Chatting with Your Agent
Published March 2, 2026 · Last updated March 5, 2026 · 5 min read
This guide covers how to work with your Obvious agent through chat — creating artifacts, editing content, analyzing data, and transforming information using natural conversation.
Creating Artifacts
The fastest way to create anything in Obvious is to ask for it.
Create a workbook to track our Q1 marketing campaigns with columns for campaign name, channel, budget, spend, and ROI
The agent creates the workbook, sets up the sheet with the right field types, and opens it for you. This works for every artifact type — documents, workbooks, presentations, and more. Be specific about what you need, and the agent handles the structure.
A few more examples of what you can ask:
Write a project brief for our website redesign. Include sections for objectives, timeline, stakeholders, and success metrics.
Create a sheet to track job applicants with name, email, role applied for, interview stage, and notes.
The agent picks the right artifact type based on your request. Structured data becomes a workbook. Long-form content becomes a document. You don't need to specify the format unless you have a preference.
Editing and Refining
Once an artifact exists, ask the agent to change it. Reference what you want edited, and the agent makes targeted updates without rewriting everything else.
Add a "Priority" column to the campaigns sheet with options: High, Medium, Low
Make the executive summary shorter — two paragraphs max
Change the date column to sort newest first
Edits happen in place. The agent modifies the existing artifact rather than creating a new one, so your links, views, and references stay intact. If you want a fresh copy instead, say so explicitly: "Create a new version of this document with..."
Analyzing Data
Drop a file into the chat or point the agent at an existing sheet, and ask questions about it. The agent reads the data, runs queries, and produces answers — often with charts or summary tables.
What are the top 5 campaigns by ROI?
Show me a breakdown of spend by channel for the last three months
Are there any duplicate entries in this sheet?
The agent uses SQL under the hood to query your sheets, so it handles filtering, grouping, aggregation, and joins across multiple sheets. You don't need to write SQL — just describe what you want to know.
For deeper analysis, the agent can create new sheets with the results, build charts, or write a document summarizing its findings. Ask for what you need:
Create a summary sheet showing total spend and average ROI by channel
Build a bar chart comparing Q1 budget vs. actual spend
Transforming Data
Data cleanup and transformation is where the agent saves the most time. Describe what needs to change, and the agent applies the transformation across your entire dataset.
Normalize all the company names — capitalize properly and remove trailing spaces
Split the "Full Name" column into "First Name" and "Last Name"
Convert all dates to YYYY-MM-DD format
Flag any rows where the email address looks invalid
The agent can also enrich data by pulling in information from external sources, merging sheets, deduplicating records, or applying formulas. Complex transformations that would take hours of manual work often take a single message.
Working with Files
Drag and drop files directly into the chat — CSVs, PDFs, images, Excel files, or any common format. The agent reads the content and makes it available for your conversation.
For CSVs and spreadsheets, the agent can import the data into a workbook, clean it up, and set appropriate field types. For PDFs and documents, the agent extracts the content and can summarize, analyze, or restructure it.
Import this CSV as a new sheet and fix any formatting issues
Summarize the key points from this PDF
Extract all the email addresses from this document into a sheet
Using @Mentions and /Commands
Two shortcuts make chatting more precise. Type @ to mention a specific artifact, file, or sheet by name — this tells the agent exactly which resource you're referring to, even if your project has dozens of artifacts. As you type, results are ranked by relevance: exact matches appear first, followed by names that start with what you've typed, then alphabetical order. You'll usually find what you're looking for in the first few results.
Type / to open the command menu, which gives you access to saved shortcuts — reusable prompts you've created for tasks you repeat often.
Tips for Better Results
Be specific about outcomes. "Analyze this data" is vague. "Show me the top 10 customers by revenue with their churn risk score" gives the agent a clear target.
Reference artifacts by name. When your project has multiple sheets or documents, mentioning the specific one avoids ambiguity. Use @mentions for precision.
Iterate naturally. If the first result isn't quite right, refine it. "Make the chart horizontal instead" or "Add a column for percentage change" — the agent builds on its previous work.
Use the right mode. Quick tasks work well in Fast mode. Complex analysis benefits from Deep Work or Analyst mode. Switch modes with the mode selector or Ctrl+1 through Ctrl+9.
Next Steps
-
Agent Modes — choosing the right mode for your task
-
Your First Workbook — a hands-on walkthrough
-
Keyboard Shortcuts — speed up your workflow