A few months ago, Microsoft introduced Copilot Tuning, providing customers with a way to use low-code tools in Microsoft Copilot Studio to take advantage of highly automated fine-tuning “recipes” trained with enterprise data.
Generic Artificial Intelligence (GENAI) tools tend to relate to AI models that are trained on vast public information about the Internet and social media platforms, but companies need models that understand internal data and processes. Today, leading AI providers have instilled the power of Genai in the way business executives and IT chiefs think.
Such products aim to provide industry sector-specific AI systems compared to highly generalized models trained on freely available internet data. In theory, we should not suffer from hallucinations that plague more common AI models and closely match the way business works.
Ranveer Chandra, Vice President and Group Product Manager of Microsoft Experiences and Devices, was posted on his blog. [large language models] Also, searched augmented generation does not always understand your business in terms of specific processes, terms, and styles. “He argued that Microsoft's approach to optimizing AI models for business is to reduce the complexity that is often associated with tweaking projects.
One of the customers of Microsoft 365 Copilot (M365) Tuning's Early Access Program is accounting firm Ernst & Young. Marna Ricker, vice-chair of taxation at the company, said the company integrates corporate knowledge and expertise with LLM in the tax domain to provide a market-enhancing tax services through the M365. “This synergy has improved the quality of services and has significantly driven tax and legal research into which the relevant knowledge and intelligence is readily available in the M365 where people already work,” she added.
Gartner forecasts show that the market for specialized Genai models more than doubled to more than $2.5 billion by 2026. This is significantly smaller than Gartner's forecast for Gantner's $23 billion model, but it has shown that there is demand for such a technology business.
Roberta Cozza, senior director analyst at Gartner, said that the leading AI providers are tweaking the model as they are where companies are moving. Enterprise buyers said it's valuable to work with trusted technology providers, but they also need Genai-based tools to accommodate domain-specific things. “What we really see is a domain-specific model,” she said.
Cozza actually starts as a base from an open source model and often deploys as a small language model (SLM) that provides efficiency in terms of resource cost, but can be trained on the company's own data, providing better control.
While genai can provide value, the IT leaders of the companies she is talking to say they want to be trained with the issues, data and content of the specific industry they run. Microsoft and leading IT consulting providers are now increasing their AI business offerings to cater to companies that are currently seeking to provide business value with Genai. “They need to put their own data into the hands of a model builder or IT service provider,” Cozza said.
“The barriers to entry using basic open source models have been significantly reduced, and we have seen many small AI providers where large customers support large customers with their own small models,” she added. “They can distill their own models like ChatGpt, but many start with meta llamas and in Europe they see Mistral as their starting point.”
While 90% of the Genai model is managed by several major providers, Cozza said Gartner has submitted inquiries from IT decision makers who need to specifically deploy European AI technology.
“AI applications and technologies that are deemed at high risk are something that must regulate and comply with EU AI law,” she said. “But this covers frontier models trained with internet data.”
Cozza said models and explanationable SLMs built on internal corporate data are unlikely to require regulatory scrutiny. “Training AI and creating something more domain-specific actually improves general compliance because you can comply with policies and regulations,” she added.
Tools like the M365 Copilot Tuning will inevitably help lower the barriers to penetration for IT leaders tasked with providing Genai capabilities that can add business value, but SLM offers explanability and an alternative approach that can easily adapt to EU AI law.
