The Rise of AI Agents: Adoption, Challenges, and Future Prospects

AI Agents are Everywhere, But No One Knows Why

Companies are going all in on AI agents, but are they skewing the definition of one? Software framework LangChain recently published a report surveying over 1,300 professionals, to "learn about the state of AI agents" in 2024. While 51% of the respondents said they have already been using AI agents in production, 63% of mid-sized companies deployed agents in production, and 78% have active plans to integrate AI agents. Furthermore, the survey also revealed that professionals in non-technical companies are also willing to deploy AI agents. It stated, "90% of respondents working in non-tech companies have or are planning to put agents in production (nearly equivalent to tech companies, at 89%)." Even Research and Market's report on 'AI Agents Market Analysis' indicates an optimistic future for AI agents. "The AI agents market is projected to grow from $5.1 billion in 2024 to $47.1 billion in 2030, with a CAGR of 44.8% during 2024-2030," read the report. The numbers indicate a resounding shift in sentiment towards AI agents, moving away from the prohibitive scepticism. While a majority of the respondents in Langchain's survey revealed using AI agents for research summarization and personal assistance, a notable 35% said they use them for coding tasks. Companies, however, haven't settled on a definition for AI agents yet. A spectrum, or absolute autonomy? Earlier, Google stated that 25% of all newly written code is AI-generated, a revelation that sparked criticism. For instance, a user on HackerNews suggested Google's claim might be overstated, arguing that it primarily relies on a code-completion engine. Meanwhile, a Reddit user observed that Google was indicating "clean-up jobs for dependencies, removing deprecated classes, or changing deployment configurations". A few days ago, payment processing giant Stripe launched a software development kit (SDK) for AI agents. This SDK allowed LLMs to call functions related to payments, billing, and issuing, enabling agents to 'spend' funds and accept or decline payment authorisations. Several users on X questioned the reality of the feature and asked if it was just a fancier way to refer to API and function calls. "I mean, for me, at least, this just removes ten lines of code and proposes a more complex pricing model. At the end of the day, am I missing something?" said a user on X. At Oracle CloudWorld in 2024, the company announced over 50 AI agents in the Fusion Cloud Application suite. Oracle's executive vice president of applications development, Steve Miranda, however, was quite transparent about the definition of an agent. During an interaction with AIM, he said, "I think that early use cases will be a little bit less completely autonomous and more human-assisted." Similarly, Ketan Karkhanis, CEO of ThoughtSpot, while talking to AIM, explained that many systems today, such as Microsoft's Copilot, operate on single-turn Q&A, answering one question at a time. They lack reasoning, adaptability, and the ability to learn a user's business to be called autonomous. "There are a lot of nuances to this. If you can't coach it, then it's not an agent. I don't think you can coach a copilot. You can write custom prompts [but] that's not coaching," Karkhanis added. Even Salesforce CEO Marc Benioff has often criticised Microsoft's approach towards AI agents and accused them of falsely marketing Copilot's capabilities. While there isn't a universally accepted definition yet, companies are claiming an improvement in several operations with the use of AI agents. The survey received criticism on social media. A user on X posted, "In this day and age, surveys are the worst indication of real usage. Show us actual real usage tracking metrics that you can collect." Despite their skewed definitions, several organisations, including some of the biggest names, are achieving success with AI agents. A few weeks ago, Freshworks unveiled a new version of Freddy AI, an autonomous agent that resolved 45% of customer support requests and 40% of IT services (on beta). Even Salesforce announced the availability of Agentforce, which enabled their customers to deploy AI agents on their platform. One of Salesforce's customers, publishing company Wiley, reported a notable success with Agentforce. "With the help of AI productivity tools, Wiley was able to onboard seasonal agents 50% faster, leading to a 213% return on investment and $230,000 in savings," said Wiley wrote a blog post. Wiley also mentioned that Agentforce showed a 40% improvement in customer case resolution compared to their previous chatbot. This is in line with LangChain's survey, in which 45.8% of the participants mentioned deploying AI agents in customer support and service. Salesforce continues to remain bullish over an agentic future. "In 2025, we'll increasingly see more complex, multi-agent orchestrations solving higher-order challenges across the enterprise, like simulating new product launches or marketing campaigns and developing recommendations for adjustments," said Mick Costigan, VP of Salesforce Futures. Moreover, companies who have actively deployed AI agents, continue to improve accuracy and reduce operational costs. Amdocs, a telecommunications company, built AI agents using NVIDIA's NIM Microservices, which increased AI accuracy by 30%. Further, Amdocs reported a notable decrease in operational costs, by reducing token usage by 60% for data preprocessing and by up to 40% for inferencing. Contrary to the popular definition of AI agents operating autonomously, there's a good reason why they aren't. In LangChain's survey, a majority of respondents expressed the need for 'tracing and observability' to oversee autonomous operations. More than 35% of companies prioritise online or offline evaluation of the agents' output results. Most of the surveyed companies granted agents read-only permissions, and very few, around 10% of the companies, granted agents full read, write, and delete permissions. Even if risks and concerns are alleviated, AI agents may not fully understand the nuances of each aspect of the operation. During a conversation with AIM, Sam Mantle, CEO of Lingaro Group, stressed the importance of handling the flow of data between each individual component in an operation, which is often disconnected. "I'm interested in [knowing] who owns the data component that may sit in that application, because if we really want to streamline things, somebody has to be responsible for that data, no matter where it flows within the organisation," Mantle further said.

Microsoft quietly assembles the largest AI agent ecosystem -- and no one else is close

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Microsoft has quietly built the largest enterprise AI agent ecosystem, with over 100,000 organizations creating or editing AI agents through its Copilot Studio since launch - a milestone that positions the company ahead in one of enterprise tech's most closely watched and exciting segments. "That's a lot faster than we thought, and it's a lot faster than any other kind of cutting edge technology we've released," Charles Lamanna, Microsoft's executive responsible for the company's agent vision, told VentureBeat. "And that was like a 2x growth in just a quarter." The rapid adoption comes as Microsoft significantly expands its agent capabilities. At its Ignite conference starting today, the company announced it will allow enterprises to use any of the 1,800 large language models (LLMs) in the Azure catalog within these agents - a significant move beyond its exclusive reliance on OpenAI's models. The company also unveiled autonomous agents that can work independently, detecting events and orchestrating complex workflows with minimal human oversight. (See our full coverage of the Microsoft's agent announcements here.) These AI agents - software that can reason and perform specific business tasks using generative AI - are emerging as a powerful tool for enterprise automation and productivity. Microsoft's platform enables organizations to build these agents for tasks ranging from customer service to complex business process automation, while maintaining enterprise-grade security and governance. Building an enterprise-grade foundation Microsoft's early lead in AI agents stems from its focus on enterprise requirements that often get overlooked in the AI hype cycle. While its new autonomous agents and LLM flexibility grabbed headlines at Ignite, the company's real advantage lies in its enterprise infrastructure. The platform integrates with over 1,400 enterprise systems and data sources, from SAP to ServiceNow to SQL databases. This extensive connectivity lets organizations build agents that can access and act on data across their existing IT landscape. While enterprises can build custom agents from scratch, Microsoft has also launched ten pre-built autonomous agents targeting core business functions like sales, service, finance, and supply chain - to accelerate adoption for common enterprise use cases. The company did not provide any more detail about which types of agents customers are finding the most popular. But Lamanna said that aside from apps that IT departments are building for specific core tasks, there was a second category of apps that is more bottoms-up. This is where employees create Copilot agents to share their documents or presentations with their team or other partners, so that others can interact with the content and ask questions about it. Security and governance features, often afterthoughts in AI deployments, are built into Microsoft's core architecture. The platform's control system ensures agents operate within enterprise permissions and data governance frameworks. "We think it will show up everywhere," Lamanna told VentureBeat, "because whenever you have a technology that makes something possible that was previously impossible, all of you kind of are always shocked by how broadly it ends up being used." He compared it with the Internet, where connectivity extended from the browser to the operating system, and fundamentally changed client-server architecture. The LLM made a big breakthrough, Lamanna explains, in that it understands unstructured content - language or video or audio - and has shown the beginnings of reasoning, to make conclusions or judgments based on this data, Lamanna said. "So the browser, word processor, the core operating system experience, and the way you do sales processes and customer support processes - they all have to be reevaluated now that this capability exists...I don't think there'll be really any part of the stack in computing that doesn't have some component reimagined as a result of all the agent and AI capabilities." Early adopters are already seeing results. McKinsey reduced its project intake workflows from 20 days to just 2 days using automated routing agents. Pets at Home deployed fraud prevention agents in under two weeks, saving millions annually. Other companies using Copilot Studio include Nsure, McKinsey, Standard Bank, Thomson Reuters, Virgin Money, Clifford Chance and Zurich, Microsoft told VentureBeat. The Agent mesh: Microsoft's vision for enterprise AI At the heart of Microsoft's strategy is what Lamanna calls the "agent mesh" - an interconnected system where AI agents collaborate to solve complex problems. Rather than operating in isolation, agents can pass tasks, messages, and knowledge seamlessly across the enterprise. Copilot Studio has been associated so far with agents that are triggered via chat, but now Microsoft is emphasizing any kind of actions. Imagine an enterprise where agents collaborate seamlessly: A sales agent triggers an inventory agent to check stock availability, which then notifies a customer service agent to update the client. This architecture includes: Microsoft's research arm recently released the Magnetic-One system based on the company's Autogen framework, which establishes a sophisticated agent hierarchy: a managing agent maintains task checklists in an "outer loop" while specialized agents execute work in an "inner loop." This architecture could potentially soon embrace tools like Microsoft's OmniParser that let agents interpret UI elements, and showcases Microsoft's technical lead in computer-using agents -- matching capabilities being developed by Anthropic and Google. The company said it is working to bring this research into production, but did not specify how and when. Microsoft's approach addresses a key enterprise challenge: scaling from hundreds to potentially millions of agents while maintaining control. The platform enables companies to coordinate multiple specialized agents through its orchestration capabilities - an approach that aligns with a broader industry trend toward multi-agent systems. The platform's pricing model reflects this enterprise focus. Rather than charging per token like most AI providers, Microsoft Copilot studio prices based on the number of messages exchanged - emphasizing business outcomes over raw compute. Companies are no longer asking about which model is best, Lamanna explained. They're asking for examples of business value. "That has been a remarkable shift in the market." The race for enterprise AI agents While other tech giants are investing heavily in AI agents, Microsoft's combination of enterprise features and extensive integrations gives it an early advantage. Competitors like Salesforce and ServiceNow have introduced their own AI agent platforms, such as Agentforce (which boasted 10,000 agents built) and ServiceNow Agents, but these offerings are relatively new and lack Microsoft's established enterprise reach: Hundreds of millions of workers use Microsoft's productivity suite. The landscape includes various approaches. OpenAI focuses on direct API access but hasn't yet built an enterprise AI agent deployment framework, though its recent o1-preview model shows superior reasoning capabilities that could power more intelligent agents in the future. New entrants like Crew offer experimental agentic frameworks but lack enterprise scale. LangChain's modular framework remains popular among developers but focuses more on experimentation than enterprise-grade deployment. AWS maintains a developer-focused approach through platforms like SageMaker, while Google's AI platforms show strength in specific verticals but lack a cohesive agent framework for broad enterprise adoption. By contrast, Microsoft combines enterprise security, low-code tools, pre-built templates, and pro-code SDKs for developers, making it a more inclusive option for diverse teams. However, the technology remains nascent. Large language models can still hallucinate, and AI agents that rely on them need careful installation and management to avoid issues like infinite loops or unnecessary costs. Some customers have expressed concerns about Copilot's pricing and implementation challenges. The field is also likely to remain fragmented. A significant subset of Fortune 500 companies may opt for multi-vendor approaches, potentially using Microsoft's Copilot agents for employee productivity while choosing other frameworks for sensitive applications. Conclusion: Leading the agent-driven enterprise While Microsoft leads in enterprise AI agent deployment today, the technology remains in early stages. The company's advantage stems not from any single feature but from its comprehensive approach: enterprise-grade infrastructure, extensive integrations, and focus on business outcomes rather than raw AI capabilities. The coming year will test whether Microsoft can maintain this lead. Competitors are racing to enhance their offerings. Enterprises are moving from experimentation to full deployment. What's clear is that AI agents are moving beyond the hype cycle into the reality of enterprise IT architecture - with all the complexity and challenges that transition entails. For technical leaders, now is the time to evaluate how AI agents can transform your workflows, from automating repetitive tasks to enabling new modes of collaboration. Start small, focus on measurable outcomes, and consider pre-built agents first to accelerate your journey. Watch the full interview with Charles Lamanna embedded above to hear firsthand how Microsoft is driving the AI revolution, what AI agents mean for enterprise architecture, the rise of ContentOps, and how its affecting roles and job functions.

The new paradigm: Architecting the data stack for AI agents

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More The launch of ChatGPT two years ago was nothing less than a watershed moment in AI research. It gave a new meaning to consumer-facing AI and spurred enterprises to explore how they could leverage GPT or similar models into their respective business use cases. Fast-forward to 2024: there's a flourishing ecosystem of language models, which both nimble startups and large enterprises are leveraging in conjunction with approaches like retrieval augmented generation (RAG) for internal copilots and knowledge search systems. The use cases have grown multifold and so has the investment in enterprise-grade gen AI initiatives. After all, the technology is expected to add $2.6 trillion to $4.4 trillion annually to the global economy. But, here's the thing: what we have seen so far is only the first wave of gen AI. Over the last few months, multiple startups and large-scale organizations - like Salesforce and SAP - have started moving to the next phase of so-called "agentic systems." These agents transition enterprise AI from a prompt-based system capable of leveraging internal knowledge (via RAG) and answering business-critical questions to an autonomous, task-oriented entity. They can make decisions based on a given situation or set of instructions, create a step-by-step action plan and then execute that plan within digital environments on the fly by using online tools, APIs, etc. The transition to AI agents marks a major shift from the automation we know and can easily give enterprises an army of ready-to-deploy virtual coworkers that could handle tasks - be it booking a ticket or moving data from one database to another - and save a significant amount of time. Gartner estimates that by 2028, 33% of enterprise software applications will include AI agents, up from less than 1% at present, enabling 15% of day-to-day work decisions to be made autonomously. But, if AI agents are on track to be such a big deal? How does an enterprise bring them to its technology stack, without compromising on accuracy? No one wants an AI-driven system that fails to understand the nuances of the business (or specific domain) and ends up executing incorrect actions. The answer, as Google Cloud's VP and GM of data analytics Gerrit Kazmaier puts it, lies in a carefully crafted data strategy. "The data pipeline must evolve from a system for storing and processing data to a 'system for creating knowledge and understanding'. This requires a shift in focus from simply collecting data to curating, enriching and organizing it in a way that empowers LLMs to function as trusted and insightful business partners," Kazmaier told VentureBeat. Building the data pipeline for AI agents Historically, businesses heavily relied on structured data - organized in the form of tables - for analysis and decision-making. It was the easily accessible 10% of the actual data they had. The remaining 90% was "dark," stored across siloes in varied formats like PDFs and videos. However, when AI sprung into action, this untapped, unstructured data became an instant value store, allowing organizations to power a variety of use cases, including generative AI applications like chatbots and search systems. Most organizations today already have at least one data platform (many with vector database capabilities) in place to collate all structured and unstructured data in one place for powering downstream applications. The rise of LLM-powered AI agents marks the addition of another such application in this ecosystem. So, in essence, a lot of things remain unchanged. Teams don't have to set up their data stack from scratch but adapt it with a focus on certain key elements to make sure that the agents they develop understand the nuances of their business industry, the intricate relationships within their datasets and the specific semantic language of their operations. According to Kazmaier, the ideal way to make that happen is by understanding that data, AI models and the value they deliver (the agents) are part of the same value chain and need to be built up holistically. This means going for a unified platform that brings together all the data - from text and images to audio and video - to one place and has a semantic layer, utilizing dynamic knowledge graphs to capture evolving relationships, in place to capture the relevant business metrics/logic required for building AI agents that understand the organization and domain-specific contexts for taking action. "A crucial element for building truly intelligent AI agents is a robust semantic layer. It's like giving these agents a dictionary and a thesaurus, allowing them to understand not just the data itself, but the meaning and relationships behind it...Bringing this semantic layer directly into the data cloud, as we're doing with LookML and BigQuery, can be a game-changer," he explained. While organizations can go with manual approaches to generating business semantics and creating this crucial layer of intelligence, Gerrit notes the process can easily be automated with the help of AI. "This is where the magic truly happens. By combining these rich semantics with how the enterprise has been using its data and other contextual signals in a dynamic knowledge graph, we can create a continuously adaptive and agile intelligent network. It's like a living knowledge base that evolves in real-time, powering new AI-driven applications and unlocking unprecedented levels of insight and automation," he explained. But, training LLMs powering agents on the semantic layer (contextual learning) is just one piece of the puzzle. The AI agent should also understand how things really work in the digital environment in question, covering aspects that are not always documented or captured in data. This is where building observability and strong reinforcement loops come in handy, according to Gevorg Karapetyan, the CTO and co-founder of AI agent startup Hercules AI. Speaking with VentureBeat at WCIT 2024, Karapetyan said they are taking this exact approach to breach the last mile with AI agents for their customers. "We first do contextual fine-tuning, based on personalized client data and synthetic data, so that the agent can have the base of general and domain knowledge. Then, based on how it starts to work and interact with its respective environment (historical data), we further improve it. This way, they learn to deal with dynamic conditions rather than a perfect world," he explained. Data quality, governance and security remain as important With the semantic layer and historical data-based reinforcement loop in place, organizations can power strong agentic AI systems. However, it's important to note that building a data stack this way does not mean downplaying the usual best practices. This essentially means that the platform being used should ingest and process data in real-time from all major sources (empowering agents to adapt, learn and act instantaneously according to the situation), have systems in place for ensuring the quality/richness of the data and then have robust access, governance and security policies in place to ensure responsible agent use. "Governance, access control, and data quality actually become more important in the age of AI agents. The tools to determine what services have access to what data become the method for ensuring that AI systems behave in compliance with the rules of data privacy. Data quality, meanwhile, determines how well (or how poorly) an agent can perform a task," Naveen Rao, VP of AI at Databricks, told VentureBeat. He said missing out on these fronts in any way could prove "disastrous" for both the enterprise's reputation as well as its end customers. "No agent, no matter how high the quality or impressive the results, should see the light of day if the developers don't have confidence that only the right people can access the right information/AI capability. This is why we started with the governance layer with Unity Catalog and have built our AI stack on top of that," Rao emphasized. Google Cloud, on its part, is using AI to handle some of the manual work that has to go into data pipelines. For instance, the company is using intelligent data agents to help teams quickly discover, cleanse and prepare their data for AI, breaking down data silos and ensuring quality and consistency. "By embedding AI directly into the data infrastructure, we can empower businesses to unlock the true potential of generative AI and accelerate their data innovation," Kazmaier said. That said, while the rise of AI agents represents a transformative shift in how enterprises can leverage automation and intelligence to streamline operations, the success of these projects will directly depend on a well-architected data stack. As organizations evolve their data strategies, those prioritizing seamless integration of a semantic layer with a specific focus on data quality, accessibility, governance and security be best positioned to unlock the full potential of AI agents and lead the next wave of enterprise innovation. In the long run, these efforts, combined with the advances in the underlying language models, are expected to mark nearly 45% growth for the AI agent market, propelling it from $5.1 billion in 2024 to $47.1 billion by 2030.

How to get started with AI agents (and do it right)

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Due to the fast-moving nature of AI and fear of missing out (FOMO), generative AI initiatives are often top-down driven, and enterprise leaders can tend to get overly excited about the groundbreaking technology. But when companies rush to build and deploy, they often deal with all the typical issues that occur with other technology implementations. AI is complex and requires specialized expertise, meaning some organizations quickly get in over their heads. In fact, Forrester predicts that nearly three-quarters of organizations that attempt to build AI agents in-house will fail. "The challenge is that these architectures are convoluted, requiring multiple models, advanced RAG (retrieval augmented generation) stacks, advanced data architectures and specialized expertise," write Forrester analysts Jayesh Chaurasia and Sudha Maheshwari. So how can enterprises choose when to adopt third-party models, open source tools or build custom, in-house fine-tuned models? Experts weigh in. AI architecture is far more complex than enterprises think Organizations that attempt to build agents on their own often struggle with retrieval augmented generation (RAG) and vector databases, Forrester senior analyst Rowan Curran told VentureBeat. It can be a challenge to get accurate outputs in expected time frames, and organizations don't always understand the process -- or importance of -- re-ranking, which helps ensure that the model is working with the highest quality data. For instance, a user might input 10,000 documents and the model may return the 100 most relevant to the task at hand, Curran pointed out. But short context windows limit what can be fed in for re-ranking. So, for instance, a human user may have to make a judgment call and choose 10 documents, thus reducing model accuracy. Curran noted that RAG systems may take 6 to 8 weeks to build and optimize. For example, the first iteration may have a 55% accuracy rate before any tweaking; the second release may have 70% and the final deployment will ideally get closer to 100%. Developers need to have an understanding of data availability (and quality) and how to re-rank, iterate, evaluate and ground a model (that is, match model outputs to relevant, verifiable sources). Additionally, turning the temperature up or down determines how creative a model will be -- but some organizations are "really tight" with creativity, thus constraining things, said Curran. "There's been a perception that there's an easy button around this stuff," he noted. "There just really isn't." A lot of human effort is required to build AI systems, said Curran, emphasizing the importance of testing, validation and ongoing support. This all requires dedicated resources. "It can be complex to get an AI agent successfully deployed," agreed Naveen Rao, VP of AI at Databricks and founder and former CEO of MosaicAI. Enterprises need access to various large language models (LLMs) and also have the ability to govern and monitor not only agents and models but underlying data and tools. "This is not a simple problem, and as time goes on there will be ever-increasing scrutiny over what and how data is being accessed by AI systems." Factors to consider when exploring AI agents When looking at options for deploying AI agents -- third party, open source or custom -- enterprises should take a controlled, tactical approach, experts advise. Start by considering several important questions and factors, recommended Andreas Welsch, founder and chief AI strategist at consulting company Intelligence Briefing. These include: It's also important to factor in existing licenses and subscriptions, Welsch pointed out. Talk to software sales reps to understand whether your enterprise already has access to agent capabilities, and if so, what it would take to use them (such as add-ons or higher tier subscriptions). From there, look for opportunities in one business function. For instance: "Where does your team spend time on several manual steps that can not be described in code?" Later, when exploring agents, learn about their potential and "triage" any gaps. Also, be sure to enable and educate teams by showing them how agents can help with their work. "And don't be afraid to mention the agents' limitations as well," said Welsch. "This will help you manage expectations." Build a strategy, take a cross-functional approach When developing an enterprise AI strategy, it is important to take a cross-functional approach, Curran emphasized. Successful organizations involve several departments in this process, including business leadership, software development and data science teams, user experience managers and others. Build a roadmap based on the business' core principles and objectives, he advised. "What are our goals as an organization and how will AI allow us to achieve those goals?" It can be difficult, no doubt because the technology is moving so fast, Curran acknowledged. "There's not a set of best practices, frameworks," he said. Not many developers have experience with post-release integrations and DevOps when it comes to AI agents. "The skills to build these things haven't really been developed and quantified in a broad-based way." As a result, organizations struggle to get AI projects (of all kinds) off the ground, and many eventually switch to a consultancy or one of their existing tech vendors that have the resources and capability to build on top of their tech stacks. Ultimately, organizations will be most successful when they work closely with their partners. "Third-party providers will likely have the bandwidth to keep up with the latest technologies and architecture to build this," said Curran. That's not to say that it's impossible to build custom agents in-house; quite the contrary, he noted. For instance, if an enterprise has a robust internal development team and RAG and machine learning (ML) architecture, they can use that to create their own agentic AI. This also goes if "you have your data well governed, documented and tagged" and don't have a "giant mess" of an API strategy, he emphasized. Whatever the case, enterprises must factor ongoing, post-deployment needs into their AI strategies from the very beginning. "There is no free lunch post-deployment," said Curran. "All of these systems require some type of post launch maintenance and support, ongoing tweaking and adjustment to keep them accurate and make them more accurate over time."

Microsoft's new AI agents support 1,800 models (and counting)

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More AI agents are the talk of the enterprise right now. But, business leaders want to hear about tangible results and relevant use cases -- as opposed to futuristic, not-quite-there-yet scenarios -- and demand tools that are easy to deploy and use and, further, that support their preferred model(s). Microsoft claims to have all these concerns covered with new no-code and low-code capabilities in Microsoft 365 Copilot. Today at Microsoft Ignite, the tech giant announced that users can now build their own custom autonomous agents or deploy out-of-the-box, purpose-built agents. And, they can do this via a bring-your-own setup that provides them access to the 1,800-plus models in the Azure AI catalog. (See our separate story today about how Microsoft has quietly assembled the largest AI agent ecosystem -- and no one else is close). "Companies have done a lot of AI exploration and really want to be able to measure and understand how agents can help them be more efficient, improve performance and decrease cost and risk," Lili Cheng, corporate VP of the Microsoft AI and research division, told VentureBeat. "They're really leaning into scaling out their copilots." Supporting bring-your-own-knowledge, bring-your-own-model According to IDC, in the next 24 months, more and more companies will build custom, tailored AI tools. Indeed, vendors -- from tech giants such Salesforce and Snowflake to smaller players like CrewAI and Sema4.ai -- are increasingly pushing platforms to market that promise to revolutionize enterprise operations. Microsoft introduced Copilot in February 2023, and has now infused it with a suite of new capabilities to support agentic AI. Autonomous capabilities now in public preview allow users to build agents that act on their behalf without additional prompting. This means agents can work and act in the background without human oversight. Users can use templates for common scenarios (such as sales order and deal accelerator agents) in Copilot Studio. Or, more advanced developers can take advantage of a new Agent SDK (now available in preview) to build full-stack, multichannel agents that integrate with various Microsoft services and can be deployed across Microsoft, third-party and web channels. New integrations with Azure AI Foundry will support bring-your-own-knowledge (custom search indices can be added as a knowledge source) (now in preview) and bring-your-own- model (now in private preview). This will allow users to pull from the 1,800-some-odd models (and counting) in Azure's catalog. This element is critical, as users are demanding the ability to securely use proprietary data and combine and test different models without getting locked in to one or the other. "People want a variety of models, they want to be able to fine-tune models," said Cheng. Ready-made agents for HR, translation, project management But not all tasks require a custom solution; already-built models can be useful across enterprises. Microsoft is releasing several ready-made agents in Copilot that can handle simple, repetitive tasks or more complex multi-step processes. These include: Further, a new Azure AI Foundry SDK offers a simplified coding experience and toolchain for developers to customize, test, deploy and manage agents. Users can choose from 25 pre-built templates, integrate Azure AI into their apps and access common tools including GitHub or Copilot Studio. Cheng pointed to the importance of low-code and no-code tools, as enterprises want to accommodate teams with a range of skills. "Most companies don't have big AI teams or even development teams," she said. "They want more people to be able to author their copilots." The goal is to greatly simplify the agent-building process so that enterprises "build something once and use it wherever their customers are," she said. Tooling should be simple and easy to use so that app creators don't even know if things are getting ever more complicated on the back end. Cheng posited: "Something might be more difficult, but you don't know it's more difficult, you just want to get your job done." McKinsey, Thomson Reuters use cases Initial use cases have revolved around support, such as managing IT help desks, as well as HR scenarios including onboarding, said Cheng. McKinsey & Company, for its part, is working with Microsoft on an agent that will speed up client onboarding. A pilot showed that lead time could be reduced by 90% and administrative work by 30%. The agent can identify expert capabilities and staffing teams and serves as a platform for colleagues to ask questions and request follow-ups. Meanwhile, Thomson Reuters built an agent to help make the legal due diligence process -- which requires significant expertise and specialized content -- more efficient. The platform combines knowledge, skills and advanced reasoning from the firm's gen AI tool CoCounsel to help lawyers close deals more quickly and efficiently. Early tests indicate that several tasks in these workflows could be cut by at least 50%. "We really see people combining more traditional copilots -- where you have AI augmenting people skills and providing personal assistance -- together with autonomous systems," said Cheng. Agents are increasingly authoring processes and workflows and working across groups of people and in multi-agent systems, she noted. AI agents aren't new (but using them on top of LLMs is) While they may be all the talk now, agents are not new, Microsoft Source writer Susanna Ray emphasizes in a blog post out today. "They're getting more attention now because recent advances in large language models (LLMs) help anyone -- even outside the developer community -- communicate with AI," she writes. Agents serve as a layer on top of LLMs, observing and collecting information and providing input so that together they can generate recommendations for humans or, if permitted, act on their own. "That agent-LLM duo makes AI tools more tangibly useful," Ray notes, adding that agents will become even more useful and autonomous with ongoing innovations with memory, entitlements and tools. Cheng pointed out that Microsoft began talking about conversational AI about eight years ago. Before AI agents, conversation data "was always kind of lost and siloed." Now, agentic AI can bring intelligence to users and provide context in real time. "People just want that tooling to be more natural," she said. "It's phenomenal that we can do a lot of these things that we dreamed about. Being able to combine all these sources effortlessly is really groundbreaking."