and recommended actions.

The whole flow stays in one workbench. You do not need to create multiple pivot tables in Excel or move data between tools just to keep the analysis going.

How is this different from using DeepSeek, OpenAI, or Claude directly?

Models such as DeepSeek, OpenAI, and Claude have made AI data analysis much more approachable. Uploading a file to an AI assistant and asking it to explain the data can already save time.

But real work needs more than model capability.

Data analysis needs stable context. It needs to know which file is being used. It needs field understanding. It needs follow-up questions. It often needs files, databases, and existing data sources to sit in the same workflow.

Tukun.ai focuses on that product layer:

• uploaded files become part of the current data context
• the system infers table structure and field descriptions
• users can keep asking follow-up questions about the same file
• Excel, CSV, databases, and existing data sources can be used in one workbench
• results can be turned into summaries, reviews, and recommendations

DeepSeek, OpenAI, and Claude represent progress in foundation models. Tukun.ai focuses on placing that capability inside a stable workflow for business data analysis.

How is this different from BI or ChatBI?

Traditional BI is good for fixed dashboards. When metrics are well defined, data pipelines are stable, and charts need to be reused, BI is the right tool.

But many everyday analysis questions are not fixed dashboards:

Why did revenue drop this month?

Which customers are worth following up with?

Is this channel still worth investing in?

Are there anomalies in this order file?

These questions often start from an Excel or CSV file, and the direction changes as the analysis develops.

Tukun.ai is closer to an AI data workbench. It supports the natural-language querying pattern people associate with ChatBI, but it is not limited to one-off metric lookup. It starts from the file, follows the business question, and helps shape the result into something that can be shared.

What Excel and CSV files fit well?

Tukun.ai works well with common structured files, including:

• sales order exports
• user behavior data
• advertising and campaign reports
• finance and expense records
• inventory and procurement files
• customer lists and CRM exports
• survey results
• support ticket data
• weekly and monthly operations reports

If the file has clear headers and structured rows, it is usually a good candidate for analysis.

If the fields are specific to your business, you can add context in the question:

This is our April channel performance file. cost is spend and revenue is sales revenue. Analyze ROI by channel.

Tukun.ai will combine your explanation with the inferred field information.

Why file analysis matters

When teams talk about AI data analysis, they often start with database connections. In daily work, though, files are still the most common entry point.

Files are easy to get. They are good for temporary analysis. They often carry business context. They do not require a long integration process. They are also friendlier for non-technical users.

That is why a useful Data Agent needs to handle Excel and CSV well, not just databases.

Tukun.ai puts file upload next to the input because the first step should be small: upload the file, ask the first question, then go deeper.

Tukun.ai’s product advantages

For Excel and CSV analysis, Tukun.ai is useful because:

• the path is short: upload from the + menu beside the input
• the system infers tables, field types, and field descriptions
• users can ask in business language instead of SQL
• the same file can support follow-up questions in the conversation
• results can become summaries, reviews, and recommended actions
• files, databases, and existing data sources share one workbench flow

That is where Tukun.ai differs from generic AI chat, traditional BI, and single-purpose ChatBI tools.

FAQ

Is Tukun.ai a ChatBI tool?

Tukun.ai supports the natural-language querying pattern of ChatBI, but it is broader. It is closer to a Data Agent workbench for Excel, CSV, databases, and existing data sources.

Do I need to configure fields after uploading Excel or CSV?

Usually no. Tukun.ai infers table structure, field types, and field descriptions. You can still add business meaning in your question when needed.

Can I use it without SQL?

Yes. The core interaction is natural language. You can ask questions such as “Which channel had the lowest ROI?”, “Which customers drove the most revenue?”, or “Find unusual orders.”

How does this relate to DeepSeek, OpenAI, and Claude?

DeepSeek, OpenAI, and Claude are foundation model capabilities. Tukun.ai focuses on turning those AI capabilities into a product workflow for uploading files, understanding fields, asking follow-up questions, and producing business conclusions.

Summary

Excel and CSV are still where many data tasks begin. In the past, we analyzed them by hand in Excel or imported them into a database before writing SQL. Tukun.ai offers a more direct path: upload the file, then talk to the data.

After upload, Tukun.ai understands the tables and fields and adds the file to the current conversation context. You can begin with a simple question, continue with follow-ups, find anomalies, break down causes, and turn the result into a usable summary.

If you have an Excel or CSV file on hand, start from the + menu in the Tukun.ai workbench. Upload the file, then ask the first business question.