Download to read offline
Uploaded byDrew Madelung
Discover - Assemble - and Gain Insights into your Content with SharePoint Content AI - TechCon365 Dal.pptx
SharePoint Content AI brings advanced content services and AI together to automate processes, classification, and more, transforming the information into working knowledge for your organization in a way we've never seen before. In this workshop, you'll learn how Content AI has come to the forefront of automated knowledge management, the services and solutions it can provide, how they work together, and some proven ways to implement it in your organization. You'll walk away from this workshop with real-world SharePoint Content AI implementation, deployment, and adoption scenarios to spark the conversation at your organization.
More Related Content
Similar to Discover - Assemble - and Gain Insights into your Content with SharePoint Content AI - TechCon365 Dal.pptx
More from Drew Madelung
Discover - Assemble - and Gain Insights into your Content with SharePoint Content AI - TechCon365 Dal.pptx
- 1. Discover, Assemble, and GainInsights into your Content with SharePoint Content AI SharePoint Fest Chicago 2021 #SPFestChi TechCon365 Dallas 2025 Drew Madelung
- 3. Drew Madelung Email :drew.madelung@protiviti.com Twitter : @dmadelung Website: drewmadelung.com Director – Microsoft Solutions
- 4. Content management stateof the union What are my AI options in SharePoint? How do I set it up? Demos on Demos Agenda Discover, Assemble, and Gain Insights into your Content with SharePoint Content AI TechCon365 Dallas 2025 How should I use them?
- 5.
- 6. Challenges Lost productivityfrom manual search & processing Decisions on incomplete information Missed data opportunities Expensive information silos Data not optimized for Copilot Too much data, not enough time SharePoint isn’t Google search
- 7. Opportunities Efficiency Streamlinedprocesses Competitive advantages Cost management Data optimized for Copilot Adoption
- 8. We need metadatamore than ever Searchability and findability Content context Automation and compliance Data driven insights When documents are properly tagged with metadata (like customer names, invoice numbers, contract dates), users can search and filter easily instead of browsing manually. Without good metadata, even the best search engines struggle because they’re guessing based only on text inside documents — not structured data. Classification helps the system and users understand what a document is (e.g., invoice vs. contract vs. report). When a document is correctly classified, it can trigger the right processes (such as reviews, approvals, or data transfers). Regulatory compliance often requires that specific types of documents be retained, reviewed, or protected differently. Without proper tagging, automating compliance is almost impossible — you're relying on people to guess or remember what each file is. Classified data can feed effective reporting, analytics, and Copilot responses. If the content is not well-organized, AI can't reliably surface insights ("show me all active contracts ending this year") because the data isn't structured.
- 9. We need metadatamore than ever "Imagine searching for a specific book in a library where none of the books are sorted by title, author, or genre — and there’s no catalog. Even the smartest librarian would struggle.”
- 10. What is SharePointcontent AI? Project Cortex Nov 2019 SharePoint Syntex Oct 2020 Microsoft Syntex Oct 2022 SharePoint Premium Jan 2024 SharePoint Premium and Content Management: Re view and What’s Next in 2024 Announcing SharePoint Syntex Introducing Project Cortex Introducing Microsoft Syntex - Content AI in the fl ow of work AI Powered Features in SharePoint Oct 2025
- 11. The three pillars SharePointAdvanced Administration SharePoint Content Solutions Annotations AI-Driven rules SharePoint eSignature Agreements App AI Powered SharePoint Solutions (Content AI) Document Processing Content Assembly OCR Image Tagging Taxonomy tagging Translation for files and videos Autofill Knowledge Agent SAM - DAG Reports SAM - Site Access Reviews SAM - Restricted Access Control MGDC Microsoft 365 Backup Microsoft 365 Archive These are my personal views and may not reflect those of Microsoft. The company may no longer align with the three pillars as outlined. Slide courtesy of Gokan Ozcifci
- 12. SharePoint has anew generation of AI-powered services that understand, process, and manage content (documents, data, media) in ways that enhance productivity, compliance, and discoverability — built right into Microsoft 365 and SharePoint. AI-powered document processing Content classification and taxonomy Autofill columns based on document content Integration with M365 Copilot Knowledge Agent! Improving content in SharePoint with AI
- 13. Content services SharePointdocument processing
- 14. Document classification Metadata extraction Label assignment Metadata search Document discovery Content use/ response generation Content creation 1 Content is createdand collected through content creation process 2 Documents are classified 3 Metadata that has been assigned through classification is extracted 4 Documents are correctly labeled 5 Metadata search is used to better find documents 6 Documents are found 7 The content is used and/or an appropriate response is filed based on the scenario Typical content lifecycle
- 15. Autofill columns Streamlinethe process of managing files and their associated information by using large language models (LLMs) to extract or generate content automatically. Content assembly Automatically generate standard repetitive business documents, such as contracts, statements of work, service agreements, letters of consent, correspondence, and more. Document translation Create a translated copy of a document in a SharePoint document library, preserving the original format and structure of the file. Available for all supported languages and dialects. eSignature Send electronic requests using SharePoint eSignature, keeping your content in Microsoft 365 while it’s being reviewed and signed. Image tagging Find, sort, filter, and manage images in SharePoint document libraries. Optical character recognition Extract printed or handwritten text from images, letting you quickly and accurately find the keywords and phrases you're looking for. Prebuilt document processing Save time processing and extracting information from contracts, invoices, receipts, and other types of documents. Structured and freeform document processing Automatically extract information from documents, such as letters, contracts, forms, and invoices, where the information can appear anywhere in the document. Taxonomy tagging Automatically tag terms or term sets in SharePoint document libraries, making the files easier to search, sort, filter, and manage. Unstructured document processing Automatically classify documents that vary in composition and extract information from them. Knowledge Agent Unified agent experience of SharePoint’s AI tools including content management, content creation, and content consumption skills SharePoint document processing solutions
- 16. Conventional ML • “Blackbox” AI • Large training sets • Human expertise reviews results later Large content set AI identifies patterns AI applies tags Experts QA the AI model Evolution of AI in SharePoint
- 17. Machine teaching • Human-centricAI • Small sample sizes • Experts review results early Experts tag a small sample set AI encodes human expertise as tagging rules New content is tagged Evolution of AI in SharePoint
- 18. LLM (Copilot) • Trainedon massive datasets • Prompt based interaction • Highly flexible without retraining Evolution of AI in SharePoint
- 19. SharePoint document processingsolutions Autofill columns Streamline the process of managing files and their associated information by using large language models (LLMs) to extract or generate content automatically. Content assembly Automatically generate standard repetitive business documents, such as contracts, statements of work, service agreements, letters of consent, correspondence, and more. Document translation Create a translated copy of a document in a SharePoint document library, preserving the original format and structure of the file. Available for all supported languages and dialects. eSignature Send electronic requests using SharePoint eSignature, keeping your content in Microsoft 365 while it’s being reviewed and signed. Image tagging Find, sort, filter, and manage images in SharePoint document libraries. Optical character recognition Extract printed or handwritten text from images, letting you quickly and accurately find the keywords and phrases you're looking for. Prebuilt document processing Save time processing and extracting information from contracts, invoices, receipts, and other types of documents. Structured and freeform document processing Automatically extract information from documents, such as letters, contracts, forms, and invoices, where the information can appear anywhere in the document. Taxonomy tagging Automatically tag terms or term sets in SharePoint document libraries, making the files easier to search, sort, filter, and manage. Unstructured document processing Automatically classify documents that vary in composition and extract information from them. Knowledge Agent Unified agent experience of SharePoint’s AI tools including content management, content creation, and content consumption skills
- 20. Content services Autofillcolumns
- 21. Autofill Autofill columns automaticallyextract, summarize, or generate content from files uploaded to a SharePoint document library. By using large language models (LLMs) through generative AI, autofill columns can save metadata automatically. • Saved prompt used to process files created or uploaded • Extract, classify, summarize, and analyze the content to new metadata • Can work alongside other models • Currently English only • No person, lookup, or managed metadata support yet • file types: .csv, .doc, .docx, .eml, .heic, .heif, .htm, .html, .jpeg, .jpg, .md, .ms g, .pdf, .png, .ppt, .pptx, .rtf, .tif, .tiff, .txt, .xls, and .xlsx. • Default enabled for all sites
- 22. Autofill suggested prompts Columntype Suggested prompt Example Number <insert the question here>? Respond with a valid number. If no proper answer is found, respond with <insert default response>. What is the total number of hours agreed upon? Respond with a valid number. If no proper answer is found, respond with "Not applicable." Yes/No <insert the question here>? Respond with "Yes" or "No" and only "Yes" or "No." If the documents don't provide enough information, leave the answer blank. Is the tenant required to have renter's insurance? Respond with "Yes" or "No" and only "Yes" or "No." If the documents don't provide enough information, leave the answer blank. Date and time <insert the question here>? Format the answer as a valid C# datetime. If no proper answer is found, respond with <insert default response>. What is the deadline for the quarterly financial report submission? Format the answer as a valid C# datetime. If no proper answer is found, respond with "No deadline set." Choice Make a decision about this document based on the following categories: <Choice 1>, <Choice 2> and <Choice 3>. Reply with one of the choices only. Reply with "None of the above" if none of the choices fits. Make a decision about this sales agreement based on the following categories: Signatory Approval, Legal Review, and Finalize Sale. Reply with one of the choices only. Reply with "None of the above" if none of the choices fits. Choice Make a decision about this document based on the following categories: <Choice 1> <insert description for choice 1>, <Choice 2> <insert description for choice 2>, and <Choice 3> <insert description for choice 3>. Reply with one of the choices only. Reply with "None of the above" if none of the choices fits. Make a decision about this document based on the following categories: Approval Needed: Requires managerial review and approval before proceeding, Information Only: For informational purposes and doesn't require any action, and Urgent Update: Contains time-sensitive information that requires immediate attention and updating. Reply with one of the choices only. Reply with "None of the above" if none of the choices fits. Hyperlink <insert the question here>? Reply with the name formatted as a valid URL. Who is the insurance provider? Reply with the name formatted as a valid URL. Currency <insert the question here>? Reply in a valid U.S. currency format. If the documents don't provide enough information, leave the answer blank. What is the total cost of the project as detailed in the invoice? Reply in a valid U.S. currency format. If the documents don't provide enough information, leave the answer blank.
- 23.
- 24.
- 25. Custom Models PrebuiltModels • Unstructured document processing • Teaching method • Structured document processing • Layout method • Freeform document processing • Freeform selection • Contract processing • Invoice processing • Receipt processing Model types
- 26. Unstructured document processing •Extract sentences or specific regions of the document • Models can be created in a content center • Models applied to any library you have access to • Associated to a content types • Utilizes “classifier” to determine document • Utilizes “extractor” to pull info from documents • Not limited by AI builder credits Train on 5-10 PDF, Office, mail files, including negative examples Single class model -> (aka teaching method) Model details
- 27. Structured document processing •Used for semi-structured file formats • Forms or invoices • Relies on AI builder (models) • Admins enable per site • Supports more than 100 languages Train on PDF, JPG, PNG format, totaling 50MB/500pp and less than 20 MB Structured extraction model (aka Layout method) Model details
- 28. Freeform document processing •Unstructured and free-form formats • Documents with no set structure • Letters, contracts, and SOWs • Relies on AI builder (models) • Admins enable per site • Supports more than 100 languages Train on PDF, JPG, PNG format, totaling 50MB/500pp and less than 20 MB Freeform extraction model (aka Freeform selection) Model details
- 29. Model details Feature Unstructureddocument processing Freeform document processing Structured document processing Use for this type of content Unstructured or semi-structured file formats, for example Office documents where there are differences in the layout, but still similar information to be extracted. Unstructured and free-form file formats, for example documents that have no set structure such as letters, contracts, and statements of work. Structured and semi-structured file formats, for example PDFs for forms content such as invoices or purchase orders where the layout and formatting is similar. Model creation Model created in SharePoint. Model created in AI Builder with seamless access from SharePoint document library. Model created in AI Builder with seamless access from SharePoint document library. Classification type Trainable classifier with optional extractors using machine teaching to assign document location on what data to extract. Not applicable Not applicable Supported file types Train on 5-10 .pdf, Office, or email files, including negative examples. Files are truncated at 64,000 characters. OCR- scanned files are limited to 20 pages. Supports more than 20 file types. See supported file types. Train on .pdf, .jpg, or .png format, total 50 MB and 500 pages. Train on .pdf, .jpg, or .png format, total 50 MB and 500 pages. Integrate with managed metadata Yes, by training entity extractor referencing a configured managed metadata field. No No Supported regions Available in all regions. Relies on Power Platform. Relies on Power Platform. Supported languages Supports more than 40 languages. Supports more than 40 languages. Supports more than 100 languages.
- 30. Prebuilt Model details Contract InvoiceReceipt Sensitive information Simple document Analyzes and extracts key information from contract documents. The API analyzes contracts in various formats and extracts key contract information such as client or party name, billing address, jurisdiction, and expiration date. Analyzes and extracts key information from sales invoices. The API analyzes invoices in various formats and extracts key invoice information such as customer name, billing address, due date, and amount due. Analyzes and extracts key information from sales receipts. The API analyzes printed and handwritten receipts and extracts key receipt information such as merchant name, merchant phone number, transaction date, tax, and transaction total. Analyzes, detects, and extracts key information from documents. The API analyzes contracts in various formats and extracts key sensitive information such as social security numbers, financial account numbers, driver license identification numbers, and other personal information. Offers a flexible, pretrained solution for extracting key- value pairs, selection marks, and named entities from basic structured documents. Unlike other prebuilt models with fixed schemas, this model can identify keys that others might miss, providing a valuable alternative to custom model labeling and training. Start here when working with models
- 31. Centralized management formodel creation and management Integrated home for managing SharePoint document processing AI models and processes in Microsoft 365 Inherits content management and user experiences from SharePoint Develop and publish AI models for structured and unstructured documents Monitor classification and process with embedded analytics Process and share content with Power Automate and rich formatting Content centers
- 32. Content services Demo Workingwith models
- 33. Content services ContentAssembly
- 34. Automatically generate standardrepetitive business MS Word documents such as contracts, statements of work, or sales pitches. • Start with an existing document • Fill in data from free form or metadata from lists or libraries • Create the new document filled out with data entered Content assembly
- 35. Content services Demo ContentAssembly
- 36. Content services TaxonomyTagging
- 37. Taxonomy tagging givesyou the ability to automatically tag documents in SharePoint libraries with terms configured in your term store using AI. • Uses managed metadata column • No training necessary • .doc, .docx, .pdf, and .pptx • Only existing terms Taxonomy tagging
- 38.
- 39. Setting up pay-as-you-go Documentprocessing services billed on a pay-as-you-go basis with an Azure subscription and billed based on the type and number of transactions.
- 40. Setting up pay-as-you-go Documentprocessing services billed on a pay-as-you-go basis with an Azure subscription and billed based on the type and number of transactions. • Can be viewed via Azure cost analysis
- 41.
- 42. Knowledge Agent: AI-readycontent starts with SharePoint Your intelligent content curator, backed by enterprise-grade security & compliance Improve AI responses Make every Copilot response smarter with structured, enriched, and context-aware SharePoint content. Drive Business Processes Streamline workflows and automate tasks at scale using natural language. Keep Content Fresh Proactively monitor and improve content quality across your intranet.
- 43. Improve AI responses GivesAI the context it needs – without effort Knowledge Agent suggests metadata for SharePoint files, which Copilot and agents can use to better distinguish documents and improve response quality. Makes organized content immediately useful By suggesting autofill columns based on content and user input, Knowledge Agent ensures consistent organization and better discoverability – no more spending hours tagging. Knowledge Agent enriches a content library by adding autofill columns and populating it with metadata. Enriching content to make it “Copilot-ready”
- 44. Drive business processes Simplifiesautomation for everyone Users can describe what actions they need in natural language (e.g., “notify me when…” and the agent builds a workflow. Get insights instantly Ask the agent anything about site content without digging through files. Auto-filled metadata enables precise, grounded responses. Knowledge Agent answers questions about content, using metadata to provide precise responses. Drive intuitive, efficient site management & business processes
- 45. Keeps content fresh Intelligentsuggestions to keep content current The agent analyzes search behavior to detect content gaps on SharePoint sites, as well as flags broken links and stale pages in just a few clicks. Easy page creation Add pages with a prompt or template and get smart content suggestions from the agent as you build. Knowledge Agent flags pages that users haven’t viewed in a specified period – the site editor can quickly retire them. Continuous optimization of content
- 46.
- 47. Knowledge Agent capabilities Deliversa unified agent experience of SharePoint’s AI tools Works continuously in the background when content is added or modified Content management skills: • Organize files in a SharePoint document library with suggested columns and metadata • Automate workflows in a library • Create views in a library 1 Site management skills: • Fix broken links • Find content gaps • Retire inactive pages 2 3 Content creation skills: • Create pages from your files • Create sections based on your content that fit naturally into your page • Create an FAQ Content consumption skills: • Context aware Q&A • Summarization • File comparison • Audio overview • Generate FAQ • Metadata query support (coming next month) 4 Agent discoverability skills: • Help users discover and access custom agents on the site 5
- 48. Content management skills:Organize files in a library AI suggests metadata columns by analyzing the first 20 files in a document library. Users review and refine columns via a chat panel—edit, rename, or remove as needed. Custom instructions allow users to define new columns with specific data capture rules. New uploads are auto-tagged using saved prompts and column configurations. Supports common file and column types, with limits on file size, column count, and unsupported formats. Build a richer, more intelligent document library for streamlined content management
- 49. Content management skills:Create views & automate workflows Create automation rules using plain language—no coding needed; actions include email, move, copy, and set value. Launch “Set up rules” from the document library to configure workflows via the chat panel. Customize rule conditions and actions, and the agent generates and applies them automatically. Design custom views by describing filters, sorting, grouping, or column visibility (e.g., “Show overdue invoices grouped by vendor.”). Views and rules are saved to the library, run automatically, and are visible to all users (views are public). Make your content immediate useful with by setting up AI-powered views and rules
- 50. Site management skills:Improve your site Retire inactive pages by identifying content that hasn’t been viewed recently and deprioritizing it in search, Copilot, and agents. Add banners to retired pages to notify visitors that the content is no longer maintained. Detect content gaps using search behavior —see what users are looking for but not finding. Fill gaps with AI-generated suggestions, either by creating new pages or expanding existing ones using prompts or templates. Fix broken links that users experience and redirecting them; smart redirection suggestions are coming soon. Keep your SharePoint clean, current, and confidence-inspiring
- 51. Content creation skills:Create a page, section, or FAQ with AI Create full pages with AI using natural language prompts or templates—attach files for more context and hit “Create.” Add new sections by clicking the sparkle icon between sections; choose from AI- suggested content and preview design options. Insert an FAQ web part to build out common questions and answers on your page’s topic. Collaborate with the agent as it analyzes your content and suggests layouts that match your page’s style. Transform basic pages into rich experiences with multi-turn AI support for structure, design, and clarity. Take your pages from functional to exceptional
- 52. Content consumption skills:Ask a question, etc. Ask context-aware questions—get answers based on page content, metadata, and related files. Summarize documents with AI to quickly extract key points and insights. Compare files side-by-side to highlight differences in content, structure, or metadata. Generate audio overviews of page content for quick listening or accessibility. Get a grounded, precise answer on your content – or a quick summary or comparison
- 53. Agent discovery skills:Agents on this site Approve custom agents to offer site visitors tailored expertise as an alternative to the default site agent, as a site owner. Quickly explore and switch between agents, depending on your knowledge needs, as a site member.
- 54. Why Metadata matters forCopilot & agents
- 55. Knowledge Agent’s metadata: Thefoundation for smarter Copilot & agent experiences Agents and Copilot now use file metadata (tags, categories, dates, etc.) to filter and present results, not just keywords in the content. This means users get answers that are more precise, context-rich, and tailored to their actual business needs. Knowledge Agent answers questions about content, using metadata to provide precise responses.
- 56. Knowledge Agent’s metadata: Thefoundation for smarter Copilot & agent experiences Imagine Sage, a product supplier, maintains a library of product spec sheets. Files in a document library.
- 57. Knowledge Agent’s metadata: Thefoundation for smarter Copilot & agent experiences Without metadata – and without metadata understanding – AI doesn’t understand the request for “west coast,” as this is not information included in any of the files. If Sage wants to confirm for a customer which products are sold in a specific region, she might ask:
- 58. Knowledge Agent’s metadata: Thefoundation for smarter Copilot & agent experiences But with metadata, the same search can not only organize the results by the product iteration and other product metadata, Copilot and agents semantically understands that “west coast” refers to the states California, Oregon, and Washington. Sage’s document library, automatically populated with metadata by Knowledge Agent.
- 59. Knowledge Agent’s metadata: Thefoundation for smarter Copilot & agent experiences So, when Sage asks Copilot or an agent to share specific detail about her data, they do so more accurately than they ever have before. Copilot and agents deliver more accurate results when reasoning over metadata.
- 60.
- 61. Key advantages Increase search& prompt accuracy • Built on metadata Enable auto-classification on file upload • Scan & capture Decentralize model creation • Enable users to create their own models Enhanced document creation • Use metadata filled from model extraction
- 62. Challenges with SharePointDocument Processing • It is built on pay-as-you-go so it is good to have scenarios in mind - $ • End user training is required • Not a fit for all unstructured data • Trusting the results of AI • Not knowing where to start within your organization
- 63. Content AI FeatureValue Breakdown Different features will provide more value but take more time Feature Scenario Value Complexity Knowledge Agent Classify, discovery, enhance, and more within your sites and libraries HIGH Doc Processing Classify document, extract metadata, and apply labels VERY HIGH Autofill Columns Automatically populate metadata columns using LLM HIGH Image tagging Organize large libraries of images by tagging objects or themes detected by AI MEDIUM Taxonomy Tagging Apply predefined managed metadata taxonomy terms detected by AI MEDIUM Content Assembly Generating standard documents from templates with dynamic content fields MEDIUM OCR Extract text from scanned PDF or images to make them searchable LOW Doc Translation Automatically translate documents to support multilingual collaboration LOW Annotations Enable commenting, mark up, or highlighting directly in file preview NONE PDF Merge and Extract Combining multiple PDFs into one document or extracting specific pages NONE Content Query Enhanced search filtering within a SharePoint document library NONE
- 64. Track data from invoiceswith form processing Track information from contracts Capture information from policies Gather data from product worksheets Automate order processing Simplify Visa renewals Example scenarios
- 65. Understand capabilities • Addthe Syntex free trial to a tenant • Identify/assign technical lead • Review resources for training • Utilize built in examples & tours Assess your platform readiness • Analyze SharePoint Online usage • Understand taxonomy & content types • Run the M365 assessment tool • Review content migration & search integration Discover use cases with the business • Utilize surveys and questionnaires • Analyze existing content and processes • Perform a knowledge workshop • Focus on business value and ROI Develop strategic ownership/budget • Establish sponsor of knowledge program • Secure resourcing & licensing • Establish tactical governance for knowledge • Assign organizational roles Perform pilot • Set clear pilot goals • Decide on pilot period length • Choose testing group • Perform onboarding and comms for pilot Build deployment plan • Utilize learnings from pilot • Establish a deployment team • Document implementation architecture • Build governance and training Preparing for automation
- 66. Content services Demo M365Assessment Tool
- 67. Get your FREESharePoint —Content AI, Solutions, & Admin guide from 9 Microsoft MVPs!
- 68. Search for inthe App Store or Google Play How was the session? Fill out the Session Surveys in the TechCon 365 Dallas Event App and be eligible to WIN PRIZES!
- 69. Questions ? Email: drew.madelung@protiviti.com Twitter: @dmadelung Website:drewmadelung.com Slides: http://bit.ly/DrewSlides
- 70. Discover , Assemble, and Gain Insights intoyour Content with SharePoint Content AI TechCon365 Dallas 2025
Editor's Notes
- #23 Doc understanding Form processing Freeform
- #25 In summary we provide two types of model building experiences designed for different types of content you want to classify and extract information from. They are optimized for different file formats and layouts. The document understanding models created in the content center can be applied to multiple libraries, whereas the form understanding models are currently created for one. Multiple models can be applied to a single document library Models use content types – so these constructs and their schemas can be used for content retrieval, compliance file plans, and as triggers for business process flows built using Power Automate.
- #26 Supported file types This model supports the following file types: .csv, .doc, .docx, .eml, .heic, .heif, .htm, .html, .jpeg, .jpg, .md, .msg, .pdf, .png, .ppt, .pptx, .rtf, .tif, .tiff, .txt, .xls, and .xlsx.Supported languages This model supports all of the Latin-based languages, including: English, French, German, Italian, and Spanish.OCR considerations This model uses optical character recognition (OCR) technology to scan .pdf files, image files, and .tiff files. OCR processing works best on documents that meet the following requirements: - File format of .jpg, .png, or .pdf (text or scanned). Text-embedded .pdf files are better, because there won't be any errors in character extraction and location. - If your .pdf files are password-locked, you must remove the lock before submitting them. - The combined file size of the documents used for training per collection must not exceed 50 MB, and PDF documents shouldn't have more than 500 pages. - For images, dimensions must be between 50 x 50 and 10,000 x 10,000 pixels. Images that are very wide or have odd dimensions (for example, floor plans) might get truncated in the OCR process and lose accuracy. - For .pdf files, dimensions must be at most 11 x 17 inches, corresponding to Legal or A3 paper sizes and smaller. - If scanned from paper documents, scans should be high-quality images. - Must use the Latin alphabet (English characters). Note the following differences about Microsoft Office text-based files and OCR-scanned files (.pdf, image, or .tiff): - All files: Truncated at 64,000 characters (in training and when run against files in a document library). - OCR-scanned files: There's a 500-page limit. Only PDF and image file types are processed by OCR.Multi-Geo environments When setting up Syntex in a Microsoft 365 Multi-Geo environment, you can only configure it to use the model type in the central location. If you want to use this model type in a satellite location, contact Microsoft support.Multi-model libraries If two or more trained models are applied to the same library, the file is classified using the model that has the highest average confidence score. The extracted entities will be from the applied model only.
- #31 Use this slide as reference for additional features offered within SharePoint Syntex as it relates to the content center. Recommendation: Need a talk track here to help explain and offer specifics around what technical architects should focus on here.
- #32 Doc understanding Form processing Freeform
- #34 Use this slide as reference for additional features offered within SharePoint Syntex as it relates to the content center. Recommendation: Need a talk track here to help explain and offer specifics around what technical architects should focus on here.
- #35 Content assembly
- #38 Content assembly
- #42 Knowledge Agent acts as your intelligent content curator, backed by enterprise-grade security and compliance. It improves AI responses by enriching SharePoint content, streamlines business processes with automation, and keeps content fresh through proactive monitoring and quality improvements.
- #43 Knowledge Agent gives AI the context it needs—without extra effort. It suggests metadata for SharePoint files, helping Copilot and agents distinguish documents and improve response quality. By autofilling columns based on content and user input, it ensures consistent organization and discoverability, saving hours of manual tagging.
- #44 Knowledge Agent simplifies automation for everyone. Users can describe actions in natural language—like 'notify me when...'—and the agent builds workflows. You can ask the agent anything about site content and get instant, precise answers thanks to auto-filled metadata. This enables intuitive site management and supports efficient business processes.
- #45 Knowledge Agent continuously analyzes search behavior to detect content gaps, flag broken links, and identify stale pages—all in just a few clicks. It makes page creation easy, offering smart suggestions as you build. The agent also flags pages that haven’t been viewed recently, so editors can quickly retire outdated content and keep sites fresh.
- #46 Let’s review the core capabilities of Knowledge Agent. From content management and site optimization to intelligent creation and consumption, these features are designed to streamline your workflow and maximize the value of your SharePoint investment.
- #47 Knowledge Agent delivers a unified experience for SharePoint’s AI tools, working continuously in the background. It organizes files with suggested metadata, automates workflows, creates views, fixes broken links, finds content gaps, retires inactive pages, and supports content creation and consumption with advanced skills like summarization and file comparison. It also surfaces relevant custom agents as well.
- #48 Let’s dive into how users access the 'Organize Files' capability with Knowledge Agent in SharePoint. If you’re a library owner, simply select the floating button—the persistent action button—in the lower right corner of your SharePoint document library. From there, choose 'Organize this library.' The agent immediately begins analyzing the first 20 files in your library and suggests metadata columns tailored to your content. You’ll see these suggestions in a chat panel, where you can review, edit, rename, or remove columns as needed. If you have specific requirements, you can enter custom instructions to define new columns and data capture rules. The agent applies these rules to test files, making the process interactive and flexible. Here’s the best part: once columns and prompts are set, any new files uploaded to the library are automatically tagged by the agent—no manual effort required. This automation ensures your content is consistently organized and enriched with metadata. Why does this matter? Content with rich metadata is the foundation for smarter Copilot responses. When your SharePoint files are well-organized and tagged, Copilot and other agents can deliver more precise, context-aware answers—making your business processes faster and your insights more reliable.
- #49 To access the 'Create Views & Automate Workflows’ skills, start by selecting the floating action button in your SharePoint document library—this is available to library owners. From the menu, choose 'Set up rules' or 'Create new view.' The agent guides you through the process using a chat panel, where you can describe your intent in plain language, such as 'Email me when a contract is approved' or 'Show overdue invoices grouped by vendor.’ The agent automatically generates rules and custom views based on your instructions—no coding required. You can review, customize, and confirm rule conditions and actions directly in the chat panel. The agent can also create new metadata columns if needed to support your workflow logic. Once rules and views are set, they’re saved to the library and run automatically whenever conditions are met, such as when files are added or modified. You don’t have to manually trigger these automations—the agent handles it for you, ensuring your content stays organized and your workflows are efficient.
- #50 As a site owner or editor, you access site improvement features by clicking the floating action button in your SharePoint site. Select 'Improve this site' to launch the agent’s capabilities. The agent automatically analyzes user behavior to identify pages that haven’t been viewed recently, allowing you to retire outdated content and add banners to notify visitors. It detects content gaps by monitoring what users search for but don’t find, and suggests new pages or expansions using AI prompts or templates. Broken links are flagged and can be fixed with a click—no manual audits required. All these actions happen automatically and proactively, keeping your SharePoint clean, current, and trustworthy. Remember: well-maintained, metadata-rich sites enable Copilot and other agents to deliver more accurate, context-aware answers for your users.
- #51 To create content with Knowledge Agent, start by clicking the floating action button and selecting 'Create a page’. You can use natural language prompts or templates, and even attach files for more context. The agent analyzes your input and instantly generates a structured draft, complete with images and suggested layouts. Adding new sections is easy—just click the sparkle icon between sections for AI-driven recommendations. You can also insert an FAQ web part to quickly build out common questions and answers. The agent acts as a co-author, collaborating with you to refine content and design, transforming basic pages into rich, engaging experiences. This automation ensures your content is well-organized and tagged, making it easier for Copilot to surface relevant, precise information.
- #52 Knowledge Agent makes consuming content smarter and more interactive. From any SharePoint page, click the floating action button and select 'Ask a question' to get context-aware answers based on page content, metadata, and related files. You can also select options to summarize documents, compare files side-by-side, or generate audio overviews for quick listening and accessibility. Because the agent leverages metadata and structured content, Copilot and other agents can provide grounded, precise answers—whether you need a summary, a comparison, or a quick fact. This capability transforms SharePoint from a static repository into a dynamic, AI-powered knowledge platform.
- #53 Knowledge Agent makes consuming content smarter and more interactive. From any SharePoint page, click the floating action button and select 'Ask a question' to get context-aware answers based on page content, metadata, and related files. You can also select options to summarize documents, compare files side-by-side, or generate audio overviews for quick listening and accessibility. Because the agent leverages metadata and structured content, Copilot and other agents can provide grounded, precise answers—whether you need a summary, a comparison, or a quick fact. This capability transforms SharePoint from a static repository into a dynamic, AI-powered knowledge platform.
- #54 Metadata is the foundation for smarter Copilot and agent experiences in SharePoint. When files are tagged with metadata—like categories, dates, and custom labels—agents and Copilot can filter and present results based on business context, not just keywords. This means users get answers that are more precise, relevant, and tailored to their actual needs. The more structured your content, the better Copilot can serve you.
- #55 Agents and Copilot now use file metadata to deliver context-rich responses. Instead of relying only on keyword matching, Copilot translates user requests into queries that leverage metadata—such as tags, categories, and dates. This enables Copilot and agents to retrieve files based on business meaning, not just text inside documents. Knowledge Agent answers questions about content by using metadata, providing precise, grounded responses. Structured metadata is what makes Copilot truly intelligent and business-ready.
- #56 Let’s look at a practical example: Imagine Sage, a product supplier, maintains a library of product spec sheets with metadata about product type, version, and sales region.
- #57 When files are organized with metadata, Copilot and agents can instantly filter and group results—making it easy to answer business questions like, 'Which products are sold in my region?' Files in a document library, automatically tagged by Knowledge Agent, become immediately useful for Copilot and agents.
- #58 With metadata, Copilot and agents can semantically understand business terms. For example, when Sage asks about 'west coast' products, Copilot knows this refers to California, Oregon, and Washington—because metadata encodes that relationship. The agent organizes results by product iteration and other metadata, making answers more meaningful and business-specific. Knowledge Agent automatically populates document libraries with metadata, powering smarter, more accurate Copilot responses
- #59 When Sage asks Copilot or an agent for specific details, metadata enables the agent to deliver more accurate results than ever before. Copilot and agents can reason over metadata, not just text, to provide answers that are contextually correct and tailored to business needs. The key takeaway: Copilot and agents deliver better results when your content is organized and enriched with metadata—thanks to Knowledge Agent.
- #66 Syntex assessment