Boost your SharePoint productivity by letting AI do the heavy lifting.
In this full tutorial, you’ll learn how to set up and use Autofill Columns in SharePoint to automatically extract metadata like invoice numbers, client names, totals, and more — using just a few words and no code.
What you’ll learn:
✔️How Autofill Columns work in SharePoint
✔️Setting up prompts for metadata extraction
✔️Practical demos for invoices, dates, amounts, and even currency
✔️Cost, limitations, and upcoming roadmap
Whether you’re managing contracts or optimizing document libraries, this feature is a game-changer.
▶️ Part of the Ultimate SharePoint Content AI Series – make sure to check out the full playlist
Watch my 100+ courses on Pluralsight
Video Summary
- Autofill columns in SharePoint use AI to extract metadata like invoice numbers, client names, dates, and more—automatically, with just a few clicks and no code required.
- You can write simple or detailed prompts to guide what data gets extracted and how it’s formatted. Whether it’s “return the invoice number” or a full instruction, SharePoint Content AI gets the job done.
- It’s fast, flexible, and supports backfilling existing files. You can even apply logic like “add a 10% tip if over $5,000” or set default values when data is missing.
- It supports a wide range of file types (PDFs, Word, Excel, images, emails, etc.) but currently only works with English-language content and library columns—not site columns or managed metadata.
- It’s cost-effective at just $0.005 per page, no matter how many columns you extract. And with upcoming features like AI-suggested columns and a new processing history pane, the future of metadata in SharePoint looks amazing.
For more information, read the transcript blog below, or watch the video above!
Transcript
One of the most important things you can do to improve your experience in SharePoint is to make sure that all your documents and list items are tagged with meaningful metadata. And what if I told you that with only a few clicks, you can create columns that will automatically fill in metadata using the power of artificial intelligence? You know what? Let me take a step back because I know a lot of people don’t fully understand what is the value of metadata. And today, in the era of AI, metadata is more important than ever. Whether it’s for security and compliance and making sure that all of the different automating labeling policies that your Purview and compliance admins are doing are being applied properly to make sure that if regulations say you need to keep a contract for seven years, all your contracts are properly tagged, the dates are properly tagged, and automatically after seven years they can either be kept for at least that time or automatically deleted.
Metadata is also important for improving search and retrieval. We all know that SharePoint search has not been the most popular tool over the last decade. But if we add the proper metadata to it, your admins and site owners can create custom search experiences, giving you filters, custom views like the one you see here on the screen, or even the nice content query functionality, allowing you to add amazing filters directly on your library experience.
And of course, today we have to talk about automations. Whether it’s the more basic SharePoint rules or more advanced flows into Power Automate, you need metadata to create smart decisions. Whether it’s a condition depending on the date of an invoice or a date that a contract expires, whether it’s simply knowing the owner of a certain contract, all your automations rely on smart metadata being added.
And of course, how can we skip talking about Copilot and agents? You see, Microsoft 365 Copilot can understand metadata, and so can your SharePoint agents. And I’ve worked with so many companies that aren’t getting the best experience out of Copilot, unfortunately. And a lot of that is not due to Copilot, but it’s simply due to their SharePoint tenant being ignored for the past decade, simply being used as a file share where we dump files without any single metadata added. But Copilot and AI can actually understand your metadata, and the next time you interact with it, you will get a better result.
In this video, we will focus on autofill columns, which allow you to extract metadata using only your words. And again, nothing technical here. It will just be a few clicks. And for those of you who have been watching the Ultimate SharePoint Content AI series, this is the seventh episode. For those of you who are new to the series, well, welcome. I really hope you’re going to enjoy this video and check out the rest of the series after. But let’s get started covering autofill columns.
Before you’re able to do everything that I will show you in this video, there are only two requirements. Your admin needs to have set up Pay-As-You-Go and have enabled autofill columns on either your tenant or the specific site that you’re going to be trying this on. And if you haven’t done it yet, we have a full dedicated video on how to enable Pay-As-You-Go in your Microsoft 365 tenant.
You know what? Enough slides. Let’s go ahead and go to the demo. So here I’m in a new SharePoint site. And just to show you how easy it is, I’ll actually just get started from nothing. So I’ll go here. I will create a new document library. And once this loads here, I will call it “My Invoices.” Let’s start with a blank library. I’ll call it here “My Invoices.” Great. Show inside navigation. Perfect. Let’s click on Create.
And what I will do is what most companies do here, right? They create the library, and then they just drag and drop their documents. I don’t even have columns yet. I don’t have anything. And right now, this is how your tenant might be looking—a bunch of content but no real metadata.
What if we wanted to extract some metadata from it? You see, this is a very basic invoice where I have a logo, who I am billing—so this is my company details—who I am billing, invoice number, dates, due date, and what we are billing the customer for. Let’s say I wanted to have the invoice number as a column. What I can do is simply go here to add a column. And of course, here you can choose the one that is better for you, whether it’s a text or a number. Every company has their own invoice number format. Let’s go here with text. I will call it here “Invoice Number.”
And then if you have autofill set up, you will see this over here where it says “Autofill.” Let me click on Setup. So now I can tell SharePoint Content AI what information to extract for this column. You see, here at the top, I actually have a little loading bar under the prompt. What this will do is sometimes, depending on—and there we go, it did it exactly as I was going to explain it—depending on the name of the column, it will automatically suggest a prompt for me. And the prompt they suggest is really good: “Extract the invoice number from the document. Output the number exactly as it appears, preserving any leading zeros or formatting. If the invoice number is not found, return ‘not found’ and give it back to me as a string.”
Because I created a text column. Now, you know what? Let me get started with the fancy prompt by Microsoft, and then I’ll show you it doesn’t have to be that complicated. Then I can test my prompt against some of the files in my library. So when you do it, add a few files to your library. This way, you have things to test with. Let’s test with the Contoso PDF file. Let’s give it a second. And it gave me invoice 556. Let me go see if I can open the PDF file here. Great—556. So that was good.
I can also test it again. Let’s test it with Fabricam. Hopefully, I gave it another invoice number here. Those are all fake invoices. Let’s see what Fabricam should be. Well, Fabricam is also 556. Let me see if—ah, there we go. The Neoxy, which is the last one I uploaded here, has a different one. Let’s test the Neoxy PDF file, and I should have 524. Amazing.
But again, this might seem very complicated, even if it’s Microsoft that gave it to me. I didn’t have to do anything. Your description of what you want to get from that column can be as simple as “Return the invoice number.” Let me just actually test this. Let’s go to Contoso. Let’s see if it’s able to do it with as simple a prompt as that. I can even ask it to modify it. And now you see the beauty of this video, and the beauty of GenAI in general is that I’m never sure exactly what I’m going to get. So what I can say is “Return the invoice number. Remove any leading zeros from the number.” Let’s just try this. See if I can just get 556 as an example here. I’m not saying you should do it in your environment—you probably want to leave the invoice numbers as is—but I just wanted to show you. You see, my result number now is 556. So I was not only able to extract information, but I was also able to modify it. Amazing.
You know what? I will just put here “Return the invoice number,” and let’s click on Save here and let me save this column. Now, what else could we have for an invoice? Let me add another column here. Let’s go text. I’ll say here “Client Name.” And here, let’s go set up. And I’ll start typing. Probably, Microsoft will suggest a nice prompt for me. Well, let’s see. Man, it’s so fast. Now, “Extract the full name of the client from the document. Ensure the name is capitalized with the first letter of each word. If the client name is not found, return ‘object not found.'”
Let’s test it. This one should be Contoso for the first one. The second one should be Fabricam because it doesn’t look at the name of the PDF file—it looks inside the file. Great. And again, this is a very good prompt by Microsoft, but you can also say, “Extract the name of the client from this invoice. Only include the company name.” As an example, let me test it with Neoxy over here. Let’s see if it works. It should work as well.
At the end of the day, I don’t want you to need to write paragraphs as big as this one for autofill columns to work. Of course, as always with GenAI, the more details you add, the better your output will be. So anything that you expect from the system—if, for example, you didn’t want any leading zeros—well, tell the autofill column to remove them. You are in charge. And it might happen that the first prompt you write is not perfect. This is why we have the test functionality. Go ahead and add things to it. Let’s save this column. You’ll notice that nothing gets added immediately. However, whenever I upload a new file, it will take a few seconds—or up to 24 hours, according to Microsoft documentation—for the metadata to be extracted. In real life, I’ve found it usually takes anywhere from a few seconds to 10 minutes, depending on how many documents you upload at once.
You’ll also see that a new column appears called Processing Status. Let me refresh the document library for good luck. Still nothing yet, but the processing status is really useful if you want to do what Microsoft calls backfill. Backfill means running the process again on content that’s already in the library. For example, I already had some documents in here, and I want those nice autofill columns applied to them. I can select all three, go to the three-dot menu, and click on Autofill.
And with a little bit of luck—well, it’s funny that as soon as I did that, the one at the end finished. Let’s see how we did. I’ll open the file. My invoice number should be 476, and the billed client should be Wingtip Toys. If we go to the column in SharePoint—amazing! Invoice number 476 and the client’s name is Wingtip Toys. Exactly what we had.
Let me refresh again to see if the others updated. Not yet, but you can see the processing status is “In Progress.” So you do have visibility into whether it’s analyzing, finished, or errored out. And wow—as I was talking, it finished all of them. I have my Contoso and Fabricam files, and as we saw, they had the same invoice number. I have the correct client names and the Neoxy invoice as well.
Now let’s take it to the next level. I have an invoices library here, and I’ll upload more documents. While autofill columns do their thing, I want to show you some other ideas of information you can extract. We’ve already seen the invoice number. I have a column called Invoice Date. The prompt I used is: “Extract the date of the invoice from the document. Format the date as year-month-day.” I can test it here. It did a good job—2025, May 4th. And because my column is a date and time type, it mapped the result correctly. That’s amazing.
As we talked, and while we wait for the rest of the columns to fill, it took maybe a minute—maybe even less—to get all that information. That’s pretty awesome. I have all my invoice dates, client names, and total invoice amounts.
Let’s look at the Total Invoice Amount column. The prompt is: “What is the total amount of the invoice? Provide only the number formatted with two decimal places.” I told autofill to show the trailing zeros (e.g., 150.00) and not include currency symbols or extra text. However, it only gave me back 8000—it didn’t follow the formatting rules. That can happen. It’s not perfect.
But remember, we have all the power of SharePoint. So I can go into the column settings and set the number of decimal places to two. Bam—it’s solved. Exactly the formatting I wanted.
If I want to, I can also modify the prompt. As a test, I can say: “If the value is over 5,000, add a 10% tip.” Let me test it. This should now be 8800. There we go. But you see, the Neoxy Consultancy invoice was only 2000, so that one remains unchanged. You can add rules like that, which is quite impressive. I’ll revert it to normal—no tip added.
I can also extract the currency. For example, in the Wingtip Toys invoice, there’s no currency mentioned. But in the Neoxy one, I specify euros. My prompt is: “Find what currency the invoice is in. If a currency such as USD or GBP is not mentioned, the default should be CAD. The result should only be written as the three-letter code of the currency.” This is great because I can set a default value and standardize the output.
Let’s do one final example. I can add a Summary column. The prompt is: “Summarize the services or products listed in the invoice. Respond with the quantity or units provided and then the products or items listed in a short phrase.” Let me test it. You see, I get “40 hours of Microsoft 365 consulting.” That’s pretty awesome. I also have “80 hours of M365 consulting,” “1 Microsoft 365 monthly support,” and “1 Power Platform monthly support.”
Again, I set up the columns, and now every time I add documents, all that information is extracted for me. Isn’t that simply amazing?
Before we finish this video, there are a few more things I want to talk about because I want to cover everything. Let’s talk about supported file types. Not every file is supported, but as you can see in the slide (which I won’t read out loud), quite a few formats are. We have Word documents, text files, PDFs, Excel, PowerPoint, a ton of image formats, HTML, and even email messages. So chances are, the file format you need will be supported.
However, I live in Quebec, where both French and English are mandatory. As of today, autofill columns are only available for English-language files. Microsoft has said that other languages will be added in future releases, so I’m sure we’ll see more language support soon.
Now, what column types are supported? Again, I won’t read all of them—you can see them on the slide. It’s easier to tell you which ones are not supported: person or group, location, image, lookup, and managed metadata. For managed metadata, we actually have a separate video in this series on taxonomy extraction, which lets you do something similar and map it to a term set inside your managed metadata. So while autofill columns can’t do it directly, other tools can.
There are a few more limitations. As of today, only library columns can be autofilled columns. You cannot use site columns or content types that include autofill columns. I know many people have been asking Microsoft about this, and they’ve said they’re working on it. They want to bring this functionality to content types, but as of July 2025, it’s not available yet.
Now let’s talk about something really important—cost. And I have really good news: it’s quite cheap. It costs half a US cent—that’s $0.005 per page that you upload and that is processed for autofill columns. The cost is the same no matter how many prompts you have per page. So, whether you extract one column or ten from the same document, the price is the same. That means you should make the most of it—add all the columns you need, since you’re paying the same price anyway.
Let’s talk a bit about the roadmap, because Microsoft is heavily investing in autofill columns and has some exciting features planned. I want to get you excited not just about what you can do today, but also about what’s coming.
One of my favorite upcoming features is suggested columns. Imagine uploading a document to a library, and the AI system says, “Hey, I see this is an invoice. Here are 10 columns I think you should add.” Wouldn’t that be amazing? That’s coming soon.
Other features on the roadmap include multi-file testing, provisional backfill, and easier, better classification.
There’s also a new feature coming by the end of July 2025. Let me show you the blog post—it was just published a day or two ago. Remember the Processing Status column I showed you in SharePoint? That’s going away soon. Instead, we’ll have a Processing Pane that shows the history for the last 60 days. This will apply not just to autofill, but also to document translation and other features.
This new pane will let you see exactly what happened: for example, “Five minutes ago, we processed this file. Eight columns were processed, and all eight were filled.” Or, “This one failed because it was encrypted or password-protected.” Or, “This file had 10 columns processed, and all 10 were successfully extracted.” I wish I could show it to you live, but for now, I can only show you a screenshot. Just know that even the status and processing experience are getting better.
So, this is some amazing stuff, and that new pane should be here by the end of July 2025.
Thank you so much for watching this video. I truly hope you enjoyed it and that you’re as excited as I am about the autofill functionality. It truly takes metadata to the next level. And if users didn’t want to fill in metadata before, now you can just configure the column for them, and that’s it. It will be done automatically.
If you’ve enjoyed this video, make sure you like it, subscribe to the channel, and check out the other videos in this playlist. If you’re looking for more ways to infuse AI into SharePoint and your content management, there’s a lot more to explore. See you soon!
