Microsoft’s AI Agents Want to Do Your Work for You—But Can They Be Trusted?

Microsoft's autonomous AI agents aim to automate routine business tasks, but will they stand out amid fierce competition in the AI landscape?

\ In a move to bolster business automation, Microsoft, late last year, introduced autonomous AI agents, its latest generative AI (genAI) based offering for the Dynamics 365 platform. The agents are engineered to handle complex tasks across sales, finance, service, and supply chains — automating key business processes without needing human intervention or coding expertise.

\ Microsoft asserted that these “apps for an AI-driven world” will surpass traditional AI offerings, reasoning over data and context to make judgments—something the company hopes will give them an edge over competitors like Salesforce, IBM, and Nvidia.

\ “Every organization will have a constellation of agents — ranging from simple prompt-and-response to fully autonomous. They will work on behalf of an individual, team, or function to execute and orchestrate business processes,” said Microsoft in a blog post.

\ Microsoft has positioned its AI agents as a more autonomous alternative to standard chatbots. Unlike chatbots, these agents are designed to function with minimal supervision, executing complex workflows across various sectors.

\ “LLMs are like the big, single-node monolithic systems used to perform multiple business functions, while AI agents are like independent microservices used to perform specialized tasks – both are important and relevant on a broader scale,” Keith Pijanowski, subject matter expert AI/ML at MinIO, told me. “Think of AI agents as the automated use of an organization’s existing models. If you want to build effective agents, you need effective models.”

\ Several companies, including McKinsey & Co., have already experimented with Microsoft's new AI agents. McKinsey created an agent capable of managing client inquiries, identifying the appropriate consultant, and scheduling follow-up meetings. According to early reports, these agents reduced lead times by 90% and cut administrative tasks by 30%.

\ “It may feel more magical for businesses that are vertically integrated on Microsoft stack. The challenges autonomous AI agents face is business use cases are stack fragmentation, API access for enterprise solutions, and lack of ML models for managing generic business functions,” Nish Krishna, Founder of Fractionalize, told me. “Within industry subsegments where Microsoft dominates solutions in the enterprise stack, Microsoft AI agents may be able to move the needle on adoption substantially.”

\ But while the company highlights the revolutionary potential, many businesses are still cautious about fully embracing them. A recent research from Gartner indicates that the majority of businesses haven’t progressed past the initial testing phases for Microsoft’s Copilot AI, leaving the adoption rate slower than expected.

\ “Microsoft and many companies are bullish and spending billions on building LLMs and AI applications, however, true leaders and success is yet to be seen,” Nitin Seth, Chief Executive Officer at SMS Magic, told me. “There are still big question marks around Hallucination, identifying the right inputs, and giving predictable results. Plus the question of bias and ethics remains as strong as ever. ”

\

Microsoft isn’t alone in this race. Salesforce recently introduced Agentforce, its own AI agents platform aimed at streamlining routine business processes. Unveiled at the Dreamforce 2024 event, Agentforce also allows companies to tailor AI agents to perform specialized functions across various departments—from sales and marketing to IT and customer service. Salesforce said that Agentforce will enable AI agents to analyze vast datasets, make informed decisions, and autonomously complete tasks with minimal human input. While Salesforce relies on its Atlas reasoning engine for Agentforce, Microsoft uses language models built in-house and by OpenAI.

\ In a recent X post, Salesforce CEO Marc Benioff slammed Microsoft’s Copilot calling it “Clippy 2.0,” the infamous 90s-era assistant, implying Microsoft’s agents might fall short in sophistication. “When you look at how Copilot has been delivered to customers, it’s disappointing. It just doesn’t work, and it doesn’t deliver any level of accuracy,” said Benioff.  “I have yet to find anyone who’s had a transformational experience with Microsoft Copilot or the pursuit of training and retraining custom LLMs. Copilot is more like Clippy 2.0.”

\ Likewise, IBM launched Granite 3.0 8B, its range of AI agent models, featuring support for advanced reasoning and a structured chat template that enables developers to implement complex tool use workflows. With rivals like Salesforce and IBM betting big on agentic AI-driven automation, Microsoft must prove that its solutions can drive tangible business value and stay ahead of the curve. The tech conglomerate is already under the microscope of investors, especially with slower-than-expected returns on its hefty AI investments and partnerships. The company's stock dipped 2.8% in the September quarter, due to a slower rate of Copilot AI adoption than Wall Street expected.

\ “The market signals suggest businesses are becoming more strategic, rather than hesitant, about AI adoption. The focus has shifted from 'Do we need AI?' to 'How do we implement AI effectively?',” Jim Palmer, Chief AI Officer at Dialpad, told me. “Businesses are now prioritizing solutions that solve specific business challenges and integrate with existing workflows while being mindful of both cost and complexity. The distinction isn't about whether to adopt AI - it's about choosing the right approach and timing to ensure successful implementation.”

\ The real test will come once these “agentic AI” tools are in the hands of more businesses. Will they fulfill the tech industry’s bold claims or become another Silicon Valley experiment that fails to catch on? The next few months will determine whether autonomous agents are truly revolutionary or just another tech gimmick destined to fade into the background.

\ “The biggest impact of agentic AI will come not from no code workflows, but from leveraging reinforcement learning with human intervention to drive a high level of accuracy,” Daniel Saks, co-founder of Landbase, told me. “Satya Nadella smartly espoused that most of the value creation in any tech wave comes from a head start in the first 2-3 years of the emergence of the technology. That means Microsoft has a few years to invest and win the market.”

\n