Microsoft Ignite 2023: The Biggest News In AI, Copilot

Microsoft Copilot for Service, Maia AI chips, Copilot in Dynamics 365 Guides and Copilot Studio are some of the biggest offerings unveiled at Ignite 2023.

Microsoft-designed AI accelerator chips. Holograms to help industrial workers repair equipment. And a studio experience for using data outside Microsoft offerings to power the vendor’s Copilot generative artificial intelligence offering.

These are some of the biggest offerings unveiled by the Redmond, Wash.-based tech giant during its annual Ignite conference, being held in Seattle and online this week.

The continuing expansion and access of Microsoft’s Copilot GenAI tool was a major theme of the event. Microsoft said it was flipping the name of Microsoft 365 (M365) Copilot to Copilot in M365 and appears to be following that naming convention as it rolls Copilot out to other Microsoft products and as it unveils role-specific versions of the tool.

Ignite 2023 News On AI, Copilot, Chips

Role-specific copilots include Microsoft Copilot for Service and Copilot for Sales, a rechristened Microsoft Sales Copilot.

Copilot in M365 only became generally available (GA) this month, adding ways for users to generate content with plain language queries, a technology that has excited some solution providers and their customers and created new work for partners.

The Microsoft-designed chips are the Maia for AI and the Cobalt for cloud workloads. Holograms are part of Microsoft’s Copilot in Dynamics 365 (D365) Guides offering. And Microsoft Copilot Studio promises users the ability to leverage SAP, Workday and ServiceNow data for copilots as part of more than 1,000 built-in connectors with systems of record.

See what else Microsoft revealed during Ignite 2023.

Microsoft-Designed Chips

Microsoft revealed two chips it has designed to assist with AI and cloud workloads, both of which will start to roll out early next year to Microsoft data centers.

The Microsoft Azure Maia AI accelerator chips promise to run cloud-based training and inference for OpenAI models, Bing, GitHub Copilot, ChatGPT and other AI workloads, according to the vendor.

The Maia 100 AI Accelerator will power some of the largest internal AI workloads running on Microsoft Azure, according to the vendor. Maia promises to help train more capable models and reduce the cost of those models.

The Maia 100 has 105 billion transistors, making it one of the largest chips on 5-nanometer (nm) process technology, according to Microsoft.

The Azure Cobalt is a cloud-native chip based on Arm architecture that is optimized for general-purpose workloads. The Microsoft-designed chip promises better performance, power efficiency and cost efficiency for these workloads, according to the vendor.

Cobalt 100, the first generation in the series, is a 64-bit 128-core chip that promises to deliver up to 40 percent performance improvement over current generations of Azure Arm chips, according to Microsoft. Cobalt already powers Microsoft Teams, Azure SQL and other services.

Microsoft is also already at work on second-generation versions of Maia and Cobalt, according to the vendor.

Silicon Partnerships Expansion

Microsoft teased upcoming infrastructure options for users with expanded silicon partnerships.

One of these expanded partnerships will include AMD MI300X accelerated virtual machines (VMs) added to Azure to accelerate AI workload processing, the vendor said this week.

These VMs are meant for AI model training and generative inferencing, according to Microsoft.

In preview is a new NC H100 v5 VM series built for Nvidia H100 Tensor Core GPUs. This offering promises improved performance, reliability and efficiency for midrange AI training and generative AI inferencing, according to Microsoft.

The vendor also plans an AI-optimized ND H200 v5 VM series with the upcoming Nvidia H200 Tensor Core GPU, according to Microsoft.

Coming next year to Microsoft fleets, the H200 Tensor Core GPU will support larger model inference with no reduction in latency.

Nvidia has said the H200 will become available in systems and cloud instances starting in the second quarter of next year through HGX H200 server boards in four-GPU and eight-GPU configurations.

Microsoft Copilot Studio

One of the marquee copilot announcements at Ignite 2023 was for Microsoft Copilot Studio.

Copilot Studio is a low-code service for connecting copilots to other data with prebuilt or custom plugins and generative pretrained transformers (GPTs). The tool is an end-to-end SaaS platform with no infrastructure or management.

Its out-of-the-box LLM means no explicit natural language training or custom model are needed, according to Microsoft.

This allows users to leverage SAP, Workday and ServiceNow as part of more than 1,000 built-in connectors with systems of record, according to Microsoft. Users can also build custom connectors for any data source not already built in.

Before Copilot Studio, Copilot for M365 relied on business data within Microsoft Graph, according to the vendor.

Copilot Studio also aims to allow users to create and orchestrate custom workflows for expense management, employee on-boarding, department processes and other scenarios.

Users can leverage a graphical interface or natural language with Studio to create and publish a copilot, including one that interacts with users on a website, on social media and one that aids with quarterly earnings, among other use cases.

Multiple users can comment on a copilot project like in Microsoft Office and display coding views side by side.

IT administrators have one place for governance, control and monitoring. The admin center shows customizations, custom copilots and who is building them. Admins can publish, secure data, manage environments and more from the admin center, according to Microsoft.

Copilot Studio integrates with Azure Services and Power Platform AI models. Users can access Azure AI Studio from Copilot Studio. Users can leverage Copilot Studio to link copilots with speech recognition, sentiment analysis, entity extraction and other parts of the Microsoft Conversational AI stack.

Microsoft Copilot For Service, Copilot For Sales

Entering public preview in early December and becoming GA in the first quarter of 2024 is Microsoft Copilot for Service, which aims to bring AI assistance to frontline employees, according to Microsoft.

Copilot for Service is priced at $50 per user, per month and will include Copilot for M365, which is available by itself for $30 per user, per month. Copilot for Service is part of Microsoft’s strategy for role-based Copilot offerings, including Copilot for Sales.

Copilot for Service works with existing CRM and contact center services, not just Microsoft Dynamics, according to Microsoft. Users can also train Copilot for Service on trusted public websites, SharePoint content, knowledgebase articles, offline files and more.

Future Copilot for Service capabilities include email summaries, email drafts and meeting recaps with CRM data, according to Microsoft. It will also allow viewing and updating of CRM records in Outlook and Teams and automation of case wrap-ups, meeting scheduling, adding new contacts and other tasks.

For those automations, Copilot for Service will draw on context from emails, knowledge sources and previous customer communications— similar to the Copilot in Dynamics 365 Customer Service agent assist capabilities.

Like Copilot for Service, Copilot for Sales is $50 per user, per month and includes Copilot for M365. About 15,000 organizations are already using Sales Copilot.

Other updates to Copilot for Sales include a new integrated experience with Microsoft Word and Microsoft Teams, allowing sellers to prompt Copilot to create a meeting preparation brief in Microsoft Word.

The brief is automatically populated with an account and opportunity summary, names and titles of meeting participants, open tasks, highlights from recent meetings, email threads and other customer information, according to Microsoft.

In Teams, sellers can use meeting recaps to surface action items and tasks, conversation key performance indicators (KPIs) and sales keywords.

Copilot Renaming

Along with Copilot for Sales and Copilot for M365, Microsoft also unveiled other name changes for copilot products.

Bing Chat and Bing Chat Enterprise are now simply called “Copilot.”

Copilot Pro becomes GA on Dec. 1. It’s accessible in Windows and included in Microsoft 365 E3, E5, Business Premium and Business Standard plans. On Dec. 1, Copilot Pro will become included with M365 F3. It is available as a stand-alone product for $5 a month.

Copilot Pro has commercial data protection, meaning that chat data isn’t saved, Microsoft has no access to chat data and data isn’t used to train Microsoft AI models, according to the vendor.

Copilot is now available for M365 E3 and E5 and Office 365 E3 and E5 users.

Upcoming Copilot for M365 Updates

In February, Copilot in Forms will become available for multiple users, allowing for the generation of forms, quizzes, surveys and polls based on queries and prompts.

Early next year, Microsoft will launch a private preview of Viva integration with Copilot. Copilot will have the ability to comb Viva data and apps for resources for employees, managers, HR and other roles.

Users will have the ability to set priorities and objectives and key results. They will have access to skilling and learning experiences, too, according to Microsoft.

Copilot in Viva Insights will enter private preview in January. This feature will allow natural language prompts to query Viva Insights data to generate personalized, dynamic reports.

Microsoft is also at work on copilot experiences in Viva Goals, Engage, Learning, Glint and Amplify, according to the vendor.

Coming next year, Copilot in Whiteboard promises to automatically capture and visualize spoken discussion points and organize them into a collaboration space in Whiteboard, which all meeting participants can access.

In spring, Copilot in Outlook will gain the ability to create meeting preparation summaries ahead of appointments. The copilot will comb through invitation details, related emails and pertinent documents to build a summary.

Copilot in Word will soon have a catch-up feature to see document revisions through asking Word the question in plain English.

Copilot in Excel will soon have the ability to pull data from the internet into a spreadsheet, allowing for further data analysis.

Copilot in PowerPoint will soon have an enterprise asset library, which taps into an organization’s image library for presentations. Copilot in PowerPoint will also incorporate Microsoft Designer for personalizing visuals for individual presentations.

And Copilot in the new Microsoft Planner promises to allow natural language to create plans, tasks and goals. It also should allow users to answer questions on progress, priorities, workload and more.

Copilot In D365

In December, MIcrosoft will launch a preview for more AI-powered experiences in Dynamics 365 (D365) Field Service that include the ability for frontline workers to ask questions in conversational language and get answers from within Teams.

Also in December, Microsoft will bring to North America a full-screen view in Copilot in D365 Sales that allows sellers to use natural language or prebuilt prompts for more data on customers, deals, meetings, forecasts and more. The feature is already available in Asia and Europe.

Coming in December for Copilot in D365 Customer Insights is a preview for generating a customer profile summary, according to Microsoft.

The summary includes demographic, transactional, behavioral and analytics data. The capability is available for organizations with D365 Customer Insights and D365 Sales Enterprise or Premium subscriptions.

Copilot in D365 Field Service mobile will allow users to describe progress in Copilot by voice instead of in writing. Copilot will suggest ways to check off service tasks, add notes and update product quantities, among other ways to improve data entry, according to Microsoft.

Now in preview for Copilot in D365 Sales is the ability for sellers to use plain language for contextual insight and recommendations for leads and opportunities.

Another preview is in Copilot in D365 Customer Insights and allows for marketers to qualify leads with engagement scores and other metrics to potentially increase win rates.

Now GA is an AI-assisted work order creation and technician scheduling, according to Microsoft.

Microsoft launched a public preview of financial and inventory data flow between D365 Field Service and D365 Finance and Supply Chain Management, according to the vendor.

D365 Guides

In private preview is Copilot in D365 Guides, which combines GenAI with mixed reality for industrial frontline workers.

A future public preview will come first to Microsoft’s HoloLens 2 headset, with more mobile devices added down the road.

Guides is meant to help these workers with complex equipment and processes and ramp up new technicians.

Workers can ask Copilot for relevant information through natural language as well as looking, pointing and other gestures, saving workers time they’d otherwise use to sift through digital documents and paper manuals, according to Microsoft.

Workers can point to a piece of equipment and ask for a bolt’s torque limit, for example. New workers can ask what a certain part does, giving time back to experienced workers.

Copilot’s GenAI technology searches through technical documentation, service records, training content and any other data sources curated by customers, summarizing the information and providing holograms and other content on top of the equipment needing maintenance.

Workers can ask for information on how a part fits and the location of a sensor module, for example. Along with holograms, the tool can provide answers with speech and text.

Copilot in D365 Guides allows for progressive conversations powered by Azure OpenAI Service. Workers can have the tool walk them through disassembling a product or provide the next step in a process, for example, according to Microsoft.

Guides can connect to sensors, operational data and service records, answering when items were last inspected and the history of an equipment’s concerns, for example.

Copilot Profile, Copilot In Outlook

An upcoming feature is Topics in Copilot, which promises the ability to summarize topics across multiple connected documents, sharing organizationwide intelligence with different users.

Early next year, M365 Chat integration comes to Copilot in Windows, according to the vendor.

Newly available features include profile, which gives users more control over copilot interactions. Users can request visualized data and bullet points instead of paragraphs in results and set other preferences.

Profile will come to Word and PowerPoint first, with profile “coming soon” to other M365 applications, which include Teams and Excel.

In Outlook, copilot can provide an email thread summary with actionable steps, agenda drafts, discussion summaries, adding more clarity to the original thread and suggesting meeting times, among other actions.

For IT administrators, Copilot for M365 usage report is now GA in the M365 administration center. Admins can assess an organization’s copilot readiness and send template messages to users’ Windows 11 notification centers to encourage use and educate them on copilot.

Soon, an AI assistance category in adoption score will show metrics on content summarization and creation usage.

Microsoft has also launched a public preview of Copilot Dashboard. Powered by Viva, the dashboard assesses readiness, tracks adoption and measures the effects of copilot. The dashboard shows how many employees are eligible and ready for copilot based on M365 app use.

The dashboard breaks down copilot use across apps and shows early signals on how copilot affect productivity based on meetings, chat, email and documents, according to Microsoft.

SharePoint Premium, Loop Goes GA

Microsoft launched a preview of SharePoint Premium during Ignite 2023. Services for this offering will launch between the end of 2023 and early 2024.

SharePoint Premium offers AI-powered content management and automation plus enhanced governance, according to Microsoft.

Microsoft has made its Loop application GA. The intelligent page creation feature in Copilot in Loop can suggest pages based on past work and adapt them for new projects.

Copilot in Loop can also craft a new page with content suggestions tailored to goals. The workspace status feature acts as a dashboard for tracking team tasks and flagging upcoming deadlines, according to Microsoft.

Microsoft Teams

Copilot in Teams users now have the option to use copilot without transcription. Copilot can answer questions and provide information during the meeting without a recording created.

Copilot in Teams channels is a new feature for users to shorten long posts, get action items and review decisions in commonly used channels.

Copilot compose box in Teams chats and channels allows copilot to help users write a message or adjust the message’s tone.

Copilot in Collaborative notes will take real-time notes during Teams meetings, sharing those notes with everyone in the meeting. A user can instruct Copilot to capture specific content. Telling it to quote a specific user will transcribe that user’s remark for everyone to see.

A new voice isolation feature in Teams promises to learn users’ voices and suppress other ones in the background. In Teams Premium, users can leverage GenAI to remove background clutter and add decor to physical rooms during video calls.

Azure OpenAI Service

Before the end of the year, Microsoft will make the GPT-4 Turbo model generally available in Azure OpenAI Service as part of a push for more multimodal AI capabilities in the offering. GPT-4 Turbo is in preview for now.

GPT-4 Turbo with Vision (GPT-4V) will be in preview by the end of 2023. This integration with Azure AI Vision promises more use cases by marrying images, video and text for generating text output.

During Ignite 2023, Microsoft said it has expanded its Customer Copyright Commitment (CCC)—formerly called the Copilot Copyright Commitment—to Azure OpenAI Service users.

The commitment is for Microsoft to pay for adverse judgments if users are sued for copyright infringement by using OpenAI Service or OpenAI Service-generated output.

Microsoft has published new documentation for Azure OpenAI Service users to implement technical measures that mitigate infringing content risk. Users must comply with the documentation to qualify for the commitment.

Copilot In Microsoft Fabric

In public preview is Copilot in Microsoft Fabric, which allows natural language for creating dataflows and pipelines, writing SQL statements, building reports and developing machine learning models.

The first experiences are in Power BI, Data Factory, Data Engineering and Data Science, according to Microsoft.

Microsoft will roll out the public preview in stages, giving Fabric F64 or higher users or Power BI Premium P1 or higher users access by the end of March 2024, according to the vendor.

Interested users don’t need to sign up. The preview will be available as a new setting in the Fabric Admin Portal.

When Microsoft starts charging for Copilot in Fabric, users can leverage Copilot against existing capacity, according to Microsoft.

Some examples of Copilot in Fabric include Power BI users creating reports and putting insight into narrative summaries in seconds. In Data Factory, users can describe how to ingest and transform the data using natural language. And users in Data Engineering or Data Science can do more with notebooks.

Fabric also integrates with Azure AI Studio, giving users a place to build AI services and custom models and a place to clean data.

Azure AI Studio, Azure AI Search

Microsoft has moved Azure AI Studio into public preview, giving more AI developers access to this one stop for models, retrieval-augmented generation (RAG) data integration, intelligent search capabilities, full-life-cycle model management and content safety.

In preview in Azure AI Studio is the prompt flow orchestration tool for large language model operations (LLMOps). Prompt flow is GA in Azure Machine Learning (ML). Prompt flow is meant to simplify prototyping, experimenting, iterating and deploying AI applications.

Azure AI Search—formerly known as Azure Cognitive Search—is now available in AI Studio.

And Microsoft has a preview of new Azure AI Content Safety-powered features inside AI Studio. The goal of these features is to address LLM harms and security risks by preventing attempted unauthorized modifications and identify when LLMs generate material using third-party intellectual property (IP) and content.

Azure AI Studio users can bring in data from OneLake and any data supported by shortcuts, including ADLS Gen2 and Amazon S3.

Meanwhile, Microsoft has made vector search GA in Azure AI Search, an information retrieval platform for organizations to leverage data for personalized GenAI apps.

Vector search can improve the quality of responses and latency, according to Microsoft. Users can leverage vector search for finding images, audio, text, video and other data types.

Also GA is semantic ranker, formerly known as semantic search. Now, users can leverage the same ML-powered search re-ranking technology as Bing in their own GenAI apps.

Azure AI Search can also chunk and store vector embeddings from any of these sources through OneLake.

Copilot for Azure, Azure Cosmos DB Updates

Microsoft unveiled a public preview of Copilot for Azure, which aims to help users design, operate or troubleshoot apps and infrastructure from cloud to edge, according to the vendor.

Copilot for Azure promises to help users find new cloud functionality and speed up complex tasks.

Aimed at users of Azure AI and GitHub Copilot, Microsoft has launched an Azure AI Advantage offer to encourage use of Azure Cosmos DB.

In preview for Cosmos DB is a dynamic scaling capability to help developers save on database costs.

Another preview is for Microsoft Copilot for Azure integration in Azure Cosmos DB.

Azure Cosmos DB for MongoDB vCore and vector search in Cosmos DB for MongoDB vCore are now GA. Users can now build apps with MongoDB data stored in Cosmos DB and store, index and query vector data without transferring data outside of a Cosmos DB database, according to Microsoft.

Developers also now have access to a public preview of Azure AI extension in Azure Database for PostgreSQL for building LLMsw and GenAI services.

Windows AI

Microsoft unveiled Windows AI Studio during Ignite 2023, giving developers tools and models from Azure AI Studio, Hugging Face and other catalogs to deploy small language models (SLMs) for local use on Windows applications.

In the coming weeks, Microsoft will open up access to Windows AI Studio as a Visual Studio (VS) Code extension, according to Microsoft.

The Studio has a guided workplace setup, model configuration user interface (UI) and guided walkthroughs to finetune Phi, Llama 2, Mistral and other popular SLMs, according to Microsoft.

Developers can test fine-tuned models with Prompt Flow and Gradio templates integrated into the workspace. They can run models on the cloud on Azure, on the edge locally on Windows or across the two.

Windows AI Studio will eventually have state-of-the-art (SOTA) models optimized specifically for Windows GPUs and neural processing units (NPUs), starting with Llama2-7B, Mistral-7B, Falcon-7B, and Stable Diffusion XL.

The Windows AI development platform has a new Microsoft Defender for Endpoint plug-in for event monitoring in all running distributions.

Users can now leverage Intune to control WSL security settings. And WSL gained advanced networking controls for specifying firewall rules that apply to the WSL virtual machine.

The vendor plans to launch a public preview for new AI capabilities in Windows 365 for Cloud PC. These capabilities include resizing recommendations for forecasting and making a better-informed cloud PC investment.

Security Copilot Embedded Features

Microsoft will add embedded experiences of Security Copilot across the Microsoft Security portfolio as part of its Early Access Program.

Purview now leverages Security Copilot for summarization within Data Loss Prevention (DLP), Insider Risk Management, eDiscovery, and Communication Compliance workflows.

eDiscovery now has AI translator capabilities to use natural language to define search queries.

In Entra, Security Copilot is embedded to assist in investigating identity risks and help with troubleshooting daily identity tasks, such as why a sign-in required multifactor authentication or why a user’s risk level increased.

IT Administrators use Security Copilot in Entra for instant risk summaries, steps to remediate and recommended guidance for each identity at risk in natural language.

In Entra ID Governance, administrators can use Security Copilot for guidance when creating a life-cycle workflow for user credentials and access rights.

Security Copilot in Defender for Cloud allows security administrators to identify critical concerns to resources faster and receive guided risk exploration with summaries and contextual insight such as critical vulnerabilities, sensitive data and lateral movement, according to Microsoft.

In Intune, Security Copilot can simplify device, policy and app management tasks for IT admins and security analysts, generate policies based on prompts and analyze drafts before deployment.

And the stand-alone Security Copilot offering has added signals from Defender, Defender for Cloud, Sentinel, Intune, Entra and Purview under a unified view. The stand-alone offering also has multiple embedded copilot features for those other Microsoft security offerings already available.