You have a Microsoft 365 Copilot license, and inside SharePoint there is now an agent offering to tag your document metadata for you. You also keep hearing about SharePoint Premium, which tags your document metadata for you. Both are Microsoft, both are AI, both promise the same thing, and they cost different amounts. So which one do you actually turn on?
The honest answer is that they are built for different jobs, and the reason so many teams are confused is that the jobs overlap at the surface. Here is what each one does, where each earns its place, and how to avoid paying for the wrong one.
What is the SharePoint Knowledge Agent?
The SharePoint Knowledge Agent, which Microsoft has since renamed AI in SharePoint and sometimes refers to as Copilot in SharePoint, is an AI capability built into SharePoint that suggests and auto-fills metadata columns for a document library, builds curated views, and helps site owners clean up their sites. It comes included with a Microsoft 365 Copilot license, and it is currently in public preview.
In practice, a site owner points it at a library and it scans the documents and proposes metadata: invoice numbers, client names, due dates, and amounts, filled into columns without manual tagging. It can build views from that metadata, create simple rules in plain language like “notify me when an invoice over $500 is added,” and flag site problems such as inactive pages, broken links, and content gaps people searched for but never found. It suggests, and a person reviews before anything sticks. It does not move, rename, or delete files, and existing permissions still apply. The whole thing is low-friction and driven by site owners rather than a configuration project.
One practical note for regulated and EU or UK tenants: the agent’s advanced features run on Anthropic models, and those tenants need to enable Anthropic as an approved sub-processor in the admin center, or users will hit errors. Worth confirming before you switch it on.
What is SharePoint Premium document AI?
SharePoint Premium document AI, the capability previously branded Syntex, uses trained models to extract specific, structured fields from the content of documents at scale. Instead of general suggestions, it pulls the exact values a process depends on: from a contract, the counterparty, the value, the expiration date, and the renewal terms, filled precisely across thousands of files. It is a separately licensed tier, not included with Copilot, and it is generally available rather than in preview.
The setup reflects the precision. You define a document type, configure or train a model to recognize the fields you care about, and point it at a structured library where those documents live. The payoff is accuracy at volume: a contract library where “every agreement expiring next quarter” becomes a filtered view instead of a day spent opening PDFs. That trade of higher setup for precise, high-volume extraction is the whole case for the tier, which our guide to when SharePoint Premium document AI is worth it walks through in full.
The overlap that causes the confusion
Both tools turn document content into metadata columns, which is exactly why they look interchangeable and exactly why teams pick the wrong one. The real difference is what each is built to tag, and how much precision and volume it is built for. One is a generalist, the other a specialist.
The Knowledge Agent is the generalist. Point it at a mixed library and it makes reasonable metadata suggestions across varied content to get that library organized and Copilot-ready, with almost no setup. Premium document AI is the specialist. Point it at a known document type and it extracts exact fields at the accuracy a contract or invoice process demands, after you have told it what to look for. Same output on the surface, a column of metadata, but one is breadth with low effort and the other is depth with real precision.
When to use the Knowledge Agent
Use the Knowledge Agent when you already have Copilot licensing and you want low-friction metadata enrichment and site cleanup across general libraries, without standing up a project to do it. It is the natural first move for broad Copilot readiness across varied content, because it is included and site owners can run it themselves.
It fits best when:
- Your libraries hold mixed, general content rather than one repeatable document type.
- Site owners will drive the work, and you want organization without a separate purchase or a training exercise.
- The goal is better findability and stronger Copilot grounding across a lot of content, not exact field extraction on one high-value type.
Because it rides your existing Copilot license, the marginal cost of trying it on a few important libraries is close to zero, which makes it the sensible place to start before you consider paying for anything more specialized.
When to use SharePoint Premium document AI
Use Premium document AI when you have a specific, high-volume, structured document type where precise field extraction matters and manual tagging is never going to happen. Contracts, invoices, claims, and orders are the classic cases, because the fields inside them drive a real downstream process and the volume is too high to handle by hand.
It earns the separate tier when:
- You can name the document type and list the exact fields you need pulled from it.
- The volume is high enough that manual tagging will realistically never get done.
- A legal or procurement process consumes that extracted data, so accuracy is not optional.
- The library is structured and consistent enough for a trained model to work against.
If you cannot name the document type and the fields, that is usually a sign the content is not ready for this tool yet, and the Knowledge Agent or a round of information-architecture work is the better next step.
Can you use both?
Yes, and in a large environment you often should. The Knowledge Agent handles broad Copilot readiness across your general content estate, while Premium document AI does precise extraction on the handful of high-value, high-volume libraries that justify it. These are not an either-or choice for the whole tenant.
The mistake is treating it as one decision for everything: turning on Premium document AI across libraries that do not need that precision, or leaning on the Knowledge Agent for a contract process that genuinely requires trained extraction. Match each tool to the content it fits. A single organization can reasonably run the Knowledge Agent widely and reserve Premium document AI for the contract and invoice libraries where the fields feed a process.
Neither one fixes a messy library
Both tools amplify the structure you already have, and neither invents structure you don’t. Point either one at a disorganized library full of duplicates and no agreed content model, and you get faster, more confident mess: metadata suggestions built on top of chaos, or extracted fields no one trusts. The library’s information architecture comes first, then the tool.
This is the part worth slowing down on before you turn either on. Deciding your document types, your content model, and what the metadata is actually for is the unglamorous work that makes both tools pay off, and it is the same Copilot-ready information architecture that shapes whether Copilot returns trustworthy answers at all. Get the structure right, then pick the tool that matches the content. If the underlying document management model needs work first, that is the place to start, not the AI switch.
Frequently asked questions
What is the difference between the SharePoint Knowledge Agent and SharePoint Premium?
The Knowledge Agent, included with Microsoft 365 Copilot, suggests and auto-fills metadata broadly across general libraries and helps site owners clean up sites, with almost no setup. SharePoint Premium document AI is a separate licensed tier that uses trained models to extract precise fields from a specific, high-volume document type. One is broad and low-effort, the other precise and configured.
Does the SharePoint Knowledge Agent cost extra?
The Knowledge Agent is included with a Microsoft 365 Copilot license rather than sold separately, so if your users are already licensed for Copilot, there is no additional purchase to use it. SharePoint Premium document AI is a separate tier with its own cost, which is part of why the two are worth telling apart before you buy anything.
Is the SharePoint Knowledge Agent generally available?
As of this writing, the Knowledge Agent is in public preview and Microsoft has changed its name more than once, so expect the branding and some behavior to keep shifting. SharePoint Premium document AI is generally available. Confirm the current status before you plan a rollout around either one.
Can the Knowledge Agent replace SharePoint Premium document AI?
Not for precise, high-volume extraction. The Knowledge Agent is strong at broad, low-effort metadata across varied content, but a contract or invoice process that needs exact fields pulled at scale is what Premium document AI’s trained models are built for. Many organizations use the Knowledge Agent widely and Premium document AI on the few libraries that require that precision.
Do these tools change my file permissions?
No. The Knowledge Agent adds structure like metadata, views, and rules, and does not move, rename, or delete files, and existing permissions continue to apply. Both tools operate within the access model you already have, so they organize and enrich content without widening who can see it.
Reviewed By
