Secure AI innovations

Integrated Microsoft 365 E5 platform with Microsoft Security Copilot to automate and centralize incident response, authentication, and security monitoring across multiple digital tools.

AES deployed generative AI agents built on Vertex AI and Anthropic’s Claude models to overhaul its health and safety audit processes. They connected Claude via API calls within the Vertex AI environment, addressing security through configured service accounts and leveraging Vertex AI Model Garden for simplified deployment. This automation streamlined document reviews for large, multilingual audit files, significantly reducing manual labor.

Stairwell uses Anthropic's Claude to analyze and summarize complex security data, helping protect companies from sophisticated cybersecurity threats. Claude's large context window allows Stairwell to process security data with over 40,000 characters and its ability to understand technical data without extensive prompt engineering enables Stairwell to convert technical security findings into clear, actionable insights for security teams of all experience levels.

The school deployed Microsoft Reading Progress and Microsoft Immersive Reader, two AI-powered apps integrated within Microsoft Teams, to support individualized learning. The tools record reading sessions, provide real-time feedback on pronunciation and errors, and translate texts to enable comprehension across multiple languages. This setup streamlines classroom instruction, personalizes student learning, and reduces teacher workload.

Vodafone uses Azure OpenAI Service, Azure AI Foundry, Microsoft Copilot, and Azure AI Search to enhance their virtual assistant TOBi and develop SuperAgent to assist customer service agents. TOBi uses conversational AI to handle customer inquiries, while SuperAgent helps agents solve complex problems faster.

Ontada leveraged Microsoft’s Azure OpenAI Service Batch API and Azure AI Foundry to build its ON.Genuity platform, which processes 150 million unstructured oncology documents to automatically extract nearly 100 critical data elements across 39 cancer types. They integrated the new platform with their structured iKnowMed system using Azure Databricks for data ingestion and Azure Document Intelligence for text extraction, transforming manual chart review processes into an automated workflow that supports clinical decision-making and life science product development.

Zendesk integrated OpenAI's models to create adaptive AI service agents that autonomously manage customer conversations and execute resolution tasks. They implemented a multi-agent architecture featuring task identification, conversational RAG, procedure compilation, and procedure execution agents integrated with existing support workflows through API calls and natural language procedure definitions, while providing real-time chain-of-thought visibility. This solution transitions from traditional intent-based bots to a hybrid model of scripted and generative reasoning, streamlining customer service processes.

Chatbase integrated Anthropic's Claude into its multi-channel customer support platform to automate routine inquiries and deliver instant, personalized support across web chat, WhatsApp, Slack, and Instagram. They implemented the solution by embedding AI agents through APIs that integrate with business systems such as Stripe for order status checks and refund processing, while enabling customizable brand voice and multilingual communication with optional human oversight. This approach streamlined customer service workflows and provided advanced analytics to help teams quickly identify trends and optimize interactions.

Pensieve integrated Anthropic's Claude models (Claude 3.5 Sonnet and Claude 3.7 Sonnet) to build AI teaching assistants that automate grading and provide 24/7 personalized tutoring. The system clusters student submissions to calibrate grading rubrics and transforms static PDFs into interactive worksheets while enforcing instructor-controlled policies. This integration streamlines course management workflows and enhances real‐time classroom analytics.