In a move to accelerate enterprise AI adoption, Aisera, a Silicon Valley startup, is supercharging Microsoft’s Copilot with domain-specific intelligence.
By extending Copilot for Microsoft 365 to work across heterogeneous enterprise environments, Aisera is bridging the gap between general-purpose AI and the specialized needs of various business functions.
The Palo Alto, Calif., company’s GenAI platform is built on Azure AI, Microsoft 365, and Azure cloud, offering scalable and secure AI Copilot and AI Search solutions.
The collaboration with Microsoft allows for joint AI Copilot and AI Search solutions with higher accuracy, lower costs, and high security and privacy standards, said Aisera CEO Abhi Maheshwari.
“Aisera is an AI Copilot solution designed for enterprise-wide use, seamlessly integrating with diverse data lakes, applications, and services, including out-of-the-box connections to SaaS applications,” Maheshwari told The New Stack. “In contrast, Microsoft Copilot is specialized for Microsoft’s systems and applications.”
Meanwhile, Microsoft provides its Azure computing infrastructure and Azure OpenAI foundational models, which Aisera has fine-tuned for domain specificity.
“Aisera’s integration with Microsoft Copilot is designed to make generative AI readily accessible with security for enterprise-wide use cases, propelling diverse industries into a new era of innovation and productivity,” said Srini Raghavan, vice president of Microsoft Teams ecosystem, in a statement.
Domain-Specific LLMs
Moreover, the Aisera platform includes hundreds of integrations, thousands of AI workflows, and domain-specific large language models (LLMs) for various enterprise functions like sales, finance, HR, IT, and customer service.
A domain-specific LLM is a general model trained or fine-tuned to perform well-defined tasks dictated by organizational guidelines.
“Domain-specific Large Language Models are designed to address the limitations of Generic LLMs in specialized fields,” wrote Antonio Nucci, CTO of Aisera in a blog post. “Unlike their generic counterparts, which are trained on a wide array of text sources to develop a broad understanding applicable across multiple domains, a domain-specific LLM focuses on a particular area.”
Domain-specific LLMs often begin as Small Language Models that are subsequently expanded and refined through the process of Fine-Tuning LLM techniques, Nucci said in the post.
“This fine-tuning involves adjusting the model based on a concentrated dataset that is rich in the specific jargon, case studies, and scenarios pertinent to the domain in question,” Nucci wrote. “The goal is to equip the LLM with the ability to not only understand but also generate text that aligns with the professional standards and nuanced requirements of the field.”
AI Service Era
The name Aisera stands for “AI Service Era” and the company aims to optimize workflows and increase efficiency across enterprises, transforming operations for the AI era.
“They are one of a number of emerging AI service and solutions providers that wrap Copilot with their tools and services to customize the resulting AI solution, Rob Enderle, founder of the Enderle Group, told The New Stack about Aisera. “Generally this significantly improved the odds that the result will be successful.”
Historically, Aisera’s most important differentiators were its pre-built ontologies and libraries of human intents. These capabilities offered more human-sounding conversations than alternative offerings, as well as more accurate results, said Jason Bloomberg, an analyst at Intellyx. The Intellyx analyst and advisory firm has done some consulting work with Aisera in the past
“Today, the company has added domain-specific LLMs to the mix, further improving accuracy within its covered domains. Domain-specific LLMs are also smaller than general-purpose ones, thus requiring less infrastructure,” he said. “Its pre-built ontologies also support its dynamic knowledge graph, giving organizations more accurate search.”
Aisera’s enterprise-wide AI Search creates a dynamic knowledge graph for content, mapping relationships between people, content, and interactions. Aisera also provides a Universal Bot with a unified interface to resolve user requests across all domains.
“As part of the partnership, M365 Copilot can call into the Aisera AI Copilot for specific tasks across all supported domains,” Maheshwari told The New Stack. “Similarly, Aisera’s heterogeneous AI copilot can call into M365 copilot through a single pane of glass interface via its u-bot architecture.”
Aisera technology is multi-agentic. Its reasoning engine can perform intentness and intent-driven conversion and take actions across thousands of enterprise endpoints.
TRAPS
The company uses what it calls a TRAPS framework (Trusted, Responsible, Auditable, Private, and Secure) to ensure their Generative AI platforms and applications meet high standards.
Aisera’s products include AiseraGPT, AI Copilot, AI Search, and Agent Assist, all built on their Generative AI Platform. Their solutions are available on the Microsoft Marketplace for easy purchasing and deployment by mutual customers.
The post Domain-Specific AI: Aisera’s Answer to Enterprise Needs appeared first on The New Stack.
Aisera extends Microsoft's Copilot for enterprise-wide use, offering domain-specific LLMs and AI-powered solutions for various business functions.