Example: build a fine-tuning dataset from real saved conversations
Export a batch of saved AI conversations in ShareGPT format, then feed the resulting dataset into a model fine-tuning pipeline that expects conversation-turn structured data.
M365 Copilot export workflow
Export M365 Copilot to ShareGPT
Save M365 Copilot conversations as ShareGPT with Be Mimos. Capture chats locally, search your history, and export when you need a clean archive.
Product screenshot
See the export workflow
A real product screenshot gives search visitors and AI answer engines visual proof that the page is describing an actual workflow, not generic export advice.
What This Page Covers
This page walks through exporting M365 Copilot conversations to ShareGPT specifically: when to reach for this format instead of another one, the exact steps involved, and what it looks like once you have it. Further down, you will also find every other platform Be Mimos captures and every other format you can export to, so you are not locked into a single combination once you start saving conversations.
ShareGPT decision guide
When ShareGPT is the right export format
ShareGPT format is a structured conversation-turn export built for fine-tuning or evaluating conversational AI models — not for reading, but for feeding into a training or evaluation pipeline.
This page focuses on ShareGPT for M365 Copilot specifically, so the decision guide below is about when to reach for this format rather than a different one — not whether M365 Copilot itself is worth saving conversations from.
Use case 1Building a fine-tuning dataset for a conversational AI model
Use case 2Feeding saved conversations into tools and pipelines that expect the ShareGPT conversation-turn schema
Use case 3Evaluating model behavior against real, saved conversation examples
Use case 4Sharing conversation data with ML/AI tooling that specifically expects this format
What to check before you export
- Choose ShareGPT format only if your destination tool or fine-tuning pipeline specifically expects this schema — it's a machine-readable dataset format, not a reading format.
- Choose Markdown, PDF, or plain text instead if a human needs to read the conversation rather than a training pipeline needs to ingest it.
- Review the exported structure against your specific fine-tuning tool's requirements, since dataset schema conventions can vary by tool version.
How the export workflow works
This page is generated from the product fact registry, so it stays tied to supported providers, supported formats, and privacy guardrails.
- Install Be Mimos and open M365 Copilot. The extension attaches to the page automatically — there is no separate account to create before capture starts.
- Let the extension save the conversation while you work, or save the chat manually if you prefer to trigger it yourself. Either way, the saved copy lives locally on your device.
- Open your saved history, choose the conversation, and select ShareGPT export. The export runs against the saved copy, so the original M365 Copilot tab does not need to stay open.
- Choose ShareGPT format only if your destination tool or fine-tuning pipeline specifically expects this schema — it's a machine-readable dataset format, not a reading format. Keep the saved conversation in your history so you can re-export it in a different format later without redoing the work.
Why use Be Mimos for ShareGPT export?
The goal is not just to download a file once. It is to keep a searchable, reusable memory layer for your AI work, where ShareGPT export is one output option among several rather than the only way to get value from a saved conversation.
Local-first archive
Chat content stays local-first and is not uploaded to Be Mimos servers. That means the export step is just reformatting a copy you already have, not a new upload to a third-party service.
Search before export
Find the saved conversation from your AI history before choosing ShareGPT -- useful here for building a fine-tuning dataset for a conversational ai model -- instead of scrolling back through the original chat to locate it.
Format-specific export
Export saved conversations in ShareGPT from the supported format registry, using the same saved copy every other export format also draws from.
Once a conversation is saved, switching from ShareGPT to a different export format later does not require redoing any of the original work — the saved copy stays in your history and can be re-exported on demand.
Works Across Your AI Stack
The same local-first capture and export workflow described above runs on every supported platform, not just the one this page is about.
ChatGPT
conversations are captured and saved automatically, ready for search or export whenever you need them.
Claude
chats are saved the moment you have them, with export available whenever you go looking for the conversation again.
Gemini
every conversation is stored locally as it happens, so export is one click away instead of a copy-paste job.
Perplexity
conversations are captured and saved automatically, ready for search or export whenever you need them.
DeepSeek
chats are saved the moment you have them, with export available whenever you go looking for the conversation again.
Grok
every conversation is stored locally as it happens, so export is one click away instead of a copy-paste job.
Microsoft Copilot
conversations are captured and saved automatically, ready for search or export whenever you need them.
AI Studio
chats are saved the moment you have them, with export available whenever you go looking for the conversation again.
NotebookLM
every conversation is stored locally as it happens, so research archive is one click away instead of a copy-paste job.
Qwen
conversations are captured and saved automatically, ready for search or export whenever you need them.
Poe
chats are saved the moment you have them, with export available whenever you go looking for the conversation again.
Every Other Way You Can Export the Same Conversation
A saved conversation is not locked to one export format. Here is what each of the other supported formats is actually good for.
Markdown
Markdown is the practical choice when the conversation needs to remain readable as plain text while preserving headings, lists, links, and code-friendly structure.
PDF is the strongest fit for a stable reading copy that can be shared, printed, attached to a record, or reviewed without requiring the recipient to edit the conversation.
DOCX
Word or DOCX is useful when the conversation is only the starting point and the next step is editing, commenting, rewriting, or combining the answer with an existing document.
HTML
HTML is useful when the exported conversation should be opened in a browser, inspected as a web document, or used as the starting point for a private knowledge page.
JSON
JSON is the format to consider when the conversation must be handled as structured data for scripts, analysis, backup inspection, migration, or a custom processing workflow.
CSV
CSV is the right choice when the value of a conversation is in its structured, row-by-row data — not its prose — and the destination is a spreadsheet or a data pipeline rather than a document.
XLSX
XLSX (Excel) is CSV's richer sibling — useful when the same structured, row-based conversation data needs real spreadsheet features like multiple sheets, column formatting, or formulas, not just raw values.
ZIP
ZIP export is the practical choice when you need to move more than one conversation, or a conversation plus its attachments, in a single download instead of exporting each item individually.
Image
Image export turns a conversation (or a specific answer) into a visual snapshot — the right choice when the goal is sharing something to look at, not a file to edit or reprocess.
Google Docs
Exporting to Google Docs is the right move when the next step is collaborative editing, commenting, or sharing inside Google Workspace, rather than working from a local file.
Plain Text
Plain text strips everything down to the words themselves — the most universal, dependency-free format for when you just need the content, with no formatting, markup, or file-type compatibility concerns.
Alpaca
Alpaca format restructures a conversation into instruction/input/output triples — the standard shape used for instruction-tuning AI models, not for casual reading.
Tavern
Tavern format packages a conversation as a character-card-style chat log, built for AI roleplay and companion-chat applications rather than general document use.
Ooba
Ooba format targets Oobabooga's text-generation-webui — useful when you're running or fine-tuning models locally and want a saved conversation available as chat history inside that specific tool.
Related export pages
Start with the format you need now, then keep the same saved conversation available for other workflows.
FAQ
Can I export M365 Copilot conversations to ShareGPT?
Be Mimos supports saved AI conversation exports to ShareGPT. Use it with supported M365 Copilot conversations captured by the extension.
Does Be Mimos upload my chat content to company servers?
No. The public privacy positioning for generated pages is local-first: chat content is not uploaded to Be Mimos servers.
Can I export to other formats too?
Yes. The product fact registry includes PDF, DOCX, Markdown, HTML, JSON, CSV, XLSX, ZIP, image, Google Docs, and ML training formats — see the full list below.
Is ShareGPT format meant for reading conversations?
No — it's a structured, machine-readable dataset format built for AI model fine-tuning and evaluation pipelines. For reading, use Markdown, PDF, or plain text instead.
Do I need any special tooling to use a ShareGPT export?
You'll typically feed it into a fine-tuning or dataset-processing pipeline that already expects the ShareGPT conversation-turn schema — it's built for that kind of downstream tool, not for opening directly in a document viewer.
What does Be Mimos cost?
Available plans: Free, Pro Monthly (4.99), Pro Yearly (39.99), Pro Lifetime (99.95). ShareGPT export works from the same saved-history foundation on every plan.
Do I have to find the M365 Copilot conversation manually before exporting?
No. Users can search saved AI conversations across supported providers. Find it by keyword or date, then export straight from there.
Does Be Mimos analyze or scan the content of my conversations?
Sources say there is no analytics on chat content and no data mining of conversations.
Is my export or search activity used for advertising?
Sources say there is no advertising tracking and no third-party advertising cookies.