Azure AI Bot Service

Is this your product? Claim it free

Hidden from Google until claimed. Claimed listings appear in Google, carry your own description, and include 100 free custom questionnaires. The first 10 products also get a founding ranking slot with 100 review codes.

Start an assessment

TrustMark reading

Where this product sits

Azure AI Bot Service is classified as Enterprise AI chatbots. Enterprise AI chatbots are organisation-deployed conversational products: ChatGPT Enterprise, Claude for Work, internal knowledge chat such as Glean, and bot platforms such as Dialogflow CX or watsonx Assistant. Buyers are IT and procurement teams who need identity, retention, connectors and admin control. Consumer ChatGPT or Claude apps, and copilots embedded inside Office, SAP or Salesforce, have different category pages.

Buyers typically come to Enterprise AI chatbots because staff paste internal knowledge into consumer chat apps that lack SSO, retention control and an admin allow-list. Classification is not a finding that this product solves that job.

This is TrustMark's classification reading for buyers comparing Enterprise AI chatbots products. It is not an Assurance Score, certificate, or ranked position.

What it is

Product classification

AI Platform / Infrastructure

Azure AI Bot Service is Microsoft's managed service for building, connecting and deploying enterprise conversational bots. It works alongside Microsoft Copilot Studio and the Bot Framework so teams can publish bots to channels such as websites, apps and Microsoft Teams, with Azure security and pay-as-you-go pricing.

Enterprise AI chatbots · General-purpose AI

What it does

How this product is used

Capabilities

Not specified

Sectors

Cross-industry / not specified

Deployment

Not specified

Autonomy

ASSISTIVE

Data exposure

Not specified

Category context

Why buyers look at Enterprise AI chatbots

Enterprise AI chatbots are organisation-deployed conversational products: ChatGPT Enterprise, Claude for Work, internal knowledge chat such as Glean, and bot platforms such as Dialogflow CX or watsonx Assistant. Buyers are IT and procurement teams who need identity, retention, connectors and admin control. Consumer ChatGPT or Claude apps, and copilots embedded inside Office, SAP or Salesforce, have different category pages.

Staff paste internal knowledge into consumer chat apps that lack SSO, retention control and an admin allow-list.

This is category context, not an independent assessment of Azure AI Bot Service. Vendor-claimed differences are listed above. Assessment results appear only after examination.

Browse Enterprise AI chatbots →

Where to get more information

• Product website · azure.microsoft.com/en-us/products/ai-services/ai-bot-service
• Microsoft website
• Company listing
• Ask to be invited to review
• Claim this listing to add a supplier statement

Other products from Microsoft

• Azure AI Content Safety Claim

  Azure AI Content Safety, now presented as Content Safety in Foundry Control Plane, is Microsoft's content moderation service for generative AI applications. It detects harmful or inappropriate text and images and provides guardrails for AI apps, and can also monitor user-generated content on platforms.

• Azure AI Face Claim

  Azure AI Face is Microsoft's face detection and recognition API, offered alongside Azure Vision in Foundry Tools. It detects and analyses faces in images and supports identity scenarios such as face verification, with pay-as-you-go transaction pricing and Microsoft's responsible AI controls on access to recognition features.

• Azure AI Language Claim

  Azure Language in Foundry Tools is Microsoft's cloud natural language processing service. It provides features such as sentiment analysis, entity and key phrase extraction, PII detection, summarisation, conversational language understanding and question answering, available through Microsoft Foundry, REST APIs and client libraries.

• Azure AI Speech Claim

  Azure Speech in Foundry Tools, formerly Azure AI Speech, is Microsoft's speech AI service. It provides prebuilt and customisable multilingual models for speech-to-text, text-to-speech, speech translation, custom voices and avatars, plus the Voice Live API for real-time voice agents, deployable in the cloud or at the edge.

• Azure AI Speech to Text Claim

  Speech to text in Azure Speech in Foundry Tools is Microsoft's transcription capability for converting spoken audio into text. It supports real-time and batch transcription across many languages, custom speech models tuned to domain vocabulary, and real-time speech translation of audio streams for apps and voice agents.

• Azure AI Translator Claim

  Azure Translator in Foundry Tools is Microsoft's cloud neural machine translation service. It translates text and documents across many languages and dialects through REST APIs, supports custom translation models for domain terminology, and powers translation in a range of Microsoft products.

• Azure AI Video Indexer Claim

  Azure AI Video Indexer is Microsoft's video and audio analytics service. It applies AI models to stored or edge video to extract insights such as spoken words, on-screen text, faces and speakers, making media libraries searchable and easier to reuse, with cloud and edge deployment options.

• Azure AI Vision Claim

  Azure Vision in Foundry Tools, formerly Azure AI Vision, is Microsoft's computer vision service. It provides image analysis, optical character recognition (OCR), spatial analysis and related models through APIs, so developers can add visual understanding to applications and agents with pay-as-you-go pricing.

• Azure Content Understanding Claim

  Azure Content Understanding in Foundry Tools is a Microsoft service that uses generative AI to process documents, images, video and audio and extract structured fields into user-defined outputs. It runs on a Microsoft Foundry resource, with Content Understanding Studio for advanced configuration, to automate content-processing workflows.

• Azure Digital Twins Claim

  Azure Digital Twins is Microsoft's IoT platform for creating live digital models of physical environments such as buildings, factories, energy networks and cities. It models relationships between people, places and devices, connects live IoT data and supports analytics that help organisations track and optimise operations.

• Azure Machine Learning Claim

  Azure Machine Learning is Microsoft's enterprise service for the end-to-end machine learning lifecycle. Data scientists and engineers use it to prepare data, train, evaluate, deploy and monitor models, with MLOps tooling, responsible AI dashboards, automated ML and a browser-based studio alongside Python SDK and CLI access.

• Azure Machine Learning AutoML for image object detection Claim

  Automated ML (AutoML) in Azure Machine Learning automatically trains and tunes models from the data and criteria a user provides, including computer vision tasks such as image classification and object detection. It runs many pipelines in parallel to compare algorithms and parameters, via Python SDK, CLI or Azure ML studio.

• Azure Machine Learning Designer Claim

  Designer in Azure Machine Learning studio is Microsoft's drag-and-drop interface for building machine learning pipelines. Users assemble data preparation, training and scoring components visually, using either classic prebuilt components or custom v2 components, then run and reuse pipelines within Azure Machine Learning workspaces.

• Azure Managed Feature Store Claim

  Managed feature store in Azure Machine Learning lets data teams develop features once and reuse them in production. Users supply a feature set specification and the service handles feature materialisation, serving, security and monitoring, removing the need to build and maintain separate feature engineering pipelines.

• Azure Speech in Foundry Tools Claim

  Azure Speech in Foundry Tools gives developers multilingual speech models inside Microsoft Foundry for building voice-enabled apps and agents. It covers recognition, synthesis with neural and custom voices, speech translation, text-to-speech avatars and the Voice Live API for low-latency speech-to-speech conversations.

• Azure SRE Agent Claim

  Azure SRE Agent is Microsoft's AI agent for site reliability engineering on Azure. It diagnoses incidents, orchestrates mitigations and helps optimise cloud operations to reduce operational toil and improve uptime, integrating with DevOps workflows. Microsoft offers a 30-day trial without always-on charges.

• Copilot in Excel Claim

  Copilot in Excel is the AI assistant inside Microsoft Excel that helps build and edit workbooks, for jobs such as updating budgets, building financial models and analysing data. It uses Excel's own tools, including tables, charts, PivotTables and formulas, so results stay editable. Microsoft says it can edit worksheets, generate data and formulas across sheets, create charts, PivotTables and dashboards, answer questions about data with trends and outliers, sort, filter and highlight, pull data from other workbooks and from the web with citations, and follow custom skills for repeatable tasks. It has three modes: edit mode changes the workbook directly, plan mode proposes a plan to confirm first, and chat mode gives insights without changing anything. It runs in Excel on Windows, Mac, the web, iPad, iPhone and Android, and needs an eligible Copilot licence.

• Copilot in Power BI Claim

  Copilot in Power BI is Microsoft's generative AI assistant inside Power BI Desktop and the Power BI service. Business users can chat with their data in three ways: a Copilot pane beside an open report, a full-screen standalone Copilot that finds and answers questions about any report or semantic model they can access (in preview), and Copilot scoped to a Power BI app, which can surface verified answers from the app's authors (also in preview). Microsoft says it can summarise reports, answer questions about data, create visuals and report pages, write DAX queries and help improve semantic models. When a question concerns the semantic model, Copilot answers from that model; otherwise it uses the language model's general knowledge. It needs paid Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher) in a supported region, with the admin setting switched on. Trial and free capacities and sovereign clouds are not supported.

• Copilot Search in Bing Claim

  Microsoft Corporation sells Windows, Azure, Microsoft 365 and Copilot AI products. Public pages cover cloud, productivity and developer APIs used by enterprises and consumers. Copilot Search in Bing is Microsoft's software product, from Microsoft Corporation. Published at bing.com/copilotsearch. Directory filing: AI search engines. Vendor legal name on this product record: Microsoft Corporation. Copilot Search in Bing is Microsoft's software product, from Microsoft Corporation. Published at bing.com/copilotsearch. Directory filing: AI search engines.

• Fairlearn Claim

  Fairlearn is an open-source, community-driven Python project that helps data scientists assess and improve the fairness of AI systems. The toolkit provides fairness metrics for comparing how a model performs across groups of people, plus mitigation algorithms for reducing the disparities it finds. Alongside the library, the project publishes guides and worked use cases, such as assessing fairness issues in credit-card default models, and treats fairness as sociotechnical: both societal context and technical choices shape who might be harmed. It installs from PyPI with pip, has API documentation and a contributor guide, and its community talks on Discord.

• FLAML Claim

  FLAML is a lightweight, open-source Python library from Microsoft for fast, low-cost automated machine learning and tuning. For common tasks such as classification and regression it finds good models for the user's own data using little compute, and it handles model selection and hyperparameter optimisation under resource limits. It also tunes other workflows, including inference settings for foundation models, MLOps and LLMOps pipelines and software configurations, and copes with large search spaces, uneven evaluation costs, constraints and early stopping. FLAML is built on research from Microsoft Research and collaborators including Penn State, integrates with MLflow, and powers AutoML and hyperparameter tuning in Microsoft Fabric Data Science. The former AutoGen module now lives in its own project.

• GitHub Copilot for Azure Claim

  GitHub Copilot for Azure is a GitHub Copilot extension for Visual Studio Code that lets developers use natural language to learn about Azure services, deploy resources, query existing resources and diagnose or troubleshoot problems, without leaving the editor.

• InterpretML Claim

  InterpretML is an open-source, community-driven Python package that brings machine learning interpretability techniques together under one API. Teams can train glass-box models that are interpretable by design, or explain black-box systems they already run, looking at a model's overall behaviour or the reasons behind individual predictions. Its headline model is the Explainable Boosting Machine (EBM), developed at Microsoft Research, which uses bagging, gradient boosting and automatic interaction detection on top of generalised additive models; the project's published benchmarks show EBMs matching random forests and gradient-boosted trees on accuracy while giving exact explanations that domain experts can edit. Uses include model debugging, feature engineering, spotting fairness issues and regulatory compliance. It installs with pip or conda on Python 3.10 and later, on Linux, Mac and Windows, with interactive visualisations and what-if analysis.

• Kosmos-2.5 Claim

  Kosmos-2.5 is a multimodal literate model from Microsoft's Document AI research for machine reading of text-heavy images such as receipts, documents and screenshots. Pre-trained on large volumes of text-intensive images, it handles two related transcription tasks: producing text blocks with their position coordinates in the image, and turning an image into structured Markdown that keeps headings, styles and layout. Both come from one shared decoder-only Transformer driven by task prompts, and it can be fine-tuned for other text-rich image tasks. The weights are published on Hugging Face under the MIT licence and run in Hugging Face Transformers 4.56 and later, with the research described in the paper Kosmos-2.5: A Multimodal Literate Model. Microsoft notes that as a generative model it can hallucinate and does not guarantee accurate OCR or Markdown output.

• Microsoft 365 Copilot Claim

  Microsoft 365 Copilot, which Microsoft now sells to businesses as Microsoft Copilot, is the AI assistant built into Word, Excel, PowerPoint, Outlook and Teams. Microsoft says it drafts content, summarises information, recaps meetings, catches up on email and helps find answers grounded in a company's own work data, with a choice of large language models. Cowork takes on multi-step work from a goal through to a finished result, Copilot Notebooks keeps project content in one workspace with audio summaries, and Autopilot runs ongoing work toward a set role and goal. Pre-built agents cover research and data analysis, and teams can build their own with Copilot Studio. For small business, Microsoft lists Microsoft 365 Business Premium with Copilot at $32.00 per user per month on a yearly plan, with usage-based billing for some advanced AI features.

• Microsoft 365 Copilot Scheduling Assistant Claim

  The Microsoft 365 Copilot scheduling assistant is the AI help for arranging meetings in Outlook. Outlook's own Scheduling Assistant shows attendees' free and busy times and suggests rooms and resources, and Copilot in Outlook builds on it. Through Copilot chat, a user can ask for something like a 30-minute Teams meeting with a colleague about a report, and Copilot checks every attendee's calendar and recommends a time that suits everyone. Microsoft says Copilot can also draft meeting invitations and agendas, summarise an email thread to give a meeting its context, suggest times based on availability, time zones and working hours, and manage RSVPs. Chat-assisted scheduling works in Outlook for Microsoft 365, and the Copilot features need a Microsoft 365 Copilot licence on a work or school account.

• Microsoft Agent Framework Claim

  Microsoft Agent Framework is Microsoft's open-source framework for building production AI agents and multi-agent workflows in Python and .NET, with a separate Go SDK. Microsoft calls it the enterprise-ready successor to AutoGen, built for teams taking agents from prototype to production. It supports graph-based orchestration patterns such as sequential, concurrent, handoff and group collaboration, with checkpointing, streaming, human-in-the-loop and time-travel. Other features include middleware for request and response processing, built-in OpenTelemetry tracing, agents defined in YAML, agent skills, a developer UI for testing, and hosting on Microsoft Foundry. It works with several model providers, including Microsoft Foundry and Azure OpenAI.

• Microsoft Copilot Claim

  Microsoft Copilot is Microsoft's AI assistant, which Microsoft describes as AI built for work. It connects to a user's work content and the apps they already use, and brings together several leading AI models, with an Auto setting that picks the model and effort for each request by weighing accuracy, speed and cost. Copilot is built into Word, Excel, PowerPoint, Outlook and Teams, and Work IQ grounds it in a company's context, people and projects. Home brings Chat and Cowork together, Copilot Code builds dashboards, trackers and apps from a description with no coding needed, and Autopilot is an always-on personal agent, currently in private preview. Agent Builder creates work agents in plain language. Copilot runs on desktop, mobile and web, and Microsoft offers plans for individuals, business, enterprise and education.

• Microsoft Copilot Studio Claim

  Microsoft Copilot Studio is Microsoft's no-code platform for creating and managing custom AI agents, workflows and apps. Teams can build chat or voice agents, add autonomous agents that plan and escalate work, and orchestrate several agents, including ones from partner providers, on complex processes. Agents can use default or custom models or the 11,000+ models in Azure AI Foundry, and connect to business data through Work IQ, 1,500+ pre-built connectors and MCP servers. Copilot Studio mixes traditional workflow automation with agent reasoning, runs automated evaluations, publishes agents internally, to websites or social channels, and supports full dev-test-prod lifecycle management with real-time analytics. Microsoft says 90% of the Fortune 500 has used it. Pricing is pay-as-you-go or a prepaid Copilot Credit commitment, and an Azure subscription is required.

• Microsoft Defender for Cloud Claim

  Microsoft Defender for Cloud is Microsoft's cloud security platform, and its AI security posture management capabilities, part of the Defender Cloud Security Posture Management (CSPM) plan, secure generative AI applications across Azure, Amazon Web Services and Google Cloud. It continuously discovers deployed AI workloads in Azure OpenAI Service, Microsoft Foundry, Azure Machine Learning, Amazon Bedrock and Google Vertex AI, and builds an AI bill of materials covering application components, data and AI artefacts from code to cloud. It scans source code and container images for vulnerable generative AI libraries such as TensorFlow, PyTorch and LangChain, flags infrastructure-as-code misconfigurations such as public AI endpoints or key-based authentication, and issues recommendations on identity, data security and internet exposure. Attack path analysis traces how grounding or fine-tuning data could be exposed, and externally reachable AI endpoints are highlighted. Microsoft notes that from 1 July 2026 agent-level discovery and posture for Foundry and third-party agents needs a Microsoft Agent 365 licence.

• Microsoft Defender XDR Claim

  Microsoft Defender XDR is Microsoft's extended detection and response suite, coordinating prevention, detection, investigation and response across endpoints, identities, email, collaboration tools and cloud apps. It joins signals from Defender for Endpoint, Defender for Office 365, Defender for Identity, Defender for Cloud Apps, Defender Vulnerability Management, Defender for Cloud, Entra ID Protection, Purview Data Loss Prevention and Insider Risk Management, and related services, so security teams see the full scope of an attack in one incident queue in the Microsoft Defender portal. Automatic attack disruption uses high-confidence correlated signals to contain in-progress attacks and limit lateral movement, and AI-driven automated remediation returns compromised devices, identities and mailboxes to a secure state. Analysts can also hunt across 30 days of raw endpoint, identity, email and cloud app data with custom queries. Defender XDR only correlates signals from the Microsoft security products an organisation has licensed, and is enabled at security.microsoft.com once licensing requirements are met.

• Microsoft Designer Claim

  Microsoft Designer is Microsoft's generative AI graphic design and photo editing app, licensed for personal, non-commercial use with a free Microsoft account. Users describe what they want in words and Designer creates images, stickers, greeting cards, invitations, social posts, monograms, collages, banners, wallpapers, icons and colouring book pages. Its editing tools can erase unwanted objects, remove or blur backgrounds, upscale images, crop, resize, add text and apply filters, and it offers design suggestions as you work. Designer runs on the web, on iOS and Android, and inside other Microsoft apps: Edit with Designer in the Windows Photos app, image creation in Word and PowerPoint through Copilot, and the Microsoft 365 Copilot app. Designs can be saved to OneDrive. Microsoft Entra work or school accounts cannot sign in to the AI editing features in Photos.

• Microsoft Fabric data agent Claim

  Fabric data agent is a generally available feature of Microsoft Fabric for building conversational question-and-answer agents over an organisation's data in OneLake. Users ask questions in plain English and the agent picks the most relevant source, then generates, validates and runs a read-only query: SQL for lakehouses and warehouses, DAX for Power BI semantic models, KQL for KQL and Eventhouse databases, plus ontologies and Microsoft Graph. Each agent can use up to five data sources, and builders tune it with instructions and up to 100 example queries per source. The agent runs under the requesting user's own permissions, respects row- and column-level security and Microsoft Purview policies such as data loss prevention, and never creates, updates or deletes data. Agents can be published to Microsoft 365 Copilot, Copilot Studio, Foundry and Teams, and managed with Git integration and deployment pipelines. It needs an F2 or higher Fabric capacity (or Power BI Premium P1 or higher), works in English only, does not read unstructured files such as PDFs, and caps answers at 25 rows and 25 columns.

• Microsoft Foundry Claim

  Microsoft Foundry, formerly Azure AI Studio, is Microsoft's enterprise platform for building, grounding and governing AI apps and agents. It brings together Foundry Models (a catalogue of more than 11,000 models from Microsoft, OpenAI, Anthropic, Meta, xAI, Hugging Face and others), Foundry Agent Service and the open-source Microsoft Agent Framework for building agents, Foundry IQ for grounding agents in organisational knowledge, Foundry Tools for tasks such as OCR, translation and speech, Azure Machine Learning, and Foundry Local for running models on-device and at the edge. Agents can act on business systems through more than 1,400 prebuilt connectors or custom tools over the Model Context Protocol. A shared control plane, integrated with Entra, Purview and Defender, handles governance, monitoring, evaluation and content-safety controls. Teams can explore it at ai.azure.com without an account, but building needs an Azure subscription, and pricing is consumption-based per service used.

• Microsoft Foundry IQ Claim

  Foundry IQ is the knowledge and retrieval layer of Microsoft Foundry, which Microsoft describes as the evolution of Azure AI Search. It gives AI agents reusable knowledge bases built over indexed or remote sources such as SharePoint, Fabric OneLake, Azure Blob Storage, the web, MCP servers, Work IQ and Fabric IQ. Its agentic retrieval plans queries, searches and iterates to find relevant context, while automated incremental indexing pipelines reduce up-front data preparation. Answers respect identity and document-level permissions through access-control list synchronisation and query-time trimming. Foundry IQ knowledge bases need an Azure AI Search resource and are set up in the Foundry portal or Azure portal, with a serverless developer tier for agent workloads. It sits alongside Work IQ (work signals) and Fabric IQ (business data) in Microsoft's intelligence layer.

• Microsoft Foundry Models Claim

  Microsoft Foundry Models is the model catalogue inside Microsoft Foundry for discovering, comparing and deploying AI models. It lists more than 10,000 models, from foundation and reasoning models to small, multimodal and industry-specific ones, with about 50 added each month. Models sold by Azure are hosted and supported by Microsoft under its Product Terms, with enterprise SLAs and Microsoft's Responsible AI review, while models from partners and the community, including Anthropic's Claude and open models from the Hugging Face hub, are supported by their providers and billed through Azure Marketplace. Models can run as serverless pay-per-token or provisioned deployments, with global, data zone or regional processing, or on dedicated managed compute. The catalogue includes benchmarks and leaderboards, fine-tuning for supported models, and Azure AI Content Safety filters by default on serverless language models.

• Microsoft Planetary Computer Claim

  The Microsoft Planetary Computer is a Microsoft platform that uses the cloud to support environmental sustainability and Earth science. It has three parts: a Data Catalog of petabytes of global environmental monitoring data, such as Landsat, Sentinel-2, NAIP and ASTER imagery, hosted on Azure in consistent, analysis-ready formats and free to use; APIs, built on the open SpatioTemporal Asset Catalog (STAC) standard, for searching that data across space and time; and applications built by partners that turn the results into information for conservation and sustainability work. The STAC API can be used anonymously, and most data can be downloaded without an account, with throttling. Many components are open source on GitHub, including the API deployment, example notebooks and a Python SDK. Microsoft describes the Planetary Computer API as being in preview.

• Microsoft Power Automate Claim

  Microsoft Corporation sells Windows, Azure, Microsoft 365 and Copilot AI products. Public pages cover cloud, productivity and developer APIs used by enterprises and consumers. Microsoft Power Automate is Microsoft's software product, from Microsoft Corporation. Published at microsoft.com. Vendor legal name on this product record: Microsoft Corporation. Microsoft Power Automate is Microsoft's software product, from Microsoft Corporation. Published at microsoft.com.

• Microsoft Purview Data Loss Prevention Claim

  Microsoft Purview Data Loss Prevention (DLP) helps organisations stop people from inappropriately sharing sensitive information such as financial data, health records and card numbers. Administrators define DLP policies in the Microsoft Purview portal, from predefined templates or custom rules built on sensitive information types and sensitivity labels, and apply them across Exchange email, SharePoint, OneDrive, Teams chats and channels, Office apps, Windows and macOS devices, on-premises file shares, Fabric and Power BI workspaces, non-Microsoft cloud apps, and Microsoft 365 Copilot (in preview). Detection uses deep content analysis, including keywords, regular expressions, validation functions and machine learning, not just simple text matching. Inline web traffic protection through Edge for Business and network integrations can stop sensitive data reaching AI apps such as ChatGPT, Gemini, DeepSeek and Copilot. When a rule matches, DLP can show a policy tip, block with or without override, or quarantine files. Policies can run in simulation mode before enforcement, with alerts flowing to Microsoft Defender XDR.

• Microsoft Purview Data Security Posture Management for AI Claim

  Microsoft Purview Data Security Posture Management (DSPM) for AI, previously known as Microsoft Purview AI Hub, is a central place in the Microsoft Purview portal for securing data used by Copilots, agents and other generative AI apps, including third-party sites such as ChatGPT and Gemini. It gives insight into AI activity across an organisation, ready-made one-click policies to detect and prevent sensitive data being shared in AI prompts, compliance controls for handling and retaining AI interactions, and guided support for AI regulations through Compliance Manager templates. A weekly data risk assessment runs automatically on the top 100 SharePoint sites to find potentially overshared content before Microsoft 365 Copilot surfaces it, with remediation through sensitivity labels, data loss prevention, restricted content discovery and retention policies; Fabric workspaces can be assessed too. Activity explorer shows prompts, responses and sensitive information types by app and user. Microsoft has now folded this classic experience into a new, broader Data Security Posture Management that covers AI apps and agents.

• Microsoft Search Claim

  Microsoft Search is the search experience built into Microsoft 365, reached from the search box in apps such as Outlook, SharePoint, OneDrive for Business and the Microsoft 365 home page. It returns people, files, sites, organisation charts and answers relevant to the app being used, and personalises results with Microsoft Graph insights while only ever showing content the user already has permission to see. Administrators can publish answers to common questions, target content to groups such as new joiners, customise result verticals, and bring outside content into the index with Microsoft 365 Copilot connectors. Microsoft says the service uses AI to work out query intent and ranks results by relevance; it does not search across tenants, and hybrid SharePoint setups can include on-premises content. Search history is personal to each user and can be downloaded or cleared from their account settings.

• Microsoft Security Copilot Claim

  Microsoft Security Copilot is Microsoft's AI for security and IT teams, built into Microsoft Defender, Entra, Intune and Purview, and integrated with Microsoft Sentinel and Defender for Cloud. Microsoft says it summarises large volumes of security signals into key insights, gives incident context and step-by-step response guidance, translates natural language into query scripts, reverse-engineers malware scripts and writes stakeholder reports. It includes ready-made agents from Microsoft for jobs such as phishing triage, vulnerability remediation and alert triage, agents built by partners, and no-code agents that teams build themselves, alongside embedded skills and promptbooks. Microsoft cites studies in which analysts using the Phishing Triage Agent found malicious emails up to 550% faster. Security Copilot agents are included with Microsoft 365 E5 and E7, and Microsoft also sells it separately.

• Microsoft SharePoint Claim

  Microsoft SharePoint is the cloud service in Microsoft 365 that organisations use to share and manage content, knowledge and applications. Teams create SharePoint sites to store documents, co-author files in real time and share them with colleagues and, where admins allow it, guests outside the organisation. SharePoint also underpins company intranets, with communication sites, team sites and hub sites that tie related sites together, and admins decide whether users can create their own sites. It is deeply integrated with Microsoft Teams: files shared in Teams are stored in SharePoint sites. Governance tools from Microsoft 365 apply to SharePoint content, including retention policies, sensitivity labels, data loss prevention and Content Search. Content can be moved in from file shares, other cloud providers or on-premises SharePoint Server with Migration Manager and the SharePoint Migration Tool, and developers can extend it with the SharePoint Framework. SharePoint works alongside OneDrive, which holds each user's personal files.

• Microsoft Teams Premium Claim

  Microsoft Teams Premium is a per-user add-on licence for organisations that already have Microsoft 365 or Office 365 and Teams licences; it does not replace a Teams licence. It adds intelligence features such as live translated captions and transcripts in 40 languages, AI-generated meeting notes and tasks, auto-generated recording chapters and personal timeline markers. Meeting protection features include watermarking, end-to-end encryption for meetings of up to 200 people, control over who can record and transcribe, detection of sensitive content during screen sharing, and Purview sensitivity labels for customers on E5-level plans. Organisations can brand meetings with custom themes, backgrounds and templates, and get advanced Bookings virtual appointments with SMS reminders (available in the UK) and the Queues app for Teams Phone call queues, which needs a Teams Phone licence. Some features apply to the organiser and are shared with attendees, while others benefit only licensed users. Since 1 April 2026 some features that were Premium-only are included in Teams Enterprise. Admins can start a 30-day trial with 25 licences, and users can start a 60-day self-service trial.

• Microsoft Translator Claim

  Microsoft Translator is Microsoft's family of translation apps and services for personal, business and education use. The Translator apps translate real-time conversations, menus and street signs (including offline), websites and documents. In classrooms it provides live captioning and cross-language conversations so students and parents can follow along in their own language. For businesses, the same technology is offered to developers as the Translator API and Speech service in Azure, which has its own listing as Azure AI Translator.

• ONNX Runtime Claim

  ONNX Runtime is Microsoft's open-source, cross-platform engine for speeding up machine learning inference and training in an existing technology stack. It runs models on CPUs, GPUs and NPUs, optimising for latency, throughput, memory use and binary size, and works on Linux, Windows, Mac, iOS, Android and in web browsers. Developers can use it from Python, C#, C++, Java, JavaScript and Rust. ONNX Runtime Web runs PyTorch and other models in the browser, ONNX Runtime Mobile brings models to Android and iOS apps, on-device training lets an app personalise a model locally, and the onnxruntime-genai package supports generative AI and large language models. Microsoft says it powers AI in Windows, Office, Azure Cognitive Services and Bing.

• Phi-4 Claim

  Phi-4 is a small open language model from Microsoft Research, released on 12 December 2024 under the MIT licence. It has 14 billion parameters in a dense decoder-only Transformer, takes text in a chat format, and has a 16,000-token context window. Microsoft trained it on 9.8 trillion tokens mixing synthetic textbook-style data, filtered public web pages, and acquired academic books and Q&A sets, with a focus on reasoning and logic. Microsoft positions it as a building block for generative AI features where memory, compute or latency are tight, mainly in English, and says developers should evaluate accuracy, safety and fairness before using it in high-risk settings. The weights are on Hugging Face, and the knowledge cut-off for public data is June 2024.

• Phi-4-mini-instruct Claim

  Phi-4-mini-instruct is a lightweight 3.8-billion-parameter open model in Microsoft's Phi-4 family, released under the MIT licence. It supports a 128,000-token context window and more than 20 languages, including English, French, German, Spanish, Arabic, Chinese and Japanese. Microsoft trained it on synthetic and filtered public web data focused on reasoning, then tuned it with supervised fine-tuning and direct preference optimisation, and says it improved instruction following and function calling over earlier Phi models. Microsoft says it suits commercial and research use in memory- or compute-limited and latency-sensitive apps that need strong maths and logic reasoning. It is available on Hugging Face, with ONNX versions, and through Azure.

• Phi-4-multimodal Claim

  Phi-4-multimodal-instruct is a lightweight open multimodal model in Microsoft's Phi-4 family that takes text, images and audio in one network and replies in text. It has 5.6 billion parameters, built on Phi-4-mini-instruct as the language backbone with added vision and speech encoders and adapters, and a 128,000-token context window. Because one model handles all three inputs, it can work across speech, pictures and text together rather than chaining a separate transcription model. Microsoft lists uses including speech recognition, speech translation and summarisation, image and chart understanding, optical character recognition, multi-image comparison, maths and logic reasoning, and function calling. Text works in more than 20 languages, audio in eight including English, French, German and Chinese, and vision in English. It was tuned with supervised fine-tuning, direct preference optimisation and RLHF, is released under the MIT licence on Hugging Face, and is also offered through Azure. Microsoft asks developers to evaluate accuracy, safety and fairness before high-risk use.

• Phi-4-reasoning Claim

  Phi-4-reasoning is an open-weight reasoning model from Microsoft Research, released on 30 April 2025 under the MIT licence. It is fine-tuned from the 14-billion-parameter Phi-4 using supervised fine-tuning on chain-of-thought examples plus reinforcement learning, with training focused on maths, science and coding. It has a 32,000-token context window, and each answer comes as a reasoning section followed by a summary. Microsoft says it is designed and tested for maths reasoning, recommends a specific system prompt and sampling settings, and asks developers to check accuracy, safety and fairness before using it for other or high-risk purposes. The weights are published on Hugging Face.

• Playwright Workspaces Claim

  Playwright Workspaces is a fully managed Azure service that runs cloud-hosted browsers for testing web applications, automating browser workflows and giving AI agents a browser to work in. Teams run existing Playwright tests at scale through the Playwright Test Runner, or connect directly to cloud browsers over the Chrome DevTools Protocol, for end-to-end, regression and cross-browser checks in CI/CD pipelines. A remote Model Context Protocol (MCP) server lets compatible AI tools and agents navigate sites, fill in forms and extract information through managed browsers. Microsoft lists session recordings, traces, logs and screenshots for debugging, a Live View with Take Control for stepping into a running session, connectivity to private and staging applications, and Microsoft Entra ID sign-in. Workspaces are regional (including West Europe and Switzerland North), and Microsoft says customer data is not stored or processed outside the workspace region.

• Presidio Claim

  Presidio is an open-source toolkit for finding and removing personal data (PII) in text, images and structured data. Its analyzer detects PII using named entity recognition, regular expressions, rule-based logic and checksums with context, in multiple languages, and can connect to external PII detection models. The anonymizer then masks, replaces or encrypts what it finds, the image redactor uses OCR to redact PII from images including DICOM medical scans, and a structured module handles tables and semi-structured data. It runs from Python or PySpark, Docker or Kubernetes. Presidio started at Microsoft and has moved to an independent, community-governed project under the Data Privacy Stack organisation on GitHub, still under the MIT licence. The project warns that automated detection cannot guarantee it finds all sensitive information.

• PyRIT Claim

  PyRIT, the Python Risk Identification Tool for generative AI, is Microsoft's open-source framework that helps security professionals and engineers red-team generative AI systems and find risks before attackers or users do. It gives teams tooling to probe models and AI applications for harmful or unintended behaviour, and is documented on its project website with install and contribution guides. Microsoft publishes a whitepaper, PyRIT: Democratizing AI Red Teaming Through Open-Source Tooling, vulnerabilities are reported privately through the Microsoft Security Response Center, and the team and community talk on Discord. The project has moved from the Azure organisation to github.com/microsoft/PyRIT.

• Responsible AI Mitigations Library Claim

  The Responsible AI Mitigations Library (the raimitigations Python package) is an open-source Microsoft library that helps data scientists and ML developers measure how balanced their datasets are across cohorts and apply targeted fixes where a model underperforms. It has three modules: DataProcessing, with mitigation techniques such as encoders, scalers and imputers for improving results on specific cohorts; DataBalanceAnalysis, with metrics for diagnosing errors caused by imbalance in labels or feature values; and Cohort, for building separate pipelines and models per cohort and combining them to meet group fairness goals. It is part of Microsoft's Responsible AI Toolbox and is designed to follow on from tools like Error Analysis in the Responsible AI Dashboard. It installs with pip on Python 3.7 to 3.10, with documentation on Read the Docs and example notebooks in the repository.

• Sales in Microsoft 365 Copilot Claim

  Sales in Microsoft 365 Copilot, formerly Microsoft 365 Copilot for Sales and also called the Sales agent, is Microsoft's AI assistant for sales teams. It brings customer and deal information from Dynamics 365 Sales or Salesforce into the Microsoft 365 tools sellers use every day, including Copilot, Outlook and Teams. Microsoft says sellers can ask about CRM data in plain language, prepare for customer meetings, draft sales emails and personalised content, capture insights and action items from Teams calls, and get AI-generated insights and next-step recommendations, while sales managers get AI-powered tools to track performance and coach their teams. A Microsoft 365 admin installs it, and users must consent before it connects their work account to the CRM. It is not available for on-premises Dynamics 365 Customer Engagement or in US government clouds, and specific licence terms apply.

• Semantic Kernel Claim

  Semantic Kernel is Microsoft's open-source, model-agnostic SDK for building, orchestrating and deploying AI agents and multi-agent systems, released under the MIT licence for Python, .NET and Java. It connects to OpenAI, Azure OpenAI, Hugging Face, NVIDIA and other model providers, or local models through Ollama, LM Studio and ONNX. Agents can use tools and plugins written as native code, prompt templates, OpenAPI specs or Model Context Protocol (MCP) servers, along with memory and planning, and it integrates with vector stores such as Azure AI Search, Elasticsearch and Chroma. Microsoft now names Microsoft Agent Framework, which reached version 1.0, as the enterprise-ready successor to Semantic Kernel and publishes a migration guide for existing users.

• Surface Laptop (8th Edition) Claim

  Surface Laptop (8th Edition) is Microsoft's clamshell Copilot+ PC line, sold in 13.8-inch and 15-inch sizes alongside a smaller 13-inch model in the same family. The 8th Edition models run Qualcomm Snapdragon X2 Plus or X2 Elite processors with a built-in neural processing unit for on-device AI features, take up to 64 GB of RAM, and use 120 Hz HDR PixelSense touchscreens. Microsoft quotes up to 20 hours (13.8-inch) and 19 hours (15-inch) of local video playback, USB4 ports, Windows Hello face sign-in, and a haptic touchpad on these sizes. On configurations with 16 GB of RAM or more, Copilot+ PC features such as describing a file in plain words to find it through Windows search are available, subject to Microsoft's language and format limits. Microsoft also publishes self-repair guidance and replacement parts for out-of-warranty repair.

• VibeVoice-1.5B Claim

  VibeVoice-1.5B is an open-source text-to-speech model from Microsoft for generating long, expressive conversations with several speakers, such as podcasts, from a script. It uses continuous acoustic and semantic speech tokenisers running at a very low 7.5 Hz frame rate, a large language model (Qwen2.5-1.5B in this release) to follow the text and dialogue flow, and a diffusion head to produce detailed audio. This release has a 64K-token context and can generate up to about 90 minutes of speech with up to four distinct voices. It is trained on English and Chinese and released under the MIT licence, with a technical report, project page and code on GitHub. Microsoft limits intended use to research, and rules out voice cloning without recorded consent, disinformation, live deepfake voice conversion and non-speech audio.

Why complete verification

Completing verification can improve the quality of the public evidence available on this product profile. Useful verified information can deepen the profile and make public questions about the named product more answerable. It does not buy directory ranking, review stars, a higher Assurance Score, a favourable certificate outcome, or a guaranteed Google ranking.

Payment remains independent from rankings and assessment outcomes. Start an assessment · Methodology