A wide range of Microsoft product updates for AI developers this week aims to help businesses stuck in an experimental phase with generative AI.
After 18 months of intense hype from major cloud providers, generative AI specialists and startups such as OpenAI and Anthropic as well as virtually every enterprise IT vendor, there is still only lukewarm signs of the actual use of technology by large companies. In fact, a recent survey from a research firm suggests that the so-called “trough of disillusionment” may already be arriving.
So far, more than half – 52% – of AI projects have failed to enter production, according to Gartner’s 2023 “AI in the Enterprise” survey of 703 people in management roles. management in companies in the United States, United Kingdom and Germany.
“There’s still a high degree of failure or discovery that it’s not as game-changing as organizations think, so they don’t act on it,” said Jason Wong, an analyst at Gartner. “It’s either finding out that they don’t really have the data quality to make it a full production product, or trying it and saying, ‘Will this be worth the cost?’ “
As during the first wave of promotion for AIOps Tools By leveraging other forms of AI and machine learning, many companies have discovered that the effectiveness of generative AI tools is highly dependent on the quality of IT automation practices and data management before these tools were applied, Wong said.
“People say, ‘Well, our people can’t find anything. Our search is broken. … So we’ll hand it over to (Microsoft) Copilot and that will make sense of the mess,'” Wong said. . “But you have to label your data, categorize it, classify your data – and if you had done that in the first place, you would have better research.”
Early examples of successful generative AI applications tend to be relatively simple projects that take an interactive approach while the technology is still rapidly evolving.
Microsoft strengthens ties with AI and guidance for developers
Microsoft introduced Fabric, a new version of a SaaS data management platform, to address data integration and analysis, during his Ignite conference in November. The Copilot chatbot within Fabric, intended to help users more easily create data pipelines, became generally available during Microsoft Build on Tuesday.
A preliminary feature called AI skills in Fabric is intended to help “analysts, creators, developers, and even those with minimal technical expertise… create intuitive AI experiences with data to unlock insights,” according to a release Microsoft press release.
Updates to AI development tools have followed suit, adding ease-of-use features and guardrails for software engineers creating AI applications. For example, Microsoft Commercial Enterprise IDE version 17.10, Visual Studioofficially integrates into a generally available and production supported integration with the GitHub Copilot code wizard.
This integration was previewed for several versions of Visual Studio and Microsoft’s free, lighter software. Visual Studio Code The publisher has long supported GitHub Copilot. But since Visual Studio Preview 3, “GitHub Copilot and Copilot Chat can be installed as a single extension combining Copilot and Copilot Chat into a single package…integrated and recommended by default in all workloads,” according to Microsoft Documentation.
Microsoft has also integrated GitHub Copilot with its existing Visual Studio code completion tool, IntelliSense. While some of the overlap between the two may initially be confusing, one IT professional said this combination could be particularly attractive to businesses.
“IntelliSense is able to find linked libraries, but now they use Copilot to add the variables into the library to complete the call,” said Nick Cassidy, chief innovation engineer at Blue Shield of California, who emphasized that his opinions do not reflect those of his employer. “If I were writing new code, I would probably use Copilot without IntelliSense. But (for) making changes to existing code, I could see it being useful.”
Integrating newer technologies into familiar tools could help AI developers understand more quickly how to use them, said Larry Carvalho, an independent analyst at RobustCloud, a cloud consulting firm.
“Developers will choose the IDE they are most familiar with, and Microsoft tools are popular with a wide segment of developers,” Carvalho said.
Taking context into account in Visual Studio’s GitHub Copilot integration will “keep Microsoft ahead of the market.” competition to help developers,” he said.
Likewise, a set of updates to Azure AI Studio unveiled this week could help a Microsoft customer overcome barriers to wider use of GenAI in production. These updates include updated reference architectures, landing zone accelerators, and service guides for the Azure OpenAI service; an overview of AI-as-a-service models; and new support for monitoring LLM application performance.
“AI adoption in many businesses still faces big challenges, from recruiting AI skills to defining an architecture on top of technologies that evolve or are replaced every quarter, or to find the hardware resources and partners necessary to support critical service levels,” said Nuno Guedes. , head of cloud computing at Millennium BCP, Portugal’s largest private bank, headquartered in Lisbon. “This year’s announcements demonstrate a comprehensive effort to lower these barriers to entry.”
Microsoft 365 users will also have access to preview versions of third-party extensions for tools like Atlassian Jira in Copilot Studio and Teams Toolkit from 365 for Visual Studio this week. This feature, Wong said, is notable for non-technical users and citizen developers who want to build custom chatbots.
“Based on our conversations with our customers, Copilot Studio is seeing very strong interest and following among organizations that have begun implementing Microsoft 365 Copilots. But the question is often: “How can we get more role-based data from other systems, other data sources, applications?” Some customers have tried this, but the full plugin approach is still a work in progress. »
IT pros sort through a glut of tools and ROI confusion
In addition to enterprise hesitancy towards generative AI due to data quality issues and concerns over securitydata confidentiality and governanceAt this point, there are simply too many generative AI development and infrastructure automation tools to choose from, Cassidy said.
The adoption of AI in many companies still faces major challenges, whether it is recruiting AI skills, defining an architecture based on technologies that evolve or are replaced every quarter, or even find the hardware resources and partners needed to support critical service levels.
Nuno GuédésHead of Cloud Computing, Millenium BCP
“What concerns me right now — and I think a lot of the industry is maybe starting to realize this — is how do I compare Watsonx to GitHub Copilot now that there is real competition in this market? It’s just background noise.”
Gartner clients testing GenAI products also find the in-depth proof-of-concept comparisons between the products confusing, especially when it comes to assessing their ROI, Wong said.
“We’re seeing companies look at this and say, ‘We need a return on investment here. We just can’t pay for all of this based on faith in the return,'” he said. declared.
AI development code assistants were likely adopted earlier by companies, in part because it is relatively easy for some companies to evaluate whether they improve the productivity of the developer team, Wong said. They can compare a team that uses them to a team that doesn’t use them.
For Cassidy, however, this isn’t necessarily simple.
“There are so many variations and factors that can factor into a developer’s performance that what I find is that it can be easier to just ask about the developer’s experience. developer: ‘Are you happier using a code wizard?'” he said. “But I can easily see junior developers relying a little too much on the code wizard and just assuming it’s okay.”
At Guedes’ company, a few generative AI projects other than code wizards have made it to production so far despite these challenges, he said. He attributes this to an approach that focuses on relatively small, short-term iterations to address specific user stories with a relatively well-defined ROI. Successful projects using this approach so far include improving access to internal information within product and customer support teams and multimodal conversions between content formats, such as speech to text, Sentiment analysis and quality assessment.
“Accepting that the current pace does not allow for long-term stable deployments, rather than trying to figure out what the next best solution will be, we are focusing on being iterative, testing it in the real world and to make decisions based on that.” he said.
Beth Pariseau, senior editor for TechTarget Editorial, is an award-winning veteran of IT journalism covering DevOps. Do you have any advice? Send him an email or contact @PariseauTT.
