Use SharePoint’s Pre-Built AI Models to Extract Metadata INSTANTLY

Adding metadata manually is tedious—so why not let SharePoint do it for you? In this episode of the Ultimate SharePoint Content AI Deep Dive Series, Vlad and Microsoft MVP Gokan Ozcifci explore how to use pre-built AI models to quickly extract important information from your documents—no complex setup required!

What you’ll discover:
✔️How to leverage Microsoft’s built-in AI models to automate metadata extraction
✔️Why pre-built models are faster, cheaper, and simpler than creating your own
✔️Real-life examples with invoices, contracts, receipts, and more!
✔️Tips on when to use pre-built models versus custom-built solutions

If you’re looking for the easiest and most cost-effective way to enhance your SharePoint document libraries with AI, this video is for you!

This is part of our Ultimate SharePoint Content AI series — check out the full playlist to go even deeper.

Watch my 100+ courses on Pluralsight

Video Summary

• Start with pre-built models in SharePoint—they’re incredibly easy to use and the most affordable option for automating metadata extraction from documents.
• Choose from five ready-made models (Contract, Invoice, Receipt, ID, and Simple Document Processing) and apply them with just a few clicks—no training or coding needed.
• Upload PDF files (Word docs aren’t supported yet), and the model will automatically detect and extract relevant fields like totals, invoice numbers, and more. You can accept, rename, or change column types as needed.
• Deploy your model to other document libraries effortlessly. Once applied, it will automatically extract metadata from new files uploaded to those libraries.
• Perfect for common business documents like invoices and receipts—unless you have highly custom needs, this is the fastest, cheapest, and easiest way to get started with AI in SharePoint.

For more information, read the transcript blog below, or watch the video above!

Transcript

We all know that adding metadata to your SharePoint documents is beneficial. And for those of you who have been following the channel, we’ve now looked at multiple ways on how to create AI models to do the boring work for us. But what if I tell you that you might not even need to create a model because Microsoft already did the hard work for you?

In this video, I’m joined again by the amazing Microsoft MVP Gokan Ozcifci, who will deep dive into how we can use pre-built models to automate content extraction.

Yes, thank you so much for the introduction, Vlad. Gokan from Belgium, and I’m happy to be on that YouTube series and explain to you the pre-built models within SharePoint Content AI. But just before that—well, you mentioned the series—so for those of you who have been following the series for every video, well, thank you so much for being back. We hope you’re still enjoying the content. And for those of you who are new, this is the sixth episode in the Ultimate SharePoint Content AI series. We hope that after you listen to this video, you’ll listen to all the other ones with us.

But now, Gokan, I feel like we’ve covered two ways already, and you told me before this video that this one—I’m not going to believe how easy it is. And I already found the other ones easy, so I’m ready.

Yeah, so we’ve seen the unstructured data processing, then we have seen the structured data processing, and we’ve seen the SharePoint way—the AI Builder way—but there’s a third way that not everyone is using, but which I think is phenomenal from two perspectives. First one is it’s easy because Microsoft does the models for you. So you just have to apply this as we showed in the two previous videos on the document library, and you can extract the data. And secondly, well, it’s the price. It is extremely cheap compared to the unstructured and structured data processing models.

I like cheap.

It is very cheap. It’s the cheapest of all three. And the definition of my rule when I present this to customers is: always go pre-built first. Because you know, it might be that one of those pre-built models can extract all the data you need from that content. And you know, when we have an invoice, 99% of the same organizations—we all need the same content, right? Value, VAT, invoice number, and so on and so on. Maybe one of those pre-built models can do it for you. If that’s not the case, if you can’t achieve what you have to do with the pre-built model, then go for the unstructured, for the SharePoint way, even for structured data. Because you can do it, right? You can add the before and the after label—remember the video we showed—and you can use that because it’s again cheap compared to the structured one. And only go for the AI Builder-based AI models because, you know, it has a certain extra cost with AI Builder.

I feel like we’re doing the models the same way that we do everything: use out-of-the-box first, the pre-built first, before you build your own.

There you go. And that’s the golden rule I give to my customers. I only have one slide, and that’s the one that you see. I thought the more we advance in those models, the fewer slides we have.

Indeed.

And you will see, it’s like we open the Syntex box, and you will get the set of a pre-built model. You just select your model, and that’s it. There is only one definition that I like: the pre-built models in SharePoint use OCR combined with deep learning models to identify and extract text and data fields common to your document types.

What does that mean?

That means that Microsoft will use deep learning and OCR to analyze your data and then bring that data into your columns. But you can only choose the ones that they create for you. So you’re not able to add more columns.

Shall I show that to you? Maybe it’s easier to understand right now.

Of course.

With pre-built, it’s cheap, it’s efficient, but you’re limited.

There you go. That’s the trade-off you always have.

There you go. Within my SharePoint site, I have a document library called “Pre-Built Model.” Again, I click on the three dots, and I’ll see “Classify and Extract” here. I’ll create a new one, and on the second tab, you will see the pre-built models. We have five in total: the Contract, Invoice, Receipt, ID, and the Simple Document Processing. They all have specific purposes. I always use the first one—the Contract Processing—for contracts, Invoice for invoicing, Receipt for receiving, and then the Simple Documents is like the latest addition to the pre-built one.

And if I click on that, well, again, as in the two other videos, it shows you what can be extracted as an example. And if I click on “Next,” and that’s the only thing you need to do—give it a name. Let’s say “Gokan’s Expenses.”

Again, you’re an expensive person.

You spend too much money.

I should send you an invoice.

No, it’s just to have the same examples everywhere. And again, the content type and the sensitivity label, retention labels—we already explained that twice, so I’ll just skip that and click on “Create.” My model is now being created in my document library as a pre-built model.

Oh, it already exists. That’s the reason why I told you—you have too many expenses.

There you go.

Again, there you go. And now it’s opening. Remember the second video we did with the four boxes? Now I get like three. I just have to add my files here.

And you know what? I’m going to add something specific just for you, just to show you.

Just for me?

What do you think if I add Word files into my model? Would that work?

Why not? Let’s go.

So I have a few Word documents here, right? I’m going to say, okay, select those and click on “Add.”

They are not supported yet.

I see red. Red is bad.

Yes. So the Word documents—still not a big thing with the pre-built model today. So I would highly suggest that you to go with PDF files. Those are supported. When you click on “Add,” well, you will now see—hey, that error message has disappeared. And now I can add those documents to my model. If I click on “Next,” my job is done. And that’s it. My job is done. Now the AI model is analyzing my PDF files.

And now on the right side, you will see—oh my gosh—all the extractors it could find just for you.

That was so fast.

Yeah, like it could find four: the subtotal, the total, the trip fare, and the American Express blah blah blah. And then you can accept or deny that. So you can say, “Okay, I’m loving those.” So accept those. You can give it a name, and then you can just change the column type.

Wow.

Right? That’s so easy. And then if you’re happy with that, well, you just say, “Okay, I want those four. I accept them as they are. Yes, yes, yes, and yes too.” And then you just click on “Save and Exit.” And then you can just review all the files you imported into your model. You have like number three, and it will again analyze the files and then try to find extractors within your PDF file.

And now you see it has again found something.

And then it even found more.

Did you find more?

Yeah, it found like complaints, identification number, license plate, operator, and then even like a PayPal for some reason just here. You know, like the more you add files, the more you can actually—but some of them don’t make sense, like PayPal. I don’t care.

Don’t select it.

You don’t select it. And if you like this, you just select this, give it a name, change the column type, click on “Next,” and when you’re happy, you just save and exit. And your job is done.

How easy can it be? Like, this is like—anyone who has ever used SharePoint can build an AI model, which is pre-built. Again, as you said, it has its limitations. You cannot add your extractor, your columns. But did you see how much it found? Just like 10 columns. And I’m pretty sure you—

So now, how do I deploy it to another document library?

Yeah, so you just—you can click here on “Apply Model.” And as I showed in the other videos, you can create document libraries or select the one where you built it. So I can click on “Add,” and now it will be added to the pre-built model.

Right. It’s just so easy. I want to see it work to believe it.

There you go. And now you will see on that page all the extractors, you know, like the ones that we selected. But remember, on the first one, we had the option to add more? But here you don’t have this.

Right. No, it’s removed. It’s grayed out.

You can add models here. You cannot add the columns. And maybe we can add more files to train your model. Now, if I come to my site just here and then come to “Pre-Built” and let me upload some files—files, uploads—and then come here, just choose that one and then go on “Next.” Now you will again see that message like, “We’re analyzing your file.” I got the total, subtotal, American Express, trip fare—you know, like I got the total, subtotal, American Express, trip fare—you know, like all the columns we selected before. And then you get the two columns for “Processed” and “Processed Status” just here. Now we just have to wait—it could be a few seconds, could be a few minutes—and then those columns will be filled in with the data that got extracted.

That’s how easy it is, Vlad, to create a pre-built model in SharePoint.

That is amazing. Let’s do a quick refresh here. Let’s see if it worked. I’ll just click here… refresh… and the first one already—

Yeah, already did a bunch of stuff.

I was going to say we’re just going to cut until one comes, but it’s already there. Look—it’s already the second one, the operator, and it will just go, you know, one by one, and I will get more and more. But you see, like, I get all the information—all the things that I need to know—it’s all coming into my document library.

And the easiest part is it’s three clicks only. Like, you upload your files, select the values you want to extract, and apply it to a document library.

That is amazing. Of course, it’s limited, but if you have receipts, if you have invoices, contracts—unless you have some crazy needs—use this one. Especially since you told me it’s the cheapest one too.

It is. It is the cheapest one. And we can probably also add the prices on the screen the prices, but this is like the cheapest one that you can get from an AI model in SharePoint.

Awesome. Well, Gokan, this was—wow—I’m still mind-blown.

It is. It is mind-blowing, right? Pre-built models—because what people think when they hear about AI models is: it’s going to be complex, I need to maintain this, I need to create that, it’s going to be very complex to understand, to explain to people. But when you see it’s like maybe three or four clicks, and then you get the info from your PDF files—it’s amazing.

Oh my gosh, I love it. I love it. Well, Gokan, thank you so much for this amazing video. I can’t wait to go try it out.

And for everybody else, I really hope you have enjoyed the video. Of course, check out all the other videos in the series—you’ll see the playlist appear on your screen right about now. We’re not done with ways to extract information from content. There’s one more coming up, which I hope you will love. And you can check it out by simply clicking on the playlist which will appear on your screen right now. And of course, make sure you subscribe to the channel to get notified as soon as new videos come out.

See you in the next one!