Athena Intelligence makes every file multiplayer for people and AI agents. Teammates and Athena's agents edit and comment in the same sheet, document or deck at once, including a deck the CEO downloads to a laptop, which stays live through the Office add-in. Every change is attributed, and one agent session's changes can be rolled back on their own.
Shared workspaces for people and AI agents are getting a lot of attention. The question that decides whether one works is where the work lives. Athena's answer is the files a company already runs on: the spreadsheet model, the board deck and the memo, in your template. Making those files agent native means an agent is one of the people in the file.
This dispatch covers what multiplayer looks like when an agent is in the room, one board deck followed from first draft to a copy on a laptop, and the controls that let many agents work in the same file without anyone losing track of who did what. Figure 1 shows the whole idea.
Contents
01What problem does multiplayer for AI agents solve?
Multiplayer removes the extra copies. When an agent builds a deck in its own file, or a teammate edits a local version, the team ends up with several files and a question about which one is current. Putting everyone in one file, agents included, removes the question.
Those files are where the work already lives. The most widely used office suite passed 450 million paid commercial seats in January 2026. In finance, which produces much of the reporting that executives and boards read, the Association for Financial Professionals found in its 2025 benchmarking survey that 96% of FP&A professionals use spreadsheets for planning and 93% for reporting on a daily or weekly basis. Gartner reports that 51% of finance leaders still build a detailed bottom-up plan in spreadsheets and then load it into planning software, and workplace telemetry from the same vendor shows that edits to slide decks spike 122% in the final 10 minutes before a meeting. Decks, models and memos are where executives and enterprises do the work, so that is where agents have to be.
The version question is just as common. In a 2019 survey of 1,500 office workers by Vanson Bourne for M-Files, 96% said they have difficulty finding the most recent version of a document or file. The 2025 Work Trend Index describes agents joining teams as digital colleagues that take on specific tasks at human direction. A digital colleague that works in a separate copy brings the version question back at machine speed. Figure 2 sets the two ways of working side by side.
02What does it look like when an agent is in the file?
It looks like a teammate. Athena has a name and a cursor in the file, takes requests in comment threads, edits the slide, cell or paragraph in place, and replies with its source. Figure 3 shows the thread from the demo.
In the demo, Priya Shah writes "@Athena can you compare Q3 to Q2 on this chart?" Athena adds a lighter Q2 bar next to each region's Q3 bar and replies, "Done. Q2 is now the lighter bar next to each region." Morgan Lee asks where the Q2 numbers come from, and Athena replies, "Q2_Model.xlsx, sheet Model, row 2." The figure on the slide carries a citation back to that cell.
People see each other too. Every participant has a name, a color and a cursor, and avatars in the header show who is in the file. The same approach works in Athena sheets, where many subagents can write to one sheet at once without conflicts. The Studios page describes the document, deck and sheet editors.
03How does it work? One deck, start to finish
The demo follows one board deck from an agent's first draft to a synced copy on a laptop. Every step happens in a file that someone on the team already uses. The video shows the whole run, and the steps below follow it.
- An Athena agent builds the Q3 board deck. It fills in revenue, gross margin, opex and net retention and draws the regional chart, and each figure carries a Cited chip.
- Two teammates, Morgan and Priya in the demo, join while the deck is still building. Their names and cursors appear on the slide, and their avatars appear in the header next to a Synced chip.
- Priya comments on the chart and asks Athena to compare Q3 to Q2. Athena adds the Q2 bars and replies in the thread. Morgan asks where the Q2 numbers come from, and Athena answers with the model file, the sheet and the row.
- The spreadsheet model and the memo fill with people and agents too.
- The workspace view shows a wall of live files with cursors moving across every one.
- One agent session gets it wrong. A session called Refresh Q3 numbers, run by Athena for Jordan Kim in the demo, changes Opex to 16.2M. Dana Ruiz opens the Activity panel, finds that session and clicks Rollback. The dialog says it will undo 2 edits from newest to oldest, Opex returns to 10.7M, and a chip reads Rolled back 2 edits. Everyone else's edits stay.
- The CEO, Sam Ortiz in the demo, downloads the deck. That copy is now a local file on his laptop, and the live animation stops for it.
- He opens it on his laptop. The Athena add-in reads the deck link, joins the deck's room and reports Live sync on, linked on another device. The other files start moving again.
- Edits from the team and the agents land in his deck, in the company template. Morgan adds FINAL DRAFT to the title, and Athena updates a chart value.
Table 1 puts the manual version and the multiplayer version side by side.
| Task | The manual way | With multiplayer Athena |
|---|---|---|
| Getting a teammate's edits | Wait for their copy by email and merge the changes by hand | Teammates edit the same file, each with a name and a cursor |
| Asking for a change | Message someone and wait for a new version | Mention Athena in a comment, and Athena edits in place and replies |
| Checking a number | Hunt for the source file and the cell | Athena answers with the file, sheet and row, cited on the slide |
| Including a file on a laptop | Email it back and forth | Download it, and the Office add-in keeps that copy live with everyone |
| Keeping the template | Reapply the formatting after each merge | The file stays in its original format, in your template |
| Undoing one author's work | Work out what changed and redo it | Roll back one agent session and keep everyone else's edits |
04Does it work in sheets and documents, and on a file on my laptop?
Yes. Sheets, documents and decks work the same way, and a copy on a laptop stays in the same room through the Office add-in. In the demo, the CEO downloads the deck and opens it on his laptop. The add-in reads the deck link, joins the room and turns on live sync, and there is no merge step afterward.
Once the copy is live, what teammates and agents change lands in the file on the CEO's laptop while he works, and what he changes reaches the cloud copies the same way. It stays a normal file in its original format, with your masters, fonts and formulas, so there is nothing to reconcile when the meeting starts.
The sheet, document and deck editors in Athena sit on a shared-editing layer, and the Python libraries for the three formats are open source on PyPI, so developers can script the same shared files. The surfaces overview and Athena Anywhere cover the other places Athena works.
05What happens when an agent gets something wrong?
One agent session's changes can be rolled back without undoing anyone else's. Every change is attributed to a person or an agent, and each agent session is labeled with the person it works for, such as Athena via Jordan Kim. Figure 4 shows the Activity panel from the video.
In the video, the Activity panel groups edits under you, other humans and agents. A session called Refresh Q3 numbers changes the Opex figure to 16.2M. Dana Ruiz selects that session and clicks Rollback. The dialog explains that 2 edits will be undone from newest to oldest and that rolling back again undoes the rollback. Opex returns to 10.7M, a chip confirms it, and Priya's, Morgan's and Dana's own edits stay.
Attribution works down to the cell and links to the exact tool call that produced a value, and a rollback can span about 90 tool calls, with any irreversible action flagged. That control is what makes it practical to put many agents in one file. Athena has run 95 subagents writing to a single sheet without conflicts, and each of their changes stays attributed. Where you want a person to sign off before a write lands, human approval can be required on any tool. The attribution and reversibility page has the detail.
06How is this different from a shared AI workspace?
The work can stay in your own files. A shared space for people and agents is a good idea, and Athena applies the idea to the sheets, documents and decks a company already runs on.
The same agent works in the file the team edits in the browser and in the copy on a laptop, so nobody needs to keep a second version in sync. Templates, formulas and masters stay intact because the file format never changes. Attribution and rollback work per person and per agent session, which makes it practical to give many agents the same file. The whole arrangement runs in your VPC, on-premises or air-gapped, with 60+ models from every frontier provider behind it, so the model can change without moving the work.
07Is it secure enough for regulated companies?
Yes. Athena Intelligence is SOC 2 Type II and HIPAA compliant and runs in managed cloud, your VPC, on-premises or air-gapped, so the files stay inside your environment. Athena has zero-data-retention agreements with Anthropic, OpenAI and Gemini, does not train on customer data and does not store data from connected sources. Permissions are checked in real time, and a person can be required to approve any tool. A global consulting firm has run Athena in production for over a year. Figure 5 shows the boundary, and the deployment page has the options.
08How do teams get started?
Open any sheet, document or deck in Athena, or install the Office add-in and link the file you already have. Invite a teammate, mention Athena in a comment and watch the edit land. A recurring deck or model, such as a board pack, is a good first file.
Curious how it would work with your own templates? We're happy to show you.
09Questions teams ask
What does multiplayer mean for AI agents?
Multiplayer means people and AI agents work in the same file at the same time. In Athena, teammates and agents each have a name and a live cursor, edit in place, and talk to each other in comment threads. Nobody works in a separate copy, so there is one version of the file and every change has an author.
Can people and AI agents edit the same sheet, document or deck at once?
Yes. The sheet, document and deck editors in Athena are live for teammates and agents together, and the Office add-in keeps a copy on a laptop in the same room. In the demo, an agent builds a board deck while two teammates join, comment and edit, and the model and the memo fill with people and agents as well.
Can a copy of the file on my laptop stay live?
Yes. The Office add-in adds an Athena panel inside the apps you already use. In the demo, the CEO downloads the deck and opens it on his laptop, and the add-in reads the deck link, joins the room and shows Live sync on. Edits from teammates and agents then land in the file on his laptop, in your template.
How do I ask an agent to change something in a file?
Mention Athena in a comment, for example by asking it to compare Q3 to Q2 on a chart. Athena edits the chart in place and replies in the thread. If you ask where a number came from, Athena answers with the source file, sheet and row. You can also ask from the Athena panel in the Office add-in.
What if an agent makes a mistake?
Every change is attributed to a person or an agent session, such as Athena working for a named teammate. You can roll back one agent session, and only its edits are undone, newest to oldest, while everyone else's stay in place. Rolling back again undoes the rollback. Attribution works at the cell level and links to the exact tool call that produced the value.
Does the file keep our template and formatting?
Yes. The file stays in its original format, so your masters, fonts, formulas and layouts come with it. In the demo, every edit from the team and the agents ends up in the deck on the laptop, in the company template.
Where does my data go, and can it run in our VPC?
Athena runs in your own cloud environment: managed cloud, your VPC, on-premises or air-gapped. Athena has zero data retention agreements with Anthropic, OpenAI and Gemini, does not train on customer data and does not store data from connected sources. SOC 2 Type II and HIPAA apply, and you own the data plane.
Run this on a schedule
Scheduled briefs that land in shared files your team edits with the agent.