10 NEW Excel Copilot Features You Need to Know

Summary of Technological Concepts & Copilot Features (Top 10)

The video demonstrates 10 new Microsoft Excel Copilot capabilities focused on accelerating spreadsheet work, reducing manual formula-writing, and improving data transformation. The emphasis is on practical AI workflows where Copilot can:

• Generate formulas
• Classify data using context (without strict keys)
• Automate formatting and cleanup
• Merge data from multiple sources

---

1) Auto-filling Excel formulas (web Excel)

• Copilot can suggest and extend real Excel formulas as soon as you type (starting with the = sign).
• Example: it generates a formula for net income (gross income minus discount) and then fills it down the column automatically.
• If inputs change, results update automatically, confirming the formulas remain linked to Excel cells.
• Limitation: this specific auto-fill behavior works in Excel on the web; desktop users should rely on other Copilot features.

---

2) Natural-language generation of conditional logic (IF)

• Copilot can interpret conditional requirements in plain language.
• Example: commission rate depends on whether net income is above 10,000 (4%) or not (2%).
• The speaker notes Copilot correctly generates the logic you’d typically build with an IF statement.

---

3) AI-based classification of rows using context (no lookup keys needed)

• Copilot can classify transactions even when explicit category labels don’t exist—so traditional XLOOKUP/VLOOKUP won’t help.
• Example classifications:
  • Starbucks coffee → “food”
  • Uber ride → “transportation”
• Approach: prompt Copilot with instructions like “classify these transactions” and provide context.

---

4) Using Copilot with more external/operational data (e.g., stock prices)

• The video claims Copilot access has improved beyond only internal datasets.
• Example: ask Copilot to find stock prices for tickers, and it returns data for the corresponding tickers.
• Note: the speaker also highlights an occasional AI reliability issue—minor inconsistencies/mistakes, such as creating an unnecessary extra column.

---

5) Personalization settings for output formatting

• Copilot personalization can enforce consistent formatting based on user preference.
• Example: date formatting set to “day, month, year”—useful when importing files that use different conventions (e.g., US vs. Japan).
• Demonstration: ask Copilot to list dates in a given month; output follows the configured format.
• Downside: personalization is tied to the user account and may not transfer to others.

---

6) “Book rules” (rules stored with the workbook)

• Copilot supports book rules so formatting behavior can be shared consistently across users/files.
• Example rules:
  • Currency displayed in pounds sterling
  • Negative values shown in parentheses and highlighted red
  • Percentages shown with one decimal place
• Demonstration: generating a sample income statement shows formatting applied consistently.

---

7) Skills (automate repeated formatting/cleanup tasks)

• Skills let users reuse instruction sets without rewriting prompts each time.
• Workflow:
  • Press @ to select an existing skill (e.g., “clean and format”).
  • Create custom skills via OneDrive by editing a skills.md definition in a duplicated folder.
• Demonstrated outcomes:
  • Payment terms standardized (e.g., “net 30” casing/format)
  • Margins/formatting normalized
  • Faster dataset cleanup
• Tip: the speaker suggests using tools like Claude or ChatGPT to verify formatting correctness when needed.

---

8) Editing modes: Chat vs Planning vs Edit + change mapping

Copilot supports different modes for applying changes:

• Chat: explains/inspects data without modifying the sheet.
• Planning: proposes changes and applies them only after approval.
• Edit permission / direct edit: applies changes directly based on the request.
• Scheduling is also mentioned as an additional customization.

If using direct edit and you’re unsure what changed:

• Copilot provides “Show Changes” / change mapping
• This displays all modifications (including Copilot-made ones), with diffs and highlighted changed columns.

---

9) Switching model providers in Excel Copilot

• Users can choose underlying model providers instead of relying on automatic selection.
• Examples mentioned:
  • GPT 5.5 (OpenAI)
  • Claude 4.8 (Anthropic)
• Speaker guidance:
  • GPT models tend to be better for VBA/code
  • Claude is strong for financial modeling, error detection, and text analysis
• Overall recommendation: prioritize crafting a strong prompt/query over focusing solely on model selection.

---

10) Merging multiple files automatically via OneDrive (Excel + other formats)

• Copilot can read and merge data from multiple files without manual downloading.
• Example:
  • Merge regional sales files (Americas, EMEA, APAC) into one table using OneDrive.
  • Copilot adds a source file column indicating each row’s origin.
  • The demo claims Copilot recognizes both the files and their schema from OneDrive automatically.

Also stated to work beyond Excel:

• PDFs: Copilot can extract invoice details from multiple PDFs and add them as new rows in Excel via prompts like:
  • “add PDF invoices ending with these values… fill fields based on invoice data”
• The example shows three matching PDF invoices added as new records.

---

Main Speakers / Sources

• Primary speaker: the video’s host (name not explicitly provided in the subtitles; branded references “Kenji”)
• Sponsor/source mentioned: HubSpot
• Model providers mentioned:
  • OpenAI (GPT 5.5)
  • Anthropic (Claude 4.8)
• Other third-party tools mentioned: Beautiful.ai, ChatGPT, Claude