Meet Agentic AI - Your New Digital Coworker

For a while, we’ve been seeing AI in its playful stage and it’s ready to move into creating real business value. It’s sort of how we once used to drink fake beers from iPhones and now we answer emails and order plane tickets from the same device.

But what does AI in the workforce look like? The answer is agentic AI – or AI agents – whatever you call them, they work for you. Think of them like your own personal assistant, that knows your company data, your work profile, and any other contextual information you need it to know.

But we’re getting ahead of ourselves, because what even is an agent?

With everything moving so quickly in the field of AI, it’s no wonder that there’s still some confusion regarding AI agents. The term agents, was chosen because it means a person or thing that takes an active role or produces a specified effect. And that’s what an AI agent is; it makes things happen for you and uses AI to do it.

The confusion happens because the one term covers quite a lot of technological and thematical variations. For example, you can have an ERP agent that helps you manage quotes and automatically turn them into sales orders. But if you use Microsoft Copilot studio, you have a much broader and more versatile toolset – letting you connext to data from other Microsoft Platforms, like Sharepoint, or connect data from 3rd party software vendors.

So, think of an agent as your digital coworker, that you train to perform specific tasks for you. These agents can be highly specialized and only do one task in a larger process (but do it like really, really well) or more generic with broader knowledge and more jobs to do.

There are a lot of options for building agents out there – but for business we love Microsoft Copilot studio: it’s easy to get started and it integrates with your existing Microsoft tools.

Depending on the use case, you can use AI in different ways, through automation or interactive dialogue-based agents.

Copilot Studio agents can help you with various work-related tasks through natural language dialogue. It’s the most common use of AI still, where you have conversation with OpenAI’s large language model (LLM). Initially designed as simple chatbots, these agents can understand and respond to natural language more effectively. This means you can interact with them using everyday language, and they will use the LLM to understand your requests and perform tasks for you.

For example, dialogue-based agents can answer questions about your technical documentation from SharePoint, data from your CRM or PPM systems stored in Dataverse, or use external data sources, or even public websites.

An autonomous agent is a system powered by AI that:

• Perceives its environment

• Reacts to changes and decides what to do based on goals

• Acts to acheive those goals

• Learns and adapts from the outcomes.

Autonomous agents can collaborate with other agents to handle even more complex workflows. The great thing about a well-designed autonomous agent is that they respond in real-time and operate across systems.

Let’s say, you run a professional services firm. In this instance you could design an autonomous agent that detects when a project is getting behind schedule in your PPM solution and then find and alert the relevant people on teams. It can even recommend a resource reallocation based on data available in your ERP solution and it’s up to you to say yes or no. Suddenly, a lot of stuff you had to monitor daily is now picked up when it’s relevant – allowing you to take action.

Before getting into the technical side of things, we always recommend starting with your internal business processes and making sure they’re in tip top shape (link to defining before designing). They’ll influence the type of agent you need, what it needs to know, and how you get your employees in the loop.

But if you’re ready to go, stay tuned for our next blog, where we give you the quick and dirty guide to getting started with Agentic AI.