When it comes to AI and integration, most companies are still experimenting, even if they claim otherwise — it’s still about tinkering with pilots, testing models or embedding AI into isolated tools. But according to Asmo Urpilainen, CTO at Frends, the real transformation begins when organizations stop adding AI to existing processes and start letting AI lead — something that will soon be available to customers.

“In the fully autonomous stage, you give AI the reins and allow AI to make decisions, rather than just use it to enhance existing solutions,” said Asmo in the recent Frends webinar on the future of Agentic AI. “You flip the script.”

Watch now the webinar “From Automation to Autonomy: What Agentic AI means for the future of Integration”:

From exploration to autonomy: A new maturity model

To help organizations understand where they are and where to go next, Frends has developed an Agentic AI Maturity Index, from the early exploration phase to fully autonomous integration.

“We see the vast majority of our customers still at the exploration stage,” Asmo explained. “They’ve started testing AI — maybe a few pilots or successful use cases —, but AI isn’t yet integrated into their core business processes.”

The Intelligent AI Connector, a new feature being introduced in the Frends platform, is designed to help users move from experimentation to integration, enabling simple yet powerful AI use cases to become part of everyday workflows.

“The goal is to enable that seamless transition… with as little effort and as much transparency as possible.”

AI in action: Smarter routing, faster decisions

The webinar featured a real-world example: an organization with a shared email inbox that handles messages for sales, IT, and support. Traditionally, routing these emails meant maintaining complex rule-based systems.

With the Frends AI Connector, this gets streamlined: “With a very simple prompt and a pre-selected model, we can categorize the email using AI,” said Asmo. “And even analyze the tone of voice to automatically set ticket priority.”

This kind of enhancement makes workflows faster and more resilient, with built-in logic to handle new scenarios, without constant human tweaking.

True autonomy: When AI takes the wheel

The real frontier, though, is Agentic AI: systems that don’t just follow instructions but actively make decisions and adapt their plans dynamically.

“You provide the AI with a target, allowing the AI to actually do the reasoning itself on what would be the best path to reach that goal,” Asmo said. “Instead of doing things in sequence.”

This approach is powered by what Frends calls the AI Toolbox, built on the MCP (Modal Context Protocol). It gives AI agents secure, governed access to business data and tools, so they can operate independently, while still being traceable and auditable.

“We need even better guardrails in place to ensure the AI doesn’t do something it’s not supposed to,” said Asmo. “That’s where Frends steps in.”

Watch the full webinar to explore how your organization can move from AI experimentation to true autonomy — securely, transparently, and with measurable value.