Powered By AI

3 min. read

Microsoft Fabric adheres to the definition of a modern data architecture for AI – the “glue” that manages data for improved efficiency powered by its own generative AI and provides the foundational AI models for practical business solutions.

A data-driven business culture or insight-driven organisation has been shown by institutions such as MIT, Harvard, and Deloitte to produce numerous and transformative benefits for businesses and organisations. As we’ve discussed in The Essentials of an AI Estate, Microsoft Fabric meets and exceeds the definition of a modern data architecture in terms of data ingestion, management, governance, analytics, reporting, and intelligence. This is powered by their sevenfold platform approach: Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Analytics, Data Activator, and Power BI.

Microsoft Fabric’s “ace in the hole” is its simplicity and ease of use without “dumbing down” the interface or capabilities for veteran data engineers or scientists to extract complex insights. This is powered, streamlined, and realised through their Generative Large Language Model (LLM) AI, Copilot.

In addition, Microsoft Azure AI services offer foundation models to innovate faster with off-the-shelf capabilities.

Using Copilot across the Fabric

Microsoft has infused the Copilot AI into every layer and workload across the fabric, creating a unified and seamless AI experience in every data platform. Microsoft says, “you can use natural language to create dataflows and pipelines, write SQL statements, build reports, or even develop machine learning models. And with our recently announced integration with Azure AI Studio, you can build generative AI solutions in Azure AI Studio tailor-made to your organization using data integrated, prepared, and modeled in Microsoft Fabric.”

Even if you have a passing familiarity with SQL – or no experience at all – one can build dataflows and reports without abstracting your stated goal into code, e.g., “Cross-reference all the sales from Q1 2022 with Q3 2023 and determine the growth in the 35-44 year old demographic” - if the unified lake has the data to support it, Copilot can recognise the statement, interpret it, and produce the desired result, which may have taken hours or days to implement using SQL or other types of code.

Copilot behind the scenes for data

The unified data lakehouse is built upon what’s known as semantic models which can be used across the Fabric. The AI, using machine learning, can sort and categorise data ingested into the Lakehouse with minimal, if any, user intervention. The AI can parse the different rows or objects of data to determine user-level access control, restricting permissions based on user access levels. It  produces replicas and other compute-saving efficiencies to speed up access and retrieval times in a non-destructive manner, which also saves money on compute units.

Using Azure AI as the foundation for AI-infused business solutions

Azure OpenAI Service, which powers Copilot, uses modern AI and large language model (LLM) that assist in complex cognitive tasks by making technology applications more intelligent, intuitive and accessible.

In addition to OpenAI, Microsoft offers a catalogue of frontier AI models that are trained on massive amounts of data to learn patterns and relationships. Other than a few organisations, training and optimising LLMs is cost prohibitive.

But using the trained LLMs from Microsoft as a foundation offers customers the perfect opportunity for infusing intelligence into task automation. For example, we take advantage of these models to implement practical and cost-effective AI solutions. First, we evaluate and select an LLM that is right for the business use case and need. Then, AI-infused solution is built, tested and deployed. This means that our customers see results from affordable AI in operation without having to wait a long time.

How we can help

Irada assists growth-minded organisations apply practical, aligned, and robust AI solutions in business operations.

References, Resources, Readings

1. https://hbr.org/webinar/2020/01/creating-a-data-driven-culture
2. https://sloanreview.mit.edu/mitsmr-connections/create-a-data-driven-culture-for-real-business-rewards/
3. https://www.deloitte.com/au/en/our-thinking/insights/topics/analytics/insight-driven-organization.html

Links to external websites were correct at the time of publishing. Articles may be behind a paywall. Irada is not responsible for the content of external websites.

The information in this article is general in nature. Your circumstances and needs may vary.

This work is licensed under CC BY-NC-SA 4.0