Microsoft Brings AI Agents To The Mainstream

Summary:

• At its Ignite conference in Chicago, Microsoft announced they are bringing to life AI-powered agents. The company went all-in on driving home the concept that the creation, deployment and usage of agents is moving into the IT mainstream.
• Microsoft also described a new set of “agentic” features they called Copilot Actions, which CEO Satya Nadella described as akin to Outlook email rules extended across the entire M365 suite.
• For those people who want to create agents that go beyond the simple templates of the pre-built agents or Copilot actions, the company also offered up a range of new options.

As exciting and impressive as the Generative AI (GenAI)-powered, prompt-driven, chatbot experience may be, the technological leap that many people have been waiting for is the move to AI-powered agents that can start performing actions on our behalf without any prompting. It’s these kind of AI agents, in particular, that are expected to drive the kinds of big productivity gains that many believe is possible with AI.

At their Ignite conference in Chicago, that’s exactly what Microsoft (NASDAQ:MSFT) announced they are bringing to life. Both via their Copilot-powered office productivity apps as well as in several types of agent creation tools, the company clearly went all in on driving home the concept that the creation, deployment and usage of agents is moving into the IT mainstream.

From an end-user perspective, the company described several different means of helping people be more effective with their existing tools. New pre-built agent actions in Microsoft 365 Copilot, for example, allow agents to be used as meeting facilitators to ensure that Teams meetings stay on track or function as live translators for multi-language environments. In SharePoint, agents can be used to tap into corporate SharePoint resources for things like automatically summarizing any of the documents included within the SharePoint site.

Microsoft also described a new set of “agentic” features they called Copilot Actions, which CEO Satya Nadella described as akin to Outlook email rules extended across the entire M365 suite. You can use Copilot actions to, for example, organize a weekly project update email that automatically requests information from project participants, summarizes that data into a bulleted list of completed tasks, and then emails the results out to the participating team members.

For those people who want to create agents that go beyond the simple templates of the pre-built agents or Copilot actions, Microsoft offered up a range of new options. For sophisticated end-users and even low-code developers, the company talked about several enhancements to their Copilot Studio, including multimodal input, voice-based control and the ability to create agents that run autonomously. For more code-focused developers, the company debuted their Azure AI Agent Service tools – part of the company’s larger Azure AI foundry app development platform – which enable the creation of more sophisticated agents.

Collectively, the announcements provide an impressively wide range of tool “levels” to assist in agent creation for users of all types. What’s also interesting is that it clearly reflects Microsoft’s belief that there will be an enormous variety of agents with an equally diverse range of capabilities.

Of course, as cool as many of these new “agentic” capabilities sound, they bring with them several challenges that still need to be addressed. First and foremost, Microsoft still has the problem of too many Copilots. It’s still easy to get confused in trying to figure out which Copilot in which app or which environment does what. Adding in new AI-enabled agents – some of which are labelled with a Copilot tag and some of which are not – unfortunately, just makes this worse. Out-of-the-box agents in Microsoft Copilot 365 versus Copilot 365 actions, for example, both sound like they could do the same kind of things yet they have different names. I recognize some of this is just the growing pains of a new and rapidly evolving technology. However, given how much confusion already exists around what an AI agent is and what it does and what Copilot is and what Copilot does, it’s something Microsoft needs to tackle head-on. And not just one time – this is an ongoing issue that speaks to the industry-wide problem of insufficient end-user education on AI and its capabilities.

I’m also a bit concerned about the management of all these different agents. While it’s great that Microsoft is bringing the power of agent customization down to the individual user, in large organizations this could lead to an agent sprawl problem that could be tough to maintain. Microsoft is clearly aware of this and talked about a Copilot Control System, but I’m still concerned this could spiral out of control if it isn’t carefully and conscientiously monitored. I have little doubt we’ll see more management tools specifically tied to agents, as well as more standardized templates that should help limit the sprawl, but organizations which quickly and aggressively adopt agents need to be prepared for these potential issues.

Despite these concerns, it’s still great to see how quickly and broadly Microsoft is bringing agents into our everyday computing environments. The potential for well-constructed and well-managed agents to improve the productivity of nearly every type of worker really is enormous. The elimination or simplification of tedious tasks is exactly the kind of efficiency boost that everyone can appreciate. Plus, it’s these kind of capabilities that highlight what I believe is the ultimate promise of AI: making our devices smarter and having them do what we want them to do – even if we don’t always know how (or the best way) to achieve that.

As with many new technologies, there are bound to be some expected – and unexpected – challenges that come with the widespread deployment of AI agents. Ultimately, though, they will unquestionably open up yet another new stage of the AI revolution that will likely lead to an even bigger impact on our day-to-day work lives than what first-generation GenAI tools have already accomplished.

Disclaimer: Some of the author’s clients are vendors in the tech industry.

Disclosure: None.

Original Source: Author

Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.