One Continuous Session Across Excel and PowerPoint
Claude can now work across your open Excel and PowerPoint files in a single continuous session — reading live cell values, building out analysis, updating slides, and carrying full context across both apps.
This lesson is original educational writing based on this video by Anthropic (published March 12, 2026). All credit for the original content goes to the creators.
1. What a cross-app session actually means
When most people first use Claude inside Microsoft Office, they use it one application at a time. They open Excel, ask Claude to explain a formula, close the sidebar, switch to PowerPoint, open the sidebar again, and start a new conversation with no memory of what just happened in Excel. Each application feels like a fresh start. The conversation history does not carry over. Any analysis Claude produced in Excel has to be re-explained or re-pasted before PowerPoint work can begin.
A cross-app session changes this entirely. When both Excel and PowerPoint are open in the same Microsoft 365 environment, Claude can hold a single continuous conversation that spans both files. You can ask Claude to analyze a data range in Excel, have it summarize the findings, then — without starting a new chat or copying anything — ask it to write those findings into a specific slide in PowerPoint. Claude remembers what it just read in Excel. The context carries. The analysis you did in one app is immediately available for the work you do in the other.
This is not a small productivity improvement. It represents a fundamental change in how AI assistance works inside a document suite. The capability mirrors how a skilled analyst actually works: they hold the data and the presentation in mind simultaneously. They do not compartmentalize. They move fluidly between understanding the numbers and communicating them, and the two activities inform each other throughout the session.
2. How live cell values feed into the session
The most technically important aspect of a cross-app session is that Claude reads live values from your files. It does not work from a snapshot taken when the session started. When you ask Claude to read a cell mid-session, it reads what is in that cell right now — the current computed value, reflecting any changes you or a collaborator made since the session opened.
This matters because real workflows are not linear. You might open a session, have Claude read some Excel figures, realize one of the inputs looks wrong, correct it in the spreadsheet, and then ask Claude to continue building the PowerPoint slide. In a snapshot-based system, Claude would still be working with the stale pre-correction numbers. In a live-reading system, it picks up the corrected value automatically.
What “live” means in practice
When Claude reads a cell value, it is reading the display value — the number Excel is currently showing in that cell, after all formulas have been calculated with current inputs. If you change an assumption cell and the downstream value updates via formula, Claude’s next read of the downstream cell will return the updated value. There is no cache to clear, no manual refresh to trigger. The read is always current.
This also means that Claude’s output in PowerPoint reflects the actual state of your Excel workbook at the moment the slide was written, not some earlier state. If you ask Claude to update a slide table with the revenue numbers from your model, and then discover you had the wrong exchange rate in your assumptions and correct it, you can ask Claude to re-run the slide update and the new values will be correct.
Context persistence across apps
Beyond live data, the other key property of a cross-app session is that Claude maintains the full conversation history as you switch between apps. This means:
- If Claude wrote a summary of your Excel data, it remembers writing that summary when you later ask it to expand on a point in PowerPoint.
- If you told Claude in the Excel portion of the session that you are building a deck for a CFO audience, it will maintain that audience context when drafting slide language in PowerPoint.
- If Claude identified a trend in your Excel analysis — “Q3 was the outlier quarter” — you can reference that finding by name in the PowerPoint portion without restating it.
The conversation is the shared context. Everything Claude learns about your data, your audience, and your goals in one app carries directly into its work in the other.
3. A complete workflow: from analysis to presentation
The practical value of a cross-app session becomes clearest when you trace a complete workflow from data to deck. Here is a concrete example of how the session flows.
Phase 1 — Excel analysis
You open your quarterly revenue workbook and your board-update presentation in the same Microsoft 365 session. You start the Claude conversation in Excel:
“I’m building a board update for our Q1 2026 performance. My audience is the board of directors — non-technical, focused on growth, margin, and guidance. Read the Revenue tab and give me the three most significant findings from Q1, with specific cell references for each finding.”
Claude reads the Revenue tab and responds with three findings — let’s say: Q1 revenue came in at $14.2M (cell D30), 8% above plan; gross margin expanded by 2.1 percentage points versus Q4 (cells E12 and F12); and the enterprise segment grew 31% year-over-year (cell H22) but the SMB segment declined 4% (cell H23).
You have not copied anything. You have not opened a second window. Claude has just read your actual numbers and produced an analysis oriented to your stated board audience.
Phase 2 — The pivot to PowerPoint
Without ending the conversation, you switch to PowerPoint. Your open presentation already has a slide titled “Q1 2026 Highlights.” You continue the same conversation:
“On slide 3, titled ‘Q1 2026 Highlights’, replace the placeholder bullets with the three Q1 findings you just identified. Write each bullet as a declarative one-line finding, not as a category label. Then write speaker notes for the slide that add context for the board discussion — specifically, what should the CEO emphasize about the SMB decline?”
Claude writes three declarative bullets into slide 3 using the findings from its Excel read. It also writes speaker notes that address the SMB decline with appropriate framing for a board discussion. The cell values are correct because Claude read them live. The language is appropriate because Claude remembered the audience context from the first prompt.
Phase 3 — Iterating across both apps
The session can continue to bounce between apps as needed. You might go back to Excel:
“Is the SMB decline primarily a volume issue or a price issue? Check the ASP and unit volume columns on the SMB tab.”
Claude reads the additional cells and returns the answer. You take that finding back to PowerPoint:
“Update the speaker note on slide 3 to specify that the SMB decline was volume-driven — ASP held flat but unit volume fell 9%. Make the language direct, not hedged.”
This back-and-forth — analysis in Excel, presentation update in PowerPoint, follow-up question in Excel, refined language in PowerPoint — is the natural rhythm of a cross-app session. Each turn builds on the prior turns. The context is cumulative.
4. Cross-app sessions versus copy-paste workflows
The obvious question is: why does this matter? You could always do this with copy-paste. Read the Excel numbers, copy them, switch to PowerPoint, paste them. Why is a cross-app session fundamentally different?
The first difference is accuracy. When you copy-paste, you introduce a manual transcription step. You are reading a number with your eyes, forming a mental representation of it, and then typing or pasting it somewhere else. This step is where errors enter decks. A number that says $14.2M becomes $14.2B when someone is moving fast. A percentage that is 8.3% becomes 83% when the decimal is missed. Claude reading a cell value directly eliminates this class of error entirely. The number in the slide is whatever is in the cell. No human hand touched the transcription.
The second difference is speed. A cross-app session can read and write a twenty-slide data-driven deck in the time it would take to manually copy-paste the numbers for five slides. The time saving scales with deck size and with how much analysis (not just raw numbers) needs to go into the slides.
The third difference is repeatability. With copy-paste, there is no record of which numbers came from where. If someone asks why slide 7 shows $4.8M rather than $5.1M, you have to trace back through your copy history or reopen the Excel file and check. With a cross-app session, the conversation history is the provenance record. Every number Claude wrote into a slide was explicitly read from a named cell in a named workbook, and that reference is in the chat log.
The fourth difference is composability. A cross-app session is not just about moving numbers. It is about moving analysis. When Claude writes a slide, it can simultaneously read from the workbook, apply business judgment to what it reads, and write language shaped by the audience context that was established earlier in the session. Copy-paste moves raw values. A cross-app session moves interpreted, audience-aware findings.
Check your understanding
4 questions · your answers are saved in this browser only
-
1. What is the defining characteristic of a cross-app Claude session?
-
2. You change an Excel assumption mid-session after Claude has already read some values. What happens when Claude next reads a downstream cell?
-
3. Which of these is NOT a key advantage of a cross-app session over a copy-paste workflow?
-
4. You want Claude to write a slide about Q3 findings, using a board-audience tone you established at the start of the session in Excel. What do you need to do before writing the PowerPoint portion of the prompt?
Build it yourself
Follow these exact steps to reproduce it yourself · estimated time: ~20 min
Prerequisites
- A Microsoft 365 subscription with Claude enabled in both Excel and PowerPoint
- An Excel workbook with at least one data table you want to present
- A PowerPoint presentation (existing or new) for the data you plan to analyze
Step 1 — Open both files in the same session
Open your Excel workbook and your PowerPoint presentation. Both files must be open and active in the same Microsoft 365 session — not in separate windows on different machines or in separate browser tabs if you are using Microsoft 365 on the web.
Step 2 — Set the context in the first prompt
Start the Claude conversation in Excel. Before asking Claude to read anything, establish the context that will carry through the entire session:
I'm building a [deck type, e.g. "board update" / "client readout" / "team review"]
for [audience description, e.g. "the board of directors — non-technical, focused
on growth and margin"]. My workbook contains [brief description of what's in the workbook].
I'll be working in both Excel and PowerPoint in this session.This framing prompt costs you thirty seconds and pays dividends throughout the session because Claude will orient all its output — including slide language — to the audience and purpose you stated here.
Step 3 — Do the Excel analysis
Ask Claude to read the relevant data and produce the analysis you need:
Read [tab name or named range] and give me the [N] most significant findings
for my [audience] audience. Cite the specific cell reference and value for each finding.Review Claude’s findings. If any finding looks wrong, ask Claude to re-read the specific cell to verify. Correct any wrong values in the workbook before proceeding — since Claude reads live values, correcting the workbook and then re-asking will return the corrected result.
Step 4 — Switch to PowerPoint without ending the conversation
Keep the Claude sidebar open (or available) and switch to PowerPoint. Navigate to the slide you want to populate. In the same Claude conversation, write a slide-writing prompt that references the findings Claude already produced:
On slide [N], titled '[Slide Title]', write the findings from your Excel analysis
as [N] declarative bullet points — one finding per bullet, stated as a fact not a
category label. Then write speaker notes that add context for the [audience]
discussion of these findings.Step 5 — Iterate across apps as needed
Continue the session as needed. If a finding needs clarification, switch back to Excel and ask Claude to dig deeper. If the speaker notes need to reflect additional context from the data, go back to Excel for it, then update the notes in PowerPoint.
Document the prompts that produced good results. Save them alongside your workbook and presentation. The next time you build this deck with updated data, re-running the saved prompts in the same session structure will reproduce the output with current numbers.
Expected result
A PowerPoint presentation with accurate, live-data-backed content produced in a single session that spanned both apps — no copy-paste, no transcription risk, and a conversation history that serves as a provenance record for every number in the deck.
Where to go next
- For a deeper look at what Claude can do inside Excel — including formula explanations, dependency tracing, and safe assumption updates — see Getting Started with Claude in Excel.
- For the full three-part PowerPoint workflow and how to use cross-app data to build a data-driven deck end to end, see Working Smarter with Claude in PowerPoint.
- For context on the broader Microsoft partnership that enables these integrations, see Claude on Microsoft Foundry.