We use ChatGPT extensively with Microsoft 365 and SharePoint as part of our internal business-systems and operational-governance workflows.
The SharePoint app is already extremely useful for retrieving and working with document content, but two limitations significantly restrict its usefulness for organisations that use SharePoint as a structured business information system rather than simply as document storage.
Requested capabilities
Native SharePoint List support
Please provide first-class access to SharePoint Lists, including the ability for ChatGPT to:
-
discover lists within sites the authenticated user can access;
-
retrieve list schema, columns and column types;
-
query and filter list items;
-
retrieve individual list items by ID;
-
follow lookup fields and relationships where permitted;
-
sort, filter and aggregate list data;
-
expose item metadata such as created/modified dates and users;
-
where write actions are enabled and authorised, create and update list items.
Access should continue to respect the authenticated user’s existing SharePoint and Microsoft 365 permissions.
First-class document and folder metadata support
OpenAI’s current SharePoint documentation states that the app can read content and metadata for files and folders. In practice, however, metadata is not currently exposed to ChatGPT in a sufficiently structured and reliable way for many business workflows.
Please expose SharePoint document-library metadata as first-class structured data, including:
-
site, library and folder;
-
filename and file type;
-
SharePoint item ID and drive/item identifiers;
-
created and modified timestamps;
-
created by and modified by;
-
content type;
-
version information;
-
SharePoint Document ID, where enabled;
-
custom library columns;
-
choice, date, Boolean, number, person/group, lookup, managed metadata and hyperlink fields;
-
approval and moderation status where applicable;
-
retention and sensitivity information where permitted;
-
sharing and permissions information where permitted.
It should also be possible to search and filter documents using metadata rather than relying predominantly on filename or document-content semantic search.
For example:
Find all documents in this site where Document Status is “Approved Current” and Authoritative Copy is “Yes”.
or:
Show all supplier agreements where Renewal Date falls within the next 90 days.
or:
Retrieve the exact SharePoint item with Document ID ABC-123 and show its metadata, version and location.
Why this matters
Many organisations deliberately use SharePoint metadata and Lists as their system of record.
Document content alone does not establish whether a document is current, approved, authoritative, superseded or merely reference material. Those distinctions are frequently stored in SharePoint columns rather than inside the document itself.
Without structured metadata and List access, ChatGPT can retrieve information from SharePoint but cannot reliably reason about the governance state surrounding that information.
This is particularly important for enterprise AI use, where source authority and provenance matter as much as the content itself.
Real-world use cases
Controlled document management
Our organisation maintains controlled documents in SharePoint.
A document may have metadata such as:
-
Document Status: Approved Current / Under Review / Superseded
-
Authoritative Copy: Yes / No
-
Document Type
-
System or Process
-
Owner
-
Review Date
Before ChatGPT relies on a document, it should be possible to verify that the SharePoint item is the approved current authoritative copy.
Without metadata access, ChatGPT can potentially retrieve the document content but cannot independently establish its governance status.
Internal systems register
We maintain a SharePoint List containing systems, ownership, implementation status, authoritative locations and integrations.
We would like to ask:
Which business systems are currently under evaluation?
Show all systems owned by Operations.
Which systems integrate with Microsoft 365?
Find systems that do not yet have an approved operating model.
This information naturally belongs in a SharePoint List, rather than in individual Word documents.
Decision log
We use a SharePoint List to record approved, rejected and superseded business-system decisions.
A useful query would be:
Has a decision already been made about adopting this system?
or:
Show the approved decisions affecting our SharePoint architecture since January.
Without List support, ChatGPT cannot interrogate the authoritative decision register directly.
Task and action management
A SharePoint List can contain implementation tasks, owners, due dates, dependencies and status.
ChatGPT should be able to answer:
What SharePoint tasks remain open?
Which overdue tasks are assigned to me?
What work depends on completion of the supplier portal?
For smaller organisations, SharePoint Lists are often entirely adequate for this purpose and avoid introducing another project-management platform merely to make structured data accessible to AI.
Contract and supplier management
Document libraries frequently contain supplier contracts combined with metadata such as:
-
Supplier
-
Agreement Type
-
Effective Date
-
Renewal Date
-
Notice Period
-
Agreement Status
This would enable questions such as:
Which supplier agreements require action during the next three months?
ChatGPT could retrieve both the agreement itself and the structured metadata governing its lifecycle.
Information architecture and governance audits
Administrators could ask:
Find documents marked as authoritative that are stored outside their designated authoritative library.
Show documents that have no Document Type assigned.
Identify files whose review date has expired.
Show duplicate Document IDs.
These are valuable governance and data-quality use cases that cannot reliably be solved from document-content search alone.
AI-assisted SharePoint administration
With appropriate permissions, ChatGPT could help administrators inspect SharePoint structures:
Show the fields and content types used by this library.
Compare the metadata schema of these two libraries.
Find lists containing a Supplier lookup field.
Show which libraries use the Controlled Document content type.
This would make the SharePoint integration useful not only as a knowledge-search connector, but as a genuine Microsoft 365 business-systems integration.
Current workaround
Because SharePoint Lists and library metadata are not currently sufficiently accessible, we have had to create an Excel workbook containing refreshed exports of several SharePoint Lists and document-library metadata.
ChatGPT reads the workbook as a non-authoritative retrieval bridge, while the live SharePoint Lists and libraries remain authoritative.
This works, but introduces unnecessary operational overhead:
-
users must remember to refresh the workbook;
-
exported information may become stale;
-
provenance is weaker than querying the source directly;
-
an additional artefact has to be governed and maintained;
-
ChatGPT cannot safely perform workflows that depend on live status.
Direct SharePoint List and structured metadata access would remove this entire workaround.
Suggested implementation model
The capability could follow the existing SharePoint app permission model and Microsoft Graph APIs.
Read operations could initially include:
-
Get sites
-
Get lists
-
Get list schema
-
Get list items
-
Get document-library items
-
Get fields/metadata
-
Search/filter items
Write functionality could remain a separately controlled action requiring explicit workspace administrator approval.
This would preserve least-privilege principles while significantly expanding the value of the SharePoint integration.
Business impact
For organisations using Microsoft 365, SharePoint is frequently not just a document repository. It is also a lightweight database, records-management platform, intranet and operational information system.
Giving ChatGPT first-class access to SharePoint Lists and metadata would enable considerably more reliable enterprise workflows, particularly around:
-
knowledge governance;
-
document control;
-
compliance;
-
contract management;
-
business-system administration;
-
project tracking;
-
records management;
-
source provenance;
-
workflow automation.
For us, this would eliminate a custom Excel export workaround and allow ChatGPT to work directly with the authoritative SharePoint information architecture.
The underlying principle is simple:
ChatGPT should be able to understand not only what a SharePoint document says, but what SharePoint says about the document.
And where structured business information is held in a SharePoint List, ChatGPT should be able to work with that information directly.